Chapter I

1. Introduction

Development of sustainable agricultural systems such as alley farming is an

attempt to reduce degradation of natural resources and to find environmentally
compatible ways of increasing production and promoting broad-scale development.

Intensification of agriculture on land currently used for traditional farming

requires a thorough knowledge of the soil as a resource and attributes of the land.
Information on distribution, potential and constraints of major soils is needed, so that the
most appropriate soil management systems can be designed. In addition knowledge on
land capability and suitability is also essential to determine the best land use for sustained
crop production.

Soil, water and plant testing is an indispensable tool for research, advisory
services and formulation of rational fertilizer recommendations, as well as designing
appropriate soil management and agronomic practices. Soil testing in particular is also
important in monitoring the various types of land degradation and the choice of measures
for land improvement.

Development of effective and efficient analytical services of soil, water, plant and
fertilizers, therefore, is an important means for increasing and sustaining land
productivity as well as crop and food production; a challenge for many developing
countries. These analytical services should be closely linked to the extension and
advisory services and should maintain functional and technical relationships with
universities and research institutions.

FAO and other organizations have assisted developing countries in establishing

and improving their soil and water laboratories (SWL) and analytical services, through
mobilization of funds (projects), supply of laboratory equipment, standardization of

1
analytical procedures, exchange of information and experience, training of human
resources and enhancing the capacity for data interpretation.
FAO has played a key role in assisting member countries in the establishment and
improvement of the functions and impact of soil and water laboratories for agricultural
development and to boost food production. Among others, FAOs efforts in this field have
been devoted to human resources development of the soil and water laboratories. The
assistance of FAO, particularly for the establishment of laboratories, has mainly been
given through specific projects (extra-budgetary resources) in many countries. Concept
development, guidelines and standardization of analytical techniques and related research
and development programmes have been disseminated on a global basis. In the recent
past and at present, the Soil Resources Management and Conservation Service (AGLS) of
the Land and Water Development Division (AGL) of FAO is supporting some eight
countries in projects which include soil laboratory components.

While many developing countries have already established SWL, there are still
considerable deficiencies in providing effective advisory services to farmers,
interpretation of data which are essential to support improvement of soil management
practices, reclamation schemes, refinement of fertilizer recommendations, fertilizer
quality control, land-use planning, monitoring of land degradation and the establishment
of land quality indicators for policy decisions.

1.1 Soils and their classification

Soil is the thin layer covering the entire earth's surface, except for open water
surfaces and rock outcrops. The properties of soil are determined by environmental
factors. Five dominant factors are often considered in the development of the various
soils: (a) the climate, (b) parent materials (rocks and physical and chemical derivatives of
same), (c) relief, (d) organisms (fauna and flora), and (e) the time factor. There are a large
number of different soils, reflecting different kinds and degrees of soil forming factors
and their combinations.

2
 Figure 1. A hypothetical soil profile.

Scientists have developed different systems of soil classification to group soils of

similar properties in one class, allowing them to exchange information on soils found in
different areas. Soil classification also helps in determining the best possible use and
management of soils. Soil classification is however a controversial subject at both
national and international levels. There is lack of agreement for a common classification
system, because soil scientists do not agree on the characteristics for differentiating and
classifying soils.

3
 The classification of soils starts with examination of soil profiles.
Morphologically, soils are composed of a series of horizons. Soil horizons are layers of
different appearance, thickness, and properties which have arisen by the action of various
soil-forming processes. The horizons are normally parallel to the surface. Collectively,
the horizons make up what is called the soil profile or soil "pedon". A soil profile is
defined as a vertical section of the soil to expose layering. Figure 1 sketches a
hypothetical soil profile having all the principal horizons, with a brief description of the
characteristics of each horizon. Individual soils have one or more of these horizons. Very
young soils may not yet have started the soil horizonization process.

In soil classification, the item to be classified is the soil profile. The classification
or study of the entire profile consists of recognising and naming the horizons which make
up the profile. In the study of soil profiles, sub-soil horizons are given greater emphasis
than surface horizons which are frequently changed by human activity to such an extent
that they bear hardly any relationship with genetic process.

1.2 PLANT ANALYSIS

Crop and pasture production are functions of environmental, genetic and
management factors. Quality and yield determine the economic value of the crop. These
are in turn controlled by the genetic potential of the crop, environmental factors and
management. The mineral nutrition of the plant plays a major role in determining the
yield and quality of the crop. Plant nutrients other than C, H and O are described as
mineral nutrients. The elemental composition of plant dry matter is:
Carbon- 40-50%
Oxygen- 42-44%
Hydrogen-6-7%
Mineral elements - < 10%

The chemical substances required by crops are known as nutrients and their supply and
absorption for growth and metabolism is defined as nutrition. Researchers have identified
that 16 elements are essential for plant growth and development. Three of the elements

4
(carbon, hydrogen and oxygen) are derived from air and water and thirteen of them
(nitrogen, phosphorous, potassium, calcium, magnesium, sulphur, iron, manganese,
boron, zinc, copper, molybdenum and chlorine) are derived from soils and fertilizers.

The mineral nutrition of the plant plays a major role in determining the yield and quality
of crops. If any of the 16 essential elements are not available or are low in the soil the
plant function will be upset and characteristic symptoms will develop. Farmers and
scientists for generations have used visual deficiency symptoms to identify nutrient
deficiencies. We have to be careful with visual deficiency symptoms because a number of
factors other than low nutrients can cause similar symptoms to nutrients. e. g. moisture
stress, high salinity, herbicide damage or disease caused by bacteria, fungus or viruses

Plant analysis, along with soil analysis and other supporting data is used as a valuable
tool for managing the nutrition of crops and pastures. The main purpose of plant analysis
is to:

 Provide information on the nutrient status of plants as a guide to nutrient

management for optimal plant production.
 Diagnose existing nutrient problems and problems likely to affect crop or pasture
production.
 Monitor crop nutrient status for optimal crop production.
 Identify and monitor environmental problems associated with over fertilization of
crops and pastures.
 Assess the quality of plant products.
 Estimate overall nutrient status of regions, districts or soil types.
 Assess nutrient levels in stock feed and food for human consumption.

Researchers, extension officers and others use the relationship between nutrient
concentration and yield of plants or plant parts to assess plant nutrient status. The
standard concentrations used for diagnosing nutrient deficiency or toxicity are based
upon the concept of “ Critical nutrient concentration” that forms the basis of most

5
methods of plant analysis to assess plant nutrient status. In real situations it is not a single
value but it is a narrow range of concentrations above which the plant is adequately
supplied with nutrients or below which it is deficient.

These values are generally obtained from properly designed sand culture, water culture,
green-house or field experiments using increasing levels of nutrients in a deficient
growing medium. An appropriate yield (generally 90% of maximum yield) is selected
and the nutrient concentration in the selected plant part at this yield is accepted as the
critical nutrient concentration (For details refer: Reuter and Robinson, 1997). Critical
concentrations for specific nutrient deficiencies or toxicities are derived through
experiments as constant values. However, in practice, they vary widely due to a number
of environmental and other factors. All these factors should be taken into consideration
when interpreting any plant analysis data. Some of the factors are:

 Plant age and part of plant sampled: As the plant grows changes in nutrient
concentration take place in the plant tissues. In perennials, the concentration of
nutrients in leaves and other organs fluctuates with seasonal flushes of shoot growth
and fruit development. It also varies between leaves of vegetative and fruiting shoots.
Therefore it is necessary to define growth stages at sampling to assist interpretation.
As critical concentrations vary with age of plant parts it is essential that parts of the
same physiological age are used, irrespective of degree of deficiency. Generally the
youngest fully expanded leaf is used.

 Critical nutrient concentrations are found to be diverse among different genotypes.

However, when values are derived from similar tissues and plants of similar
physiological age this diversity is reduced. (e. g. sulphur requirements of
graminaceous crops are less than those for leguminous crops).

 Critical nutrient concentrations for diagnosis of nutrient deficiencies vary as a result

of interactions with light, temperature, carbon dioxide concentration, diseases, soil
and other nutrients. Critical concentrations for toxicities also vary with environmental

6
 factors. In a number of species sodium can substitute for potassium. Therefore the
level of sodium in the soil can affect the critical concentration of potassium.

Functions of nutrients in crops and pastures:

All the essential nutrients are directly involved in the nutrition of the plant. Some
are required in larger quantities and are known as macronutrients whereas others are
required in small quantities and are called micronutrients or trace elements. The functions
of the nutrient elements are listed below:

Nitrogen:

The nitrogen content of plant dry matter generally ranges from 1 to 5 %. However
occasionally it may be either lower or higher than this range. Plants need a wide range of
proteins to grow, develop and mature. The main body of protein is amino acids and
nitrogen is the major component of amino acids. Nitrogen is also present in chlorophyll
(the green pigment which traps sunlight). Soil micro-organisms feed on soil nitrogen
during break down of organic materials. Nitrogen improves quality of leafy vegetables. It
promotes rapid growth and if the supply is out of balance with other nutrients flowering
and fruiting may be delayed.

Phosphorous:

The phosphorous content in plants is usually between 0.1 and 0.5 % of the dry matter.
Phosphorous simulates early root formation and growth, gives a rapid and vigorous start
to plants and stimulates flower and seed production. Phosphorous is needed in the genetic
coding material which controls cell division.

7
Potassium:

The potassium content in plants is usually between 1-5 % of the dry matter.
Potassium is essential for efficient water relationships in the plant, both for controlling
water content in cells and movement of water through tissues, and in the control of the
stomatal cells. Potassium aids in providing mechanical strength to plants and assists in
the resistance to diseases. Potassium is also associated with the formation and
translocation of carbohydrates. It improves the quality of fruits and helps in the
development of tubers.

Calcium:

The calcium content of plants is less than 1%. It promotes early root hair
formation and growth. Calcium helps to maintain strong cell walls in plants. It also
neutralises poisons produced in the plant. It encourages grain and seed production. Plants
which contain high potassium, especially grasses, will contain less calcium

Magnesium:

The magnesium content in plant dry matter is similar to that of P (0.1-0.5%). It is

the essential component of chlorophyll and acts as a carrier of P in plants. It is necessary
for the formation of sugars and promotes the formation of oils and fats.

Sulphur:

Sulphur is an essential component of many proteins. It promotes nodule formation

in legumes and stimulates seed formation. Sulphur plays a key role in chlorophyll
formation.

8
Boron:

The content of boron in plant dry matter ranges between 10 and 100-200 ppm.
Boron helps in the manufacture of sugars and carbohydrates in crops. Boron is essential
for fruit development, translocation of sugars and the development of seed and seed
quality in some crops like mungbeans. Boron aids in the utilisation of calcium, nitrogen
and phosphorous. Boron is also important in the development of young roots and shoots.

Zinc:

Plants contain 20- 100 ppm of Zinc in the dry matter. Zinc is an essential
component of many enzymes, including some plant growth hormones. Zinc is also
essential for chlorophyll formation. It plays a role in protein synthesis, seed maturity and
plant height development.

Copper:

The copper content in plants ranges from 1 to 20 ppm of the dry matter. Copper is
essential for normal seed setting in legumes and cereals. It is associated with enzymes
that convert nitrogen to protein. Copper is a constituent of the chloroplast and aids in the
stability of chlorophyll.

Manganese:

Plants contain about 20-250 ppm Mn on a dry weight basis. If the levels exceed
500 ppm toxicity symptoms will appear. Plants with a manganese level of 15-25 ppm
will exhibit deficiency symptoms. It plays a specific role in the formation of chlorophyll.
Manganese accelerates germination and maturity.

9
Iron:

The iron content of healthy plant tissue ranges from 50-200 ppm of dry matter.
Iron is essential for proper functioning of chlorophyll.

Molybdenum:

The molybdenum content of plant material is usually less than 1 ppm in the dry
matter. Molybdenum is important in the process of nitrogen fixation by legumes and also
in the process where the plants use nitrogen.

Chlorine:

The chlorine content of plants ranges from 0.2 –2.0 %. It is essential for
photosynthesis. It is also involved in the uptake, movement and efficient use of water in
plants.

Key soil attributes

The soil serves the needs of the plant by providing
• Water
• Air
• Nutrients
• Stability

The ability of a soil to provide these services may be evaluated by key soil attributes (see
table following).

Key soil attribute Relevance to plants

Wetness Water supply, exclusion of air and, consequently, exclusion of
oxygen.
Root barrier Controls the depth of soil available for roots to extract water and

10
 nutrients, and to anchor the plant
Stoniness Stones and rocks dilute the volume of soil within the root depth
that is available for water storage and nutrients
Porosity Promotes stability by allowing deep rooting. Drains excess
water, and circulates air to roots
Natural nutrient Controls nutrient supply and reserves
status
Drought proneness An interaction between climate and soil attributes

Observing the soil

 The best way of observing the soil is to dig a pit, large enough to clearly see
features to at least 70 cm depth.
 One side of the pit is cut clean with a spade to reveal structure, colour and
layering.
 Otherwise, road cuttings or other exposures can be used, but these can be
misleading because exposure to the weather may mask some important soil
attributes.
 It is necessary to dig back into the cutting (20–30 cm) to expose natural soil.
 With some experience an auger can be used to rapidly sample the soil to depths
up to 1 m. This is particularly useful after examining a pit to see how features
seen in the pit vary in the landscape.

1.3 Objectives of the study

Soil cover change in commune in the past decades threatens agricultural
production as soil fertility declines and water erosion from upland to lowland areas
occurs. Models have proven to be an important tool to assess such impacts but also to
understand drivers of land use and land cover change (Verburg et al., 2008).

In that sense, they may provide policy makers with recommendations to improve
land use decision planning for sustainable agriculture in tropical and subtropical upland

11
areas. However, applying spatially explicit models requires a consistent database of land
cover and soil maps. A reverse modeling approach could be particularly appropriate to
reconstruct the development that led to the current situation on site and identify the main
drivers responsible for this development.

In addition to the abovementioned inputs, a reverse modeling approach requires

detailed spatially explicit information on land cover history. Due to the limitation of high
resolution satellite images and historical information in the study area the objective of
this study was to develop a hybrid classification scheme and classify the change of land
cover from selected points in time – 1993, 1999 and 2007 – by using satellite images and
locally derived land use history information with a special emphasis on maize crop cover
expansion. The generated results will provide part of the required input parameters to run
the LUCIA model.

Specific objectives of this study were to obtain satellite images and classify
different land covers using remote sensing and GIS software collect field data on soil
properties and land cover collect information on current land cover and land cover
history from farmers and local authorities elaborate a decision-tree for land suitability
based on soil properties and participatory soil mapping and subsequently derive decision
rules for image classification based on farmers’ practice cross-check the image-based
land cover classification on the basis of the soil and image data.

1.4 Study Area

Land Resource of Tamil Nadu

Tamil Nadu has about 5.96 percent of the Nation’s population, occupies 4 percent of the
land area and has 3 percent of the water resources of the Nation. The demographic
changes, economic growth and social development assisted by welfare measures would in
the coming decades, exert strong and competing demands on the finite natural resources
of the states, such as land, water,

12
raw-materials, etc. In Tamil Nadu, the per capita availability of land is only 0.18 ha while
the per capita net sown area is only 0.07 ha.

Even though the share of agricultural sector in the State Domestic Product of
Tamil Nadu has declined from about 52 percent to 18 percent between 1960-61 and
2007-08, and furtherto 8.2 percent during 2010-11, agriculture continues to be a major
source of livelihood for the rural people. Agriculture still employs about 40 percent of the
workforce in the State. As a basic input for agriculture, land occupies a predominant
position among all the resources required for a modern economy.

Competition between agricultural and nonagricultural sectors for land is

intensifying due to the increasing pressure on land for food production, housing and
industrial expansion. Between 1960-61 and 2009-10 the total cultivated area in Tamil
Nadu decreased from 7.32 million ha to 5.57 million ha and the net sown area has
decreased from about 6 million ha to 4.90 million ha during the same period. However,
this reduction in cropped area has been compensated by the increase in productivity of
crops so that higher production has been possible.
Apart from agriculture, land is also required for a forestation purposes so as to maintain
ecological balance, in view of the fact that the current share of forests in total
geographical area of Tamil Nadu is only 16.3 percent as compared to the ecological norm
of 33 percent. A harmonious balance among the various uses of land, which does not
compromise on the availability of life support systems and contribute for improving the
quality of life of the various sections of the population who are dependent on land is
essential.

The long-term trends in area under different land use categories and their percentage
share in total geographical area of Tamil Nadu reveal the following:
• There has been a marginal increase in area under forests from 19.04 lakh ha to 21.20
lakh ha while there was a decline in area under barren and uncultivable lands from 8.85
lakh ha to 4.93 lakh ha between 1960 and 2000s.

13
• Due to the increasing demand for land for industrial, housing and infrastructure
development, the land put to nonagricultural uses has shown a sharp increase of 42
percent from 13.57 lakh hato 21.05 lakh ha during this period.

• The area under cultivable wastes has decreased by about 50 percent from 6.60 lakh ha
to 3.61 lakh ha.

• Though the area under current fallows increased to 16.18 lakh ha during 1980s, it
decreased to around 10.09 lakh ha during 2000s.

• One of the most disturbing trends is the sharp increase in other fallows from 6.10 lakh
ha in 1960s to 11 lakh ha in 1990s and to 15.25 lakh ha in 2000s.

• The current fallows reached the maximum during the 1980s while the other fallows
reached the maximum during the latest decade.

• Consequently, the area under total fallow lands crossed 21 lakh ha during the latest two
decades thus accounting for about 16 percent of the total geographical area of the State
and almost 50 percent of the net sown area during 2000s.

• But for the moderate increase in net sown area during the sixties and seventies, there
has been stagnation in the net sown area in the State around 56 lakh ha, which has further
decreased to about 48.2 lakh ha during the 2009-2010.

• The gross cropped area also remained stagnant around 67 lakh ha during the 50-year
period from 1950 to 2000, thus leaving the cropping intensity constant around 120
percent. However, the gross cropped area in the state in 2009-10 has dipped to about
55.71 lakh ha with the cropping intensity slipping to 114 percent during 2009-2010.

To sum up, the dynamics of land use pattern in the State over the last fifty years reveals
that there was a significant decline in the Net Sown Area, Gross Cropped Area, and the

14
Area under Cultivable Wastes and Barren and Uncultivable Wastes while there was a
sharp increase in the land put to nonagricultural uses and other fallow lands.

Challenges confronting scientific use of Land

Currently, land resources are clearly under stress; 17.16 percent of total geographical area
is degraded and the percentage is increasing. Traditional systems of land management are
either breaking down or are no longer appropriate, and the management and technology
needed
to replace them is not always available.

The primary reason for this situation is the increasing demand placed on land by the
unprecedented rate of population growth and the effects it induces. Externalities related
to global change are also becoming a constraint to sustainable land management.

Conventional land-use planning has frequently failed to produce a substantial

improvement in land management, or to satisfy the priority objectives of the land users.
As a result, rural development programmes have had mixed success in meeting
production and conservation aims.
In calling for an integrated approach to the planning and management of land
resources,the following specific needs are identified:

• Development of policies which will result in the best use and sustainable management
of land

• Improvement and strengthening of planning, management, monitoring and evaluation

systems

• Strengthening of institutions and coordinating mechanisms

15
• Creation of mechanisms to facilitate the active involvement and participation of
communities and people at local level

• An effective and practical land use policy

• Utilize techniques whereby economic, social and natural concerns can be compared and
traded off on an equal footing;

• Establish a land-use classification system which fully reflects the natural and
ecological value of land;

• Use plans and tools that are more responsive to market conditions, recognizing the
economic drivers of growth;

• Enhance the legal and administrative character of land-use plans so that they
do not readily permit exceptions;
• Consider ‘spatial planning’ as an alternative to land use planning.

Concerns in Land Use Planning

• The agricultural productivity coupled with farm and non-farm diversification has to
be increased. Technological change has to be upgraded.

• Existing Infrastructure facilities like irrigation, roads, market etc. has to be

improved and new facilities has to be created.

• The focus may be on Farm based planning.

• A transparent and people friendly mechanism has to be developed for regulation of

land, water and other natural resources, and also for appropriate land use planning.

• Efforts must be on updating the existing data and collection of new data on land

16
use for various regions.

• A mechanism has to be evolved for the integrated system of land resource

conservation, development and management.

Issues in Land Allocation

Land use may be guided and regulated in selected areas. Competition between
agricultural and non-agricultural sectors for land is intensifying due to the increasing
pressure on land for food production, housing and industrial expansion. Therefore, zoning
of land for a conflict less development with optimum use of land resources is the need of
the hour. Accordingly, four major land use zones have been identified and policies and
proposals for the zones are as follows:

• Controlled / development / regulated zone

• Highway Corridor zone

• Natural conservation zone

• Agricultural (rural) zone outside controlled/development/regulated areas.

The existing urbanizable areas and proposed built up areas have to be regulated. The
agriculture (Rural) zone within the controlled / development / regulation areas has to be
strictly adhered to. Immediate steps to be taken to declare these lands as a reserved
agricultural zone which cannot be put to any other use other than agriculture. Urban
expansion and concentration of population
are a result of new employment opportunities in non-agriculture sector. These economic
opportunities have increased the conversion of agricultural land to other non-agricultural
uses. Therefore, the following policies have been proposed.

• Existing cultivated land be conserved for agriculture use as far as possible.

17
• Measures shall be initiated for protection of private agricultural land and ensure its
needless conversion into non agricultural/ urban use.

• Utilization of less and least valuable land for urban/expansion / new urban centers/
development purposes.

• Industries should be located in organized industrial areas such as industrial estates,

industrial parks, growth centers, software parks, textile parks and Special Economic
Zones (SEZ) only.

• The developmental activities such as industry could be planned in such lands as the
waste lands that are fit for construction activities and may be environmentally suitable.
The necessary infrastructures such as roads and electricity and amenities such as schools
and hospitals could be
planned and developed.

• The Government Poramboke lands may be used for the particular purpose for which
it is earmarked. All public roads, streets, lanes and paths, bridges, rivers, stream, tanks,
seashore and canals have to be protected and free from encroachment.

The Tamil Nadu Land Encroachment Act 1905 is not very effective leading to illegal
occupation of Government lands by individuals. The Government lands can be treated
similarly as private land and legal provisions may be enacted to file criminal trespass
against the encroachers.

Strategies for Managing Problem Soil Areas

With the increasing population, there is increasing demand for food, fodder, fuel and
fibre. The solution lies in increasing the productivity of existing lands, in view of the
non-availability of additional lands for cultivation in our State. Land degradation due to
desertification, soil salinity /alkalinity, water logging, drought/flood, excessive soil

18
erosion and unscientific agricultural practices have resulted in the creation of vast
stretches of wastelands of 2.42 million hectares in Tamil Nadu, further decreasing the per
capita cultivable lands. Realizing the need for the potential utilization of land resources
for the people of Tamil Nadu, the Government of Tamil Nadu has emphasized the
development of degraded and wastelands in the State on the basis of priority and in a
phased manner over a defined period.

Strategies for Managing Problem Soil

Areas

With the increasing population, there is increasing demand for food, fodder, fuel and
fibre. The solution lies in increasing the productivity of existing lands, in view of the
non-availability of additional lands for cultivation in our State. Land degradation due to
desertification, soil salinity /alkalinity, water logging, drought/flood, excessive soil
erosion and unscientific agricultural
practices have resulted in the creation of vast stretches of wastelands of 2.42 million
hectares in Tamil Nadu, further decreasing the per capita cultivable lands. Realizing the
need for the potential utilization of land resources for the people of Tamil Nadu, the
Government of Tamil Nadu has emphasized the development of degraded and wastelands
in the State on the basis of priority and in a phased manner over a defined period.

Salt affected lands

technology and the second is the spent wash technology. The spent wash technology can

19
be used for spot reclamation of wastelands, which includes opening of pits, treatment
with spent wash, filling the pit with normal soil and planting of trees.

Tannery effluent affected lands

Tannery effluents have higher Chromium content which gets into food chain through
crops like sugarcane and sunflower. Therefore phyto remediation with eucalyptus,

Rice Field affected due to

Chromium from Tannery casuarina, acacia and neem and growing of flower crops are the
other recommendations which can be followed for the improvement of wastelands.

Degraded forests lands

Forests with less than 20 percent area are classified as degraded forests under
wasteland classification. According to Forest Survey of India (FSI), the forests with less
than 40 percent vegetative cover are grouped under degraded forests. As per the estimates
given by FSI, 6 lakh ha were classified as degraded forests and out of this, forest

20
 Degraded forests lands
department has already improved 3 lakh ha of lands by afforestation.

Degraded pasture lands

Kangayam model of silvipasture consisting of Acacia and Cenchrus ciliaris species can
be adopted in the degraded pasturelands. Goat rearing is the best animal component to be
advocated for these lands. Manila tamarind (Kodukkapuli) can be planted in area

Degraded pasture lands

Eco-restoration in Swampy and Marshy lands

Wastelands classified under swampy and marshy lands in coastal areas need not
be reclaimed and brought under cultivation as the natural swamp and marshy area
facilitates wild life habitats. Similarly, the inland waterlogged areas can be left as such
since they facilitate ground water recharge in the neighboring cultivated lands.

21
 Swampy and Marshy lands
Increasing Water Use Efficiency in Agriculture

Majority of the canal irrigation system in Tamil Nadu are supply-based systems.
It is very difficult to improve water use efficiency in supply-based systems beyond a
certain level. Therefore gradually all the canal irrigation systems should be converted to
demand based systems. Better way to beneficially use water during nights is to do
groundwater recharges
through the existing open wells and construction of injection wells. Farmers must be
encouraged to allot land for construction of farm ponds in their own lands through
subsidies. Canal water available during night

time may be stored in these farm ponds.

Sometimes, when their due share of water is more than the field demand, that water can
also be stored in their farm ponds. The poromboke lands available in the canal command
areas can also be used for the construction of tanks to store water. During periods of

22
rainfall, the supply should be reduced based on the data regarding rainfall distribution.
For this purpose, farmers and lascars should be trained to use very simple low cost rain
gauges and transmission of data to the Water
Resources Department can be done through cell phone via SMS.

Strategies for well irrigation

Compared to canal and tank irrigation, the water use efficiency is found to be always
higher in well irrigation. But there is still a good scope for improving water use efficiency
in well irrigation. If water is conveyed through pipes from well to the field instead of
field channels, the conveyance efficiency can be increased. Subsidies can be provided for
installation of plastic conveyance system. If subsidy is provided for this, then the present
popular way of applying water to coconut through tap system would also be recognized
as an efficient water conveyance
system eligible for government subsidy.

Well Irrigation
Presently, the electric power supply for irrigation is free of cost to the farmers. This is
said to be one of the prime reason for poor water use efficiency due to indiscriminate
pumping of groundwater. Without withholding of free power supply, there is a possibility
of indirectly improving water use efficiency and also reducing the electric power use. All
the electric connections should be installed with electric power meters to measure the
amount of power used. For each farmer the optimal power that is needed should be
worked out based on the land holding, crops and depth of water level. Farmers who use
less power should be provided with financial aid and based on the magnitude of power

23
saved. The farmers who use more power will naturally start using less power as they
would not get any financial aid from the government.

1.5 Chapter Organization

Chapter 1 provides an introduction to this thesis, describes the problem, and objective of
the study.
Chapter 2 gives the introduction to the discuss and related research work.
Chapter 3 discusses the image enhancement, Soil Pattern generation and plant growth
estimation techniques algorithms.
Chapter 4 this chapter presents and discusses the experimental and performance
evaluation results.
Chapter 5 concludes the thesis and discusses future avenues of research in the end.

24
 Chapter II
2. Literature Review

There are a number of methods has been discussed for the estimation of plant
growth, and we discuss some of the methods here in this section.

Soil texture classification algorithm using RGB characteristics of soil image [3],
proposes the potential of soil texture classification using RGB histograms was
investigated. Seven sites representing major Korean paddy soil series were selected, 4-6
core samples up to 50-cm depth were obtained from each site, and each sample was
segmented by 5-cm intervals. For each segmented soil sample, four surface images were
taken using a miniaturized CCD camera, and texture fractions were determined by the
pipette method. Scatter plots showed linear patterns between silt content and histogram
variables such as brightness, skewness, and mode - brightness. When 5% averaged silt
content was linearly regressed with mode - brightness, R2, RMSEC, and RMSEP were
0.96, 2.2%, and 6.3%, respectively. When soils were classified using USDA criteria, the
laboratory method and the in-situ image processing method produced the same results for
48% of the samples.

Classification of Soil Textures Based on Law’s Features Extracted from

Preprocessing Images on Sequential and Random Windows [4], Texture analysis has
been used for recognizing synthetic and natural textures. Textures are one of the
important features in computer vision for image classification and retrieval. An important
approach to region description is to quantify its texture content. In this paper ,the Soil
images has been analyzed using various image pre-processing tasks such as Gray level
thresholding, Low pass filter, Edge enhancement using Prewitt’s Horizontal filtering and
then Feature extraction using 3x3 Law’s mask convolution. The features are constructed
on preprocessed methods applied on the Soil texture image by considering different types
of windows. These features offer a better classification rate. The experimental results on
various Soil textures clearly demonstrate the efficiency of the proposed methods.

25
 Origin of the soil texture classification system used in Japan [5], reviewed the
origin of the Japanese classification systems for soil particle size ranges and soil texture.
The size range system was adopted from International Society of Soil Science (ISSS)
standards. The soil texture classification system was introduced by Tommerup in his
paper at the ISSS Commission I (Soil Physics) meeting in 1934. This soil texture
classification system was modified by Yamanaka in 1955 and was henceforth adopted as
the conventional soil texture classification system in Japan. This ISSS-defined soil
texture classification system has been in use since that time.

Exploring plant growth-promoting potential of not rhizobial root nodules

endophytes of Vigna radiate [6], plant growth promoting traits of 26 non rhizobial and
one fungal endophyte previously isolated from Vigna radiata root-nodules were assessed
for IAA and siderophore production, phosphate solubilization and hydrolytic enzymes
production. Most bacterial entophytes improved seedling vigor index while fungal
entophyte (Macrophomina phaseolina) lacked all PGP traits. Entophytes Ml, M10 and
M15 were most influential in improving Seedling Vigor Index. Three endophytes having
multiple PGP traits with maximum siderophore production: 46.77 μg mL−1 (Bacillus
anthracis; Ml), IAA production: 10.81 μg mL−1 (Paenibacillus taichungensis; M10) and
phosphate solubilization: 134.483 μg mL−1 (Paenibacillus xylanilyticus; M15)
significantly increased root length (RL), shoot length (SL), number of lateral roots (NLR)
and plant dry weight (DW) when inoculated/co inoculated with E. adhaerens (native
rhizobia) to V. radiata in a small field trial. M10 inoculation produced longest RL while
Ml when coinoculated with E. adhaerens produced highest SL and NLR. Ml inoculation
or coinoculation was most effective in improving dry weight of mature plants. Most of
the entophytes coinoculated with E. adherence improved growth parameters. We report
that non rhizobial entophytes with PGP traits in combination with native rhizobia can be
prospective candidates for use as bio fertilizer.

A Regularized Particle Filter EM Algorithm Based on Gaussian Randomization

with an Application to Plant Growth Modeling [12], study a specific data and parameter
augmentation method which gives us the opportunity to estimate more easily the

26
parameters of the initial model. For this reason, the notion of Gaussian randomization of
a model with respect to some of its parameters is introduced. The initial model can be
regarded as a sub model of the resulting extended incomplete data model. Under the
assumption that the initial model has a unique maximum likelihood estimator (MLE) and
that the likelihood function is continuous we prove that the extended model has a unique
MLE with common values for the parameters of the MLE which correspond to the initial
model. We also prove the reverse direction.

Moreover, an appropriate stochastic version of an EM (Expectation-

Maximization) algorithm is suggested to make parameter estimation feasible. In
particular, we describe how the regularized particle filter of Musso and Oudjane (1998)
can be used in this frequents based approach to performing the Monte Carlo E-step at
each iteration of the stochastic EM algorithm. This regularized version is particularly
adapted to the framework of Gaussian randomization since the last iterations of the EM
algorithm are characterized by low variance in the parameter distributions.A toy example
with available analytic solutions, a synthetic example and a real data application with
scarce observations to the LNAS (Log-Normal Allocation and Senescence) model of
sugar beet growth are presented to highlight some theoretical and practical aspects of the
proposed methodology.

Parametrization of five classical plant growth models applied to sugar beet and
comparison of their predictive capacities on root yield and total biomass [13], propose the
evaluation and comparison of five plant growth models that rely on a similar energetic
concept for the production of biomass, but with different levels of description
(individual-based or per square meter) and different ways to describe biomass repartition
(empirical or via allocation): Greenlab, LNAS, CERES, PILOTE and STICS. The models
were all programmed on the same modeling platform, calibrated on a first set of data, and
then their predictive capacities were assessed on an independent data set. First, a
sensitivity analysis was carried out on each model to identify a subset of parameters to be
estimated, to reduce the variability of the models. We were able to reduce the number of
parameters from 10 to 4 for Green lab, and from 16 to 1 for STICS. Three criteria were

27
then used to compare the predictive capacities of the models: the root mean squared error
of prediction and the modelling efficiency for the total dry matter production and the dry
matter of root and the yield prediction error.

The entire above-discussed methods use minimum factor to estimate the growth
and perform soil classification but suffers from the problem of accuracy in soil
classification and plant growth estimation.

Land use change, the physical change in land cover caused by human activities
such as agriculture and silviculture, is a common phenomenon associated with population
growth, market development, technical and institutional innovation, and related rural
development policy. Changes in land cover in turn can have various consequences on
economic growth, the level and distribution of income, and on natural resources such as
biodiversity, ecosystems, water, and soils (Müller and Zeller, 2002). Degradation of
natural resources is a global problem. In Vietnam, the mountainous upland area is the
ecological zone where resource degradation is most serious.

The uplands consist of hills, highlands and plateaus, occupying 24.4 million
hectares (74%) of the country’s total area (Leisz et al., 2005). Agriculture is more and
more practiced on steep hillsides as the population of Vietnam’s northern mountainous
regions increases. Subsequent consequences are deceasing land per person ratios. As an
example, in Yen Chau district agricultural land per person decreased from 0.5 ha per
person in 1980 to 0.2 ha per person in 1998 (Wezel et al., 2002). Farmers’ changes of
farm practices and cropping types results in change of land cover and vegetation type. A
non-permanent vegetation cover will lead to land degradation through increased soil
erosion. In that sense, it will affect upland farming activities by declining soil fertility,
and further negatively influence lowland areas by sedimentation, deterioration of water
quality and direct effects on the water balance.

Land cover changes are linked to the change of weather and climate in many
ways, as well as to human livelihoods and environmental components such as air quality

28
and water supply. They modify weather and climate through changed albedo,
transpiration rates and filtering of aerosols, thus directly influencing the Earth’s radiation
budget. Land cover change and biomass burning have impacts on emissions of CO2,
CH4, aerosols and dust. Land cover change itself can also modify the surface energy and
moisture budgets through changes in evaporation and the fluxes of latent and sensible
heat, directly affecting precipitation and atmospheric circulation as well as temperature
(Forster et al., 2007).

2.1 Remote sensing - GIS and land cover mapping

Geographic Information Systems (GIS) and remote sensing (RS) are known to be
not only powerful, but also cost-effective tools for assessing the spatial distribution and
dynamics of land cover (Giridhar, 2008; Wilson and Fotheringham, 2008; Zhiliang et al.,
2008; Tottrup, 2002; Dewan and Yamahuchi, 2009). Lentes (2006) also indicated the
necessity of using RS and GIS to improve data bases on land cover and slopes and to link
spatial models to better understand, explain and assess strategies of rural development.

There are many classification methods to distinguish and quantify land cover
based on satellite images. Generally, continuous spectral features in imagery are clumped
into discrete classes representing land cover types. Unsupervised methods are used to
produce RS-based maps of distinct zones and subsequently assign real land cover classes
to each zone. In contrast, supervised classification first calculates spectral characteristics
of spots with land covers known by maps and fieldwork on the image. Subsequently,
pixels on the image are assigned to the respective pre-defined categories of land cover
type (Wilkie and Finn, 1996). To classify images, supervised methods always need
homogeneous regions to be identified within the image. It is not possible to combine
present ground truthing points and old images because of changes in spectral response
from day to day and of ground patterns, weather, sensors and many other factors from
year to year.
RS can be a good tool for getting a more detailed impression on land cover
change (Zhou et al., 2008). The term remote sensing includes a wide range of

29
applications from digital scanning to optochemical photography and is widely used to
produce land cover information in GIS. Accuracy can reach up to more than 90% for land
cover classification (Tottrup, 2004). With the purpose of better understanding the effect
of land cover change on tropical forests by mapping, Tottrup and Rasmussen (2004)
introduced the basis and intuitive methods as one kind of hybrid classification to classify
temporal land cover change. By using ground truthing data derived from historical
interviews and multi-date LANDSAT imagery which contains information on
phenological properties such as canopy structure and pre-classification image smoothing,
it was expected to achieve higher accuracy of classification. This procedure would allow
using low resolution images to determine vegetation cover at commune level.

Satellites images are secondary data with varying spatial resolution. In terms of
cost, high resolution images such as ALOS, SPOT or aerial photographs are more
expensive. Some sources with low to average resolution such as LANDSAT are freely
available. However, for some areas in the world, availability of satellite images is limited
due to missing coverage, restricted access to existing images or high costs. Several
researchers developed and tested basis methods with supervised and unsupervised
classification. Tottrup (2002) combined RS data with social and historical data to
distinguish different tropical forest types; Leisz et al. (2005) combined RS with
laboratory analysis to create an enhanced soil map before applying a basis classification
method. Thematic information improved classification of mixed pixels in a study by
Dewan and Yamaguchi (2009).

Hybrid classification methods are not restricted to combining supervised and

unsupervised methods, but can also be linked up with socioeconomic databases (Fox et
al., 2005). Together with several combination tools to understand the interactions
between human and natural systems like ABM (Agent Based spatial computation Model),
GIS, participatory methods can also be appropriate tools to disentangle cause and effect
relationships between factors and changes in land use observations at different scales
(Castella et al., 2005). At a landscape level in small areas, hybrid classification can be a

30
good option to classify and detect land cover change and thus improve stand-alone
supervised classification methods.

2.2 Image Processing

Several researchers have used image-processing techniques for object recognition.

A series of morphological operations were implemented by Casasent et al. (2001) to
produce only an image of nutmeat from an image containing a number of nuts. Ruiz et al.
(1996) used color segmentation to locate and remove the long stems attached to
mechanically harvested oranges. Their color segmentation algorithm had 100% success
in discriminating the destemmed and the stemmed oranges. However, the algorithm
misclassified some pixels of the stem-calyx as background. El-Faki et al. (2000), used
color machine vision for the detection of weeds in wheat and soybean fields.

They used a color index for both the preprocessing and statistical discriminant
analysis (DA) for weed detection. Their experiments worked well with statistical
discriminant analysis compared to the two neural networks they trained. Aitkenhead et al.
(2003) used a simple method to discriminate plants and weeds using plant size as a
parameter. They also trained a neural network to discriminate plants and weeds. Rapid
identification of Africanized honeybees was done by used Batra (1998) using image
analysis. Hansen et al. (1997) evaluated wound status of a porcine animal model, using
color image processing. In their experiment, the differences in calibrated hue between
injured and noninjured skin provided a repeatable differentiation of wound severity for
situations when the time of injury was known.

Although researchers concentrated on both autonomous vehicles and image

processing on an individual basis, few researchers have concentrated on autonomous
vehicles that use image processing.

2.2.1 Gray Level Thresholding

31
 This is the simplest and the oldest of all thresholding processes. Hence we used
GRAY color space for the study. The advantages of the gray level thresholding are that
the algorithms are simple and easy to implement in real time. Xinwen et al. (2002)
measured the geometric features of insect specimens using image processing. In this ,
preprocessing the image included
1) Transforming the colored image to gray level image, cropping the area of interest,
using the median filter to smooth the image, widen the pixel gray level distribution.
2) Threshold input image.
3) Extract features such as the region area, boundary perimeter, number of holes etc
from the image.
4) Recognize three insect specimens using these features.

Batra (1998) used image analysis for rapid identification of Africanized

honeybees. Tao et al. (2001) used X-ray imaging to detect the foreign objects in deboned
meat. Uneven thickness of the meat resulted in the change of background that in turn
affected global thresholding. Therefore, they used a local threshold method called
adaptive thresholding. Their algorithm effectively detected foreign inclusions at all
locations of the image except for locations near the chicken filet’s periphery Casasent et
al. (2001) implemented a series of morphological operations to produce only an image of
nutmeat from an image containing a number of nuts. They used thresholding, blob
coloring and segmentation operations. Algorithm classified 99.3% of all nuts
successfully. Two errors known as under segmentation and over segmentation occurred
during the segmentation process. 0.25 % of the good nuts were slightly under segmented
and 0.7 % of the infested nuts were over segmented after nutmeat extraction.

Kim et al. (2001) used two scanning methods to detect pinholes in almonds. Their
algorithm performed better on scanned film images than on line-scanned images and it
detected both insect-damage region and germ region. This method produced 81 % correct
recognition on scanned film images and 65% correct recognition on line-scanned images.

32
 False positives were 1 % on scanned film images and less than 12 % on line-
scanned images. Neethirajan et al. (2004) studied airflow paths along the vertical and
horizontal directions in wheat samples using X-ray CT (computer tomography) scan
images. They used images of hard red winter wheat at 14% moisture content as a sample.
Local threshold was used instead of the global technique since the density of the grain
varied and which resulted in the change of background in the image. Later the image was
subjected to thinning and segmentation. A thinning algorithm strips away the outermost
layers of a figure until only a connected pixel width skeleton remains (Xu et al., 2004).

The obtained skeleton image was a single stranded subset of an original binary
image.Further processing and blob coloring the image revealed a 9% difference in
airflow path area between horizontal to vertical directions in wheat bulk.

Gray level thresholding suffers from serious drawback,” Intensity Variation”,

because of which the intensity varies from time to time. Threshold values in gray level
thresholding are fixed using only one parameter i.e. intensity. Therefore, successful
segmentation using gray level thresholding requires change of threshold values
periodically. This limitation led to the use of other thresholding techniques RGB,
HSVthresholding commonly known as color thresholding.

2.2.2 Color Thresholding

Color thresholding is the process of separating the objects in an image based on

the color values such as RGB, and HSV. A major advantage of color thresholding is that
threshold can be fixed based on more than one parameter. This reduces noise in the
output image and results in a higher classification rate (CCR). Color thresholding can be
implemented using different color spaces such as RGB, HSV, HSI, La*b* etc. In the
present study, we implemented two-color thresholding techniques using RGB and HSV
color spaces.

33
2.2.3 RGB Thresholding

RGB represents the red, green and the blue color components of the pixels in an
image. Each color ranges from 0-255. Color pictures taken by the camera frequently use
RGB color space. Hence we included this color model for the study. We fix threshold
values based on all the three-color components or using only some color components. A
simple output is a bit map image which consists of only two gray levels (black and
white), in which white generally represents the foreground pixels or the pixels of interest
and black represents the background pixels.

Sometimes intensity information is not sufficient with natural images under

varying light conditions. Bulanon et al. (2002) developed a segmentation algorithm to
segment the Fuji apples based on the red color difference among the objects in the image.

Thresholding process was used to segment the images under different lightening
conditions. They noticed that threshold calculated from the luminance histogram using
optimal thresholding method was not effective in recognizing Fuji apple while the
threshold selected from color difference of red histogram was effective in recognizing
Fuji apple. In their experiment, the maximum gray level variance of the red color
difference between the fruit and the background determined the optimal threshold.

Results indicated that threshold values varied under different lightening conditions. Ruiz
et al. (1996) used color segmentation to locate and remove the long stems attached to
mechanically harvested oranges. Their color segmentation algorithm had 100% success
in discriminating the de-stemmed and the stemmed oranges. However, it misclassified
some pixels of the stem-calyx as background. Leemans et al. (1999) proposed a method
based on Bayesian classification to identify the defects on Jonagold apples. Method used
color frequency distributions of the healthy tissue and the defects to estimate the
distribution of each class. Most defects segmented bitter pit, fungi attack, scar tissue,
frost damages, bruises, insect attack and scab.

34
 Using the RGB values of the pixels directly would sometimes result in erroneous
calculations because the RGB values are sensitive to illumination. Therefore, it is better
to use relative color indices or other color spaces (such as HSI, YCRCB), which are less
sensitive to illumination or other factors affecting the RGB gray levels. Color indices are
the combination of RGB values through simple arithmetic operations like R/(R+G+B)
(Campbell, 1994).El-Faki et al. (2000) used color machine vision to detect weeds. They
used color index for both preprocessing and statistical discriminant analysis (DA) for the
weed detection. Their experiments worked well with the statistical discriminant analysis
compared to the two neural networks they trained. A color object recognition scheme has
been developed in which the extraction of color objects was based on an aggregation
function for watersheds using local and global criteria (Lezoray et al., 2003).

2.2.4 HSV Thresholding

The HSV model is similar to the way the humans perceive colors. This color
space separates intensity and color information. HSV represents the hue, saturation and
the value/intensity of the pixels in an image. Hue represents the dominant wavelength and
ranges from 0 to 360 degrees. Saturation represents the purity of color and is represented
in terms of percentage. The lower the saturation the more grey it is and more faded it
appears. Value/Intensity represents the brightness of the color and is represented as the
percentage. The objects of similar intensities but different hues can be distinguished
because of the addition of new variables hue and saturation. In HSV thresholding, three
threshold values can be fixed making the thresholding process more efficient. HSV is
superior to RGB in terms of “how humans perceive colors” According to Wang et al.
(2003), hue is invariant to certain types of highlights, shadings and shadows. Due to this
reason we included the HSV color model for the study.

Sural et al. (2002) used a segmentation algorithm to decompose an image into

useful parts. They observed that RGB thresholding blurs the distinction between two
visually separable colors by changing the brightness. HSV color space on the other hand

35
can determine the intensity and shade variations and retains pixel information. Liu et al.
(2001) developed a vision based stop sign detection system. They divided their research
into two modules: Detection mode and recognition mode. The detection module was
implemented using color thresholding in HSV color space. A neural network was
designed for recognition mode. Hansen et al. (1997) evaluated the wound status of a
porcine animal model using color image processing. In their experiment, the differences
in calibrated hue between injured and noninjured skin provided a repeatable
differentiation of wound severity for situations when they had a track of time of injury.

This color analysis distinguished mild, moderate, and severe injuries within 30
minutes after the application of the injury. They were not able to distinguish the severity
of wounds in the first few days but when the wounds were five to seven days old,
thecorrelation re-emerged. They concluded that their technique could be adapted for
assessing and tracking wound severity in humans in a clinical setting.

Computer vision based weed identification under field conditions using controlled
lighting was developed by Hemming et al. (2001). For each single object, morphological
and color features were calculated. Their experiments showed that color features could
help in increasing the classification accuracy. They also used color for segmenting plants
and the soil. Hatem et al. (2003) used color for cartilage and bone segmentation in
vertebra images for grading beef. Hue value was used effectively in segmenting the
cartilage areas and a* in the CIE Lab was used in segmenting the bone areas. Final
segmented object was obtained after applying other morphological operations such as
hole-filling, size thresholding, erosion and dilation. A fast and robust color vision for
Monash Humanoid was developed by Price et al. (2000) using a simple, fast, modified
HSV color model. The HSV model was devised since it is invariant to lighting conditions
and aids in the process of designing accurate color filter models.

They expanded the modified HSV model beyond the normal 360 degrees of hue
so that all colors can be viewed in a continuous distribution. Their color model proved
reliable in separation and filtering of colors within images. In a research done by Lee et

36
al. (2001), hue, saturation and intensity values were used in the thresholding process
followed by masking, noise removal and neural network that led to the extraction of lean
tissue from the carcass beef. The next step in any morphological operation is the removal
of extra noise. Researchers used different methods to reduce extra noise and size
thresholding is one of the best methods available for noise removal.

2.2.5 Size Threshold

Size is one of the parameters used to remove extra noise. Size thresholding
measures the size of each object (either 4 connected or 8 connected neighbors) and
removes the objects based on the threshold (size). Threshold in this case is the number of
pixels in an object. We retain only the objects whose size is greater than the defined
threshold and remove other objects. Aitkenhead et al. (2003) in their image processing
used noise removal to remove any pixel that has less than three neighbors with values
255. Hemming et al. (2001) implemented weed identification under field conditions using
controlled lighting. Different morphological features including the size and color of each
object were calculated and used to discriminate plants and soil. Hatem et al.(2003) used
size thresholding to remove objects bigger than 10,000 and 15,000 pixels respectively
after thresholding using HSL and La*b* color spaces and filling the holes. Size
thresholding is generally followed by other morphological operations such as dilation,
which polishes the output image. Image after undergoing size thresholding may not look
like a solid object as objects may still contain holes. We generally remove such holes by
changing those pixels value to foreground. This process is called dilation or filling of
holes.

Other Methods
Birds such as geese obstruct the air traffic during takeoff and arrival of air planes.
Short et al.1999 suggested new devices that use infrared, radar, low frequency sounds,
and laser devices to detect birds to reduce the bird hazards to aircraft. Klein et al.2003
developed a millimeter-wave (MMW) radar for dedicated bird detection at airports and
air fields. The research and development department of transport Canada has been
planning to develop and evaluate a three dimensional pulse Doppler radar to provide real

37
time information to air traffic devices and flight crews. Their project involves the
optimization of an Environmental Situational Assessment Radar (ESAR) to detect birds
within 5nm of the radar. Their radar would also determine the arrival and departure of
birds and provide information for real time warning. Bruder et al. 1997 concluded that the
incorporation of digital time lapse display can provide for detection and monitoring of
bird activity in near real time.

Hoshimoto et al. (2002) proposed the utilization of a popular digital camera for
color evaluation and quantification of color changes with the growth of grape tree. The
shape of grape leaf blade has also been quantified because the morphological information
such as width of petiolar sinus and length of sinus reflects the grape tree vigor. Liu et al.
(2000) developed a machine vision algorithm to measure the whiteness of corn kernel.
They used YCRCB instead of RGB color space to overcome the problems of varying
RGB component values with varying illumination. Algorithms were developed to extract
the leaf boundary of selected vegetable seedlings by Chi et al. (2002). They fit the leaf
boundary with Bezier curves, and later derived the geometric descriptors of the leaf
shape. Terawaki et al. (2002) discussed an algorithm for distinguishing sugar beet and the
weeds. The distinction between the sugar beet and the weeds, the green amaranth, the
wild buckwheat and the field horsetail, were tested using the shape characteristics of the
leaf and the angle of the leaf tip portion based on the image processing technique. The
results of the distinction indicated that the correct distinction rate of the sugar beet was
87.2% and the error rate was less than 8%. Aitkenhead et al. (2003) used image
processing methodology that involved the use of a simple morphological characteristic
measurement of leaf shape (perimeter2/area) which had varying effectiveness in
discriminating between plants and the weeds. Variation in their case was also dependent
on plant size.

Soren et al. (1996) proved that addition of spatial attributes such as image texture
improved the segmentation process in most areas where there were differences in texture
between classes in the image. Results from this experiment showed that texture could
have strong positive effects when using threshold-based segmentations than in minimum

38
size based segmentations. Segmentations controlled by minimum size criteria produced
higher accuracies than threshold based algorithms. The test sites included a simulated
forest, natural vegetation area and a mixed–use suburban area.

Even though, most of the possible object recognition methodologies have been
discussed in this chapter, only few of them will be implemented in my project. Object
(bird) recognition algorithms, my first objective, will be implemented using image
morphology, artificial neural networks and template matching. Bird recognition will be
followed by my second objective, testing algorithms in real time on autonomous boat
built by the Price and Hall team (Price, and Hall (2000, 2002), Hall and Price (2003 a, b,
c), and Hall et al (2001, 2004, 2005)).

39
 Chapter III
3.Research Methodology

There are several approaches or techniques for processing the images of soil
samples obtained. Some of the relevant ones are described briefly. The most suitable
method was determined and used for this research.

This research proposes a novel approach to performing Agricultural plant growth

estimation and soil classification. The proposed method uses multi-attribute soil pattern
to classify the soil images and to estimate the plant growth for each of the plant being
considered. The proposed method has the following functional components namely
Image Enhancement, Soil Pattern Generation, Soil Classification and Plant Growth
Estimation.

3.1 Image Enhancement

Image enhancement technique improves the overall quality of image. Spectral
image enhancement technology can enhance the differences between pixels and
spectrum. The main purposes of image enhancement are to change the gray scale of
images, to improve image contrast, to eliminate the edge and noise, highlight the changes
in Soil conditions.

The input image is enhanced for its quality by applying multi-level filter that
removes the noise from the input image. The noise removed image is passing through
histogram equalization to enhance the clarity of the image. The enhanced image is used
to perform image classification in the next stage.

40
 Fig : Geospatial Soil images for Agricultural Plant Growth Estimation

3.2 Soil Pattern Generation

We extract the color values present in the image. There exists a number of color
values belongs to various features of the spatial image. The method separates the color
features belong to various features of the spatial image. First we extract the RGB values
and identifies the features belongs to agricultural plants and extract the features of soil.

41
 The same image is converted into gray scale image, and segmentation is
performed. Using the segmented image, the soil texture is extracted and the gray features
are extracted. With the extracted features, the method computes the water flow factor and
plant density values. The computed features are generated as a pattern that is used to
perform soil classification.

Algorithm:
Input: Enhanced Image Img
Output: Soil Pattern Sp.
Start
Read RGB values of Img.
¿ Img ¿
RGB Set Rs = ∑ ¿ ∑ RGB (Img ( i ))
i=1

¿ Rs ¿
Compute number of green Pixels Gp = ∑ ¿ ∑ Rs ( i ) . Green> GThreshold
i=1

Gp
Compute Plant Density Pd = ¿ ×100
¿ Rs ¿
¿ Img ¿
Extract soil texture St= ∑ ¿ ∑ P i ( Img ) . Black > SThreshold
i=1

Compute Histogram of St.

Hst = ∑ Distinct ( Rs ( i ) )
¿Img ¿

Compute water flow =

∑ ¿ ∑ Pi ( Img ) . Black <WThreshold
i=1
¿
¿ Img ¿
Compute Intensity deviation ID = Stddev(Hst).
Generate gray scale image gimg=GrayScale(Img).
Perform segmentation Simg=Segmentation(Gimg).
¿ Img ¿
Extract gray features Gf = ∑ ¿ ∑ Pi ( SImg ) . value<GrayThreshold
i=1

Compute Histogram of Gf.

Gst = ∑ Dis tinct(Gf ( i ))
Compute gray scale deviation GD = Stddev(Gst).

42
 Generate Soil Pattern Sp = { ID,GD,Pd,Wf}.
Stop.
The above-discussed algorithm extracts the features from the enhanced image and
generates the soil pattern.

3.3 Soil Classification

The soil classification is performed using the soil pattern generated in the
previous stage. The method maintains a number of soil patterns belong to different soil
class in the soil pattern set, using the soil patterns available in the soil pattern set; we
compute the soil texture similarity measure STSM. The soil texture similarity measure is
computed using all the features of the soil pattern.

Algorithm:
Input: Soil Pattern Set Sps
Output: Class C.
Start
For each class Ci from Soil Class Sc
Compute Intensity Similarity IS =
¿Ci ¿

∑ ¿ √ Sp ( Intensity )−Sps ( i ) . Intensity

i=1
¿
¿ Ci ¿
¿Ci¿

Compute Gray Similarity GS =

∑ ¿ √ Sp ( Gray )−Sps ( i ) . Gray
i=1
¿
¿ Ci ¿
Compute Water flow Similarity Wfs = =
¿Ci ¿

∑ ¿ √ Sp ( WaterFlow )−Sps ( i ) .Waterflow

i=1
¿
¿ Ci ¿
Compute Plant Density Similarity PDS = =
¿Ci¿

∑ ¿ √ Sp ( PlantDensity )−Sps ( i ) . PlantDensity

i=1
¿
¿Ci ¿
Computemulti attribute similarity MAS.
MAS = IS×GS×Wfs×PDS
End
Choose the most weighted class Ci = Max(MAS).
Stop.

43
 The above-discussed algorithm computes the multi-attribute similarity measure
using which the soil classification is performed.

3.4 Plant Growth Estimation

The plant growth estimation is performed using the soil pattern generated and the
pattern set available in the data base. Using the soil pattern generated from the input
image, we compute the plant growth factor for each plant being identified. The algorithm
computes the plant growth for each of the plant so that the particular plants can be
cultivated in the soil.

Algorithm:
Input: Soil Pattern Sp, Soil Pattern Set SPS.
Output: Plant Growth Estimation Set PGS
Start
For each plant class Pc
Sp(ID × GD)
Compute Soil Support SS =
(∑ Sps(ID × GD)/¿ Sps ( Pc ) )
Sp(Ws)
Compute Water Support Ws =
(∑ Sps(Ws)/¿ Sps ( Pc ) )
Sp( PD )
Compute plant support Ps =
(∑ Sps(Ps)/¿ sps ( pc ))

( SS+Ws+ Ps)
Compute Plant Growth Factor PGF =
3
Stop
The above-discussed algorithm computes the plant growth support for each of the plant
being considered and estimates the plant growth for the support of agriculture
development.

3.5 Software Review.

44
 To implement the image processing techniques described in Chapter 3, we need to
find a"suitable software package, incorporating all the essential features. Stress is laid on
the segmentation techniques offered by the package and its particle measuring
capabilities. Other important factors include consideration of the platform on which the
software operates and its cost. While most of the software available have standard filters
like Sobel, Prewitt, etc. programmed for edge detection, other segmentation techniques
such as watershed edge detection and wavelet based segmentation are difficult to find.

Matlab
Matlab 2013a software that combines numeric computation with advanced
graphics. It has application-specific group of files or functions termed as toolboxes. It
also has a special toolbox dedicated to wavelets. Toolboxes such as the morphology
toolbox (http://www.mmorph.com). wavelab (http://www-stat.stanford.edu/~wavelab ),
etc. developed by users is also made available on the web. The student edition of
mmorph for Windows and Linux and Solaris is available..However, the student version
will only operate on images which are 128 * 128 pixels in size.

For the purpose of determining the edges within an image, filters like Sobel,
Prewitt, Roberts and Laplacian of Gaussian (LoG) are available. The versatility of the
package lies in the ability to allow a user to write own Matlab functions or files. The
official web site for Matlab is http://www.mathworks.com. It uses bmp, jpg, tif and other
file formats.

3.6 Landsat Data - Imaging, Analysis, Interpretation

During the Your Turn activities, you may be asked to use some of ImageJ's tools.
If ImageJ is unfamiliar to you, here are a couple of tips that might be helpful.

This is the ImageJ control window:

Magnifier Tool
Wand Pan
Tool Tool

45
 Menu Bar
Button
Status Bar Bar

ImageJ - Zoom & Pan

 Click the Magnifier Tool button so it appears depressed. Then left-click on the
image to zoom-in and right-click on the image to zoom-out.

ImageJ - Thresholding

Let us suppose we want to determine the percentage of the Beverly image that is green
vegetation. Let's use the Beverly 4,3,2 image for this example because the near infrared
Band 4 is assigned to the red channel so that everything reflecting that near infrared
wavelengths appears red in the image. Green vegetation reflects near infrared and appears
red in this image.

 In the ImageJ control window, click File, select Open, navigate to the location of
Beverly(chs. 4,3,2).tif, and click Open.

 Click Image and select Type, then 8-bit to change the image to grayscale.

 Click Image and select Adjust, then Threshold to open the Threshold dialogue.

Each pixel in the 8-bit grayscale image has a brightness value from 0 to 255. White
(brightest with most reflected energy) is 255 and black (darkest with least reflected
energy) is 0. All values between are shown as a level of gray. In ImageJ, thresholding is a
tool that allows you to identify a specific range of pixel brightness values. All pixels in

46
the image with these values are displayed in red while all other remain unchanged. With
thresholding, you can select specific features and separate them from the background.

 A little experimentation might be necessary here. In the Threshold dialogue, move the
sliders until the selected pixels (shown in red) are about the same as those displayed
on the red channel in the 4,3,2 image.

ImageJ - Measuring

Once the desired pixels have been highlighted in red with thresholding, the next step is to
select the red areas with the Wand Tool.

 Click the Wand Tool button so it appears depressed.

 Move the crossed-hairs cursor just inside the left edge of a large red area and click it
once. The area will be selected and outlined in yellow.

 While holding down the shift key, continue clicking inside the left edge of as many
other red areas as you are able. Each will become outlined in yellow.

47
 Click Analyze and select Clear Results. Click No if asked if you want to save the
previous results.

 Click Analyze and select Set Scale. Accept the default pixel distance of 1.0 and
pixel aspect ratio of 1.0. Click OK.

 Click Analyze and select Set Measurement. Check Area and uncheck everything
else. Click OK.

 Click Analyze and select Measure. The Results window pops up with the
measurement results. Write the number in your notes.

 The Results number is the pixel count of the area highlighted in red and outlined
in yellow that represents the green vegetation.

 Click Edit, select Selection, and Select None to clear the yellow selection
outlines.

 Again, click Analyze and select Measure. A second Area measurement is shown
in the Results window.

48
The new number in the Results window is the pixel count of the entire image.

 To find the percentage of the image that is green vegetation, divide the pixel count of
the thresholded/selected areas by the pixel count of the entire image, then multiply by
100%. For example, if the pixel count of the thresholded area is 108748 and the total
pixel count of the entire image is 262144, then the thresholded area of the image is
41.5% of the entire image.

108748
× 100 %=41 . 5 %
262144

Use the same method to find the percentage of other images that represent other
features of interest.

49
Beverly, Massachusetts Landsat 5 (3,2,1)

Beverly, Massachusetts Landsat 5 (4,3,2)

50
 Beverly, Massachusetts Landsat 5 (2,3,4)

3.6.1 Soil Density and Porosity

A good understanding of soil density and porosity is essential to soil scientists,

engineers, and persons involved in crop and timber production. Soil pores contain air for
plant root metabolism and provide a reservoir for water storage. Careful irrigation
planning requires that one take into account the porosity as well as the initial moisture
content of the soil.
Bulk density values provide an extremely useful conversion factor for calculations
involving soil mass and soil volume. Bulk density measurements are also used to
calculate the total pore volume in a soil as well as the weight of soil to be moved during
an excavation. High bulk density values may indicate the presence of compact layers in
soil that could restrict root and water penetration.

There are two important characteristics of porosity – total pore space and the size
distribution of the pores. The total pore space of a soil is that portion of the soil volume
occupied by air and water. The soil texture, structure, and organic matter content
largely determine the amount or total volume of soil pore space. A well-aggregated, fine-
textured soil high in organic matter would have high pore space while a massive, coarse-

51
textured or compact soil would have low pore space. Soil pores are generally grouped
into two categories:

1. Micropores – pores less than 0.06 mm in diameter that are important for water
retention and,
2. Macropores – pores greater than 0.06 mm in diameter that are important to aeration
and saturated water flow.

Density Formulas

The density of any object is its mass divided by volume it occupies. Soil density is the
oven-dried weight per unit volume expressed in g/ cm 3 or Mg/ m3. We are interested in
measuring two different density values:

1. Bulk density (Db), and

2. Particle density (Dp).

Bulk Density, Db

Bulk density values represent the density of the oven dry soil as a whole; this includes
solids and pore space.

Db = Ws/ Vt
Where:
Ws = Oven dry mass of the sample (g)
Vt = Total volume of the sample, pore volume + solid volume (cm3).

Particle Density, Dp

52
Particle density values represent only the weight of dry soil per unit volume of the soil
solids; the pore space is not included in the volume measurement.

Dp = Ws/ Vs

Where:
Ws = Oven dry mass of the sample (g)
Vs = Volume of the solids (ONLY) (cm3).

Porosity Formula
Pore Space, PS
The pore space of a soil is that portion of the soil volume occupied by air and water.

PS = Vp/ Vt or PS = [1 - (Db/ Dp)]

Where:
Vp = Volume of the pores *
Vt = Total volume of the sample, pore volume + solid volume (cm3).

* Vp is difficult to measure, so it is more common to calculate PS from Db and Dp

3.6.2 SOIL Texture Model

Texture is a common visual experience. However, a precise definition of texture

is difficult to formulate. Traditionally, textures have been categorized as either structural
or stochastic. A structural texture is characterized by a set of primitives (texons) and
placement rules. For example, a brick wall texture is generated by tiling up bricks
(primitives) in a layered fashion (the placement rule). A stochastic texture, on the other
hand, do not contain explicit primitives (such as grass and sand). Since most natural
textures are neither complete stochastic nor structural, we attempt to characterize both
kinds of textures with a single model. The probability distribution of the intensity value
for a single pixel is completely determined by the intensity values of pixels in its spatial

53
neighborhood. More formally, let N(x, y) be the collection of pixel values in the
neighborhood of a pixel located at (x, y). Then the
value of pixel (x, y), I(x, y), can be represented as follows:

where F is a probability distribution that determines the pixel value of I(x, y) from its
neighborhoods N(x, y). The size of the neighborhood determines the “randomness” of the
texture. Textures with regular patterns such as tiled floor require large neighborhoods,
while small neighborhoods are sufficient for stochastic textures as such sand beach.
Given such a model, the goal of a texture analysis and synthesis algorithm is to estimate
the parameters of F, and generate new textures by sampling from it.

Figure 3 : The process/Algorithm in short

Texture analysis is an important problem since it conditions the quality of

segmentation and interpretation in lots of applications such as in the textile industry or for
satellite imaging. There exits two principal techniques in texture analysis.

54
 The first approach consists in characterizing a texture by using some parameters
such as the well known attributes from the co occurrence matrix or the run-length matrix
,etc. Most of attributes were proposed to characterize textured regions of an image. The
main problem concerns the difficulty to know which attributes might be used for an
application because the efficiency of each type of attribute is often badly known.

The second approach tries to define a texture model as a realization of a 2D

random field. Lots of texture models have been proposed in the literature but are usually
used for stochastic textures.

3.6.3 THE SOIL-WATER

Where detailed data about soil layers, depth to groundwater, and vegetation are not
available, hydrologists have often resorted to simple bucket models and budgeting
schemes to model near-surface hydrology. The soil-water budget is a simple accounting
scheme used to predict soil-water storage, evaporation, and water surplus. A typical
budgeting time step is one day. Surplus is the fraction of precipitation that exceeds
potential evapotranspiration and is not stored in the soil. The simple model used here
does not distinguish between surface and subsurface runoff, so surplus includes both. the
main purpose of calculating the water budget is to estimate surplus, which serves as input
to groundwater and surface water flow models. With this in mind, the basic equation for
calculating surplus is:

(1)
In Equation 1, S is surplus, P is precipitation, E is evaporation, w is soil moisture, and t is
time. Horizontal motion of water on the land surface or in the soil is not considered by
this model. Snow melt was ignored in the water budget computations for this study

55
because temperatures throughout the study region remain above freezing throughout the
year.
A major source of uncertainty in evaluating Equation 1 is estimating the evaporation.
Estimation of evaporation is based upon knowledge of the potential evapotranspiration,
available water-holding capacity of the soil, and a moisture extraction function.

3.6.4 SOIL-MOISTURE STORAGE TO YIELD SURPLUS

Soil-water budget calculations are commonly made using monthly or daily
rainfall totals because of the way data are recorded. Computing the water balance on a
monthly basis involves the unrealistic assumption that rain falls at constant low intensity
throughout the month, and consequently surplus estimates made using monthly values are
typically lower than those made using daily values. Particularly in dry locations, the
mean potential evaporation for a given month may be higer than the mean precipitation,
yet there is observed runoff, and budgeting with monthly values may yield zero surplus.
For this reason, the use of daily values is preferred over monthly values when feasible. In
this study, daily rainfall records were provided by FAO for a number of stations in the
india; however, the spatial coverage of these stations is sparse in some areas, and it is
difficult to interpolate daily rainfall over space.
Because of the difficulty in estimating daily rainfall at our regularly spaced
computational points, and because a consistent basis for comparison is needed between
the short term case (for which daily data are available) and the long term case (only
monthly data are available), monthly data were used in this study. However, the use of
monthly data does not yield enough surplus to match observed river flows in some areas
of the india even with the assumption of zero overland flow and stream losses. To resolve
this problem, a modification was made to the commonly used bucket model in which it is
typically assumed that no surplus is generated until the soil is completely saturated with
water; this assumption is not consistent with situations where the rainfall rate exceeds the
infiltration rate of the soil. In our modified model, a fraction of the precipitation is
extracted and declared runoff, before remaining precipitation is passed to the soil. This
extraction of precipitation is represented by the P*alpha term in Equation 3 below. This
scheme generates more runoff, enough to satisfy mass balance constraints in most areas

56
of the india. Without this term, the model will predict zero runoff during many months of
the year when some observed streamflow actually occurs. The runoff extraction term
roughly accounts for "event" or "quick" flow that cannot be modeled using monthly
averaged values.
Equation 3 describes how soil moisture storage is computed.

(3)

In Equation 3, w(i) is the current soil moisture, w(i-1) is the soil moisture in the previous
time step, P is precipitation, PE is potential evaporation, alpha is the runoff extraction
function, and beta is the soil-moisture extraction function. With monthly data,
computations are made on a quasi-daily basis by assuming that precipitation and potential
evapotranspiration for a given day are equal to their respective monthly values divided by
the number of days in the current month. Several conditions apply when evaluating
Equation 3: If w(i) drops below zero, then set w(i) equal to 0.01; if w(i)>w* where w* is
the water-holding capacity, then the surplus for that day is w(i)-w*+P*alpha.
The soil-moisture extraction function beta=w/w* was used for this study. Since there is
no precedent for the use of a runoff extraction function (alpha), the formulation of this
function was more speculative and deserves further study. In the mean time, a simple
relationship, alpha = w/w*, was used and yielded enough surplus to satisfy mass balance
constraints in the surface flow routing model.

57
 Chapter IV
3. Result and Discussion

The proposed method has been implemented using matlab, and the efficiency of
the proposed method in plant growth estimation and soil classification has been evaluated
using a number of data sets of spatial images. The result shows that the proposed method
has produced an efficient result in soil classification and plant growth estimation with
more accurate results.

Symbol Soil Type

S Sand
Sicl Silty Clay Loam
Sic Silty Clay
C Clay
Sl Sandy Loam

58
 Cl Clay loam
Sil Silty Loam
L Loam
LS Loamy Sand
SCL Sand Clay Loam
Sc Sand Clay

Table 1: List of soil types considered

59
Fig Sample Soil Input Images

60
 Figure Show the run window

Figure show the input image

The Figure shows the result of soil classification and plant growth estimation
produced by the proposed method. The result shows that the input image is classified as
class 2 and shows the estimated plant growth value.

Figure: Result of soil classification and plant growth estimation

The Figure shows the Soil classification and plant growth estimation result
produced by the proposed method and shows the estimated growth value.

61
 Figure sample input image is classified as estimated plant growth.

4.1 Soil support

As water is the most limiting factor in the arid to semi-arid areas, soil support
determination is of major significance. Soil support influences crop growth not only by
affecting nutrient availability, but also nutrient transformations and soil biological
behavior. Therefore soil support is routinely measured in most field trials. While it can be
assessed in the field by the neutron probe, the gravimetric approach is more flexible, as
samples can be readily taken from any soil situation.

Sp(ID × GD)
Compute Soil Support SS =
(∑ Sps(ID × GD)/¿ Sps ( Pc ) )

4.2 Water support

62
 The water-support is defined as the amount of water held in the soil after the
excess gravitational water has drained away and after the rate of downward movement of
water has materially ceased. Stage of field capacity is attained in the field of saturation;
this is the upper limit of plant-available soil moisture. We must distinguish between soil
water content, (the percent water on an oven-dry weight basis), and the soil water
potential (the energy status of water in the soil), which is usually expressed in pressure
units.
Sp(Ws)
Compute Water Support Ws =
(∑ Sps(Ws)/¿ Sps ( Pc ) )

4.3 Plant Support

Plant support is defined as the arrangement of the soil particles. With regard to
structure, soil particles refer not only to sand, silt and clay but also to the aggregate or
structural elements, which have been formed by the aggregation of smaller mechanical
fractions. The size, shape and character of the soil structure varies, which could be cube,
prism and platter likes. Depending upon the stability of the aggregate and the ease of
separation, the structure is characterized

Sp( PD )
Compute plant support Ps =
(∑ Sps(Ps)/¿ sps ( pc ))

The measurements to have practical significance, the disruptive forces causing

disintegration should closely compare with the forces expected in the field. The field
condition, particularly with respect to soil moisture, should be compared with the
moisture condition adopted for soil disintegration in the laboratory. The sampling of soil
and subsequent disintegration of clods in relevance to seed bed preparation for upland
crops should be carried out under air dry conditions for dry sieve analysis. A rotary sieve
shaker would be ideal for dry sieving.

( SS+Ws+ Ps)
Compute Plant Growth Factor PGF =
3

63
 Method Soil Classification False Plant Growth
Accuracy Classification Estimation
Ratio
RGB 87 11.6 84.7
LAW 89 9.4 87.6
EM 91 7.3 89.3
SCPGE 97.8 1.3 96.9

Table 2: Comparison of different parameters of soil classification

The Table 2, shows the comparative result on soil classification and plant growth
estimation produced by proposed method on different soil type.

Soil Classification Accuracy

100
97.8
98
96
94
92 91 Soil Classification Accuracy
90 89
88 87
86
84
82
80
RGB LAW EM SCPGE

Graph 1: Comparison of soil classification accuracy

The graph 1, shows the comparative result on soil classification produced by

different methods and the result shows that the proposed method has produced more
classification accuracy.

64
 Plant Growth Estimation Ratio
98 96.9
96
94
92
90 89.3 Plant Growth Estimation Ra-
87.6 tio
88
86 84.7
84
82
80
78
RGB LAW EM SCPGE

Graph 2: Comparison of plant growth estimation efficiency

The Graph 2 shows the comparative result of plant growth estimation produced by
different methods, and it shows clearly that the proposed method has produced more
estimation efficiency than other methods.

Chapter V
5. Conclusion

This thesis proposed a soil classification algorithm for the estimation of plant
growth of different soils using the images obtained from the satellite. The method
enhances the input image quality by applying the Gabor filter and histogram equalization.
From the enhanced image, the method extracts the features like intensity, color values,
gray values, water flow, plant density.

Using the features being extracted the method generates the soil pattern and
computes the different similarity measures to classify the soil image. Similarly, the
method computes the support different values like water support, soil support, and plant
support from the pattern set using which the plant growth is estimated.

65
 The proposed method produces more efficient soil image classification results and
produces more efficient growth estimation for each plant being considered. The method
also reduces the time complexity of classification and estimation.

6. References

[1] Bae, S.G., Yeon, I.K., Park, S.D., Kang, C.K., and Zakaullah, K. (2004). Effects of
soil texture by soil addition on the growth and quality of oriental melon (Cucumis melo
L. var. makuwa Mak.) under protected cultivation. Journal of Bio-Environment
Control, 13(3), 156-161.

[2] Breul, P., and Gourves, R. (2006). In field soil characterization: approach based on
texture image analysis. Journal of Geotechnical and Geoenvironmental Engineering,
132(1), 102-107

[3] Chung, Sun-Ok; Cho, Ki-Hyun; Kong, Jae-Woong; Sudduth, Kenneth A.; Jung, Ki-
Youl, Soil texture classification algorithm using RGB characteristics of soil images, 3rd

66
IFAC International Conference Agricontrol, Agricontrol, Volume # 3 | Part# 1, pp:34-38,
2010.

[4] R. Shenbagavalli and Dr.K. Ramar, Classification of Soil Textures Based on Law’s
Features Extracted from Preprocessing Images on Sequential and Random Windows,
Bonfring International Journal of Advances in Image Processing, Vol. 1, Special Issue,
December 2011.

[5] Hirotatsu Murano, Yusuke Takata& Toshiyuki Isoia, Origin of the soil texture
classification system used in Japan, Soil Science and Plant Nutrition, 2015.

[6] M. Pandya, M. Rajput, S. Rajkumar, Exploring plant growth promoting potential of

non rhizobial root nodules endophytes of Vigna radiate, Springer, Microbiology January
2015, Volume 84, Issue 1, pp 80-89.

[7] Tariq, M, Hameed, S, Yasmeen, T, Zahid, M, Zafar, M (2014) Molecular

characterization and identification of plant growth promoting endophytic bacteria isolated
from the root nodules of pea (Pisum sativum L.).World J. Microbiol. Biotechnol. 30: pp.
719-725.

[8] Chihaoui, SA, Mhadhbi, H, Mhamdi, R (2012) The antibiosis of nodule-endophytic

agrobacteria and its potential effect on nodule functioning of Phaseolus vulgaris. Arch.
Microbiol. 194: pp. 1013-1021.
[9] Taurian, T, Anzuay, M, Angelini, J, Tonelli, M, Luduena, L, Pena, D, Ibanez, F,
Fabra, A (2010) Phosphate-solubilizing peanut associated bacteria: screening for plant
growth-promoting activities. Plant Soil 329: pp. 421-431

[10] Dawwam, GE, Elbeltagy, A, Emara, HM, Abbas, IH, Hassan, MM (2013) Beneficial
effect of plant growth promoting bacteria isolated from the roots of potato plant. Ann.
Agric. Sci. 58: pp. 195-201.

67
[11] Ahemad, M, Khan, M (2012) Biotoxic impact of fungicides on plant growth
promoting activities of phosphate-solubilizing Klebsiella sp. isolated from mustard
(Brassica campestris) rhizosphere. J. Pest. Sci. 85: pp. 29-36.

[12] Yuting Chen, Samis Trevezas, Paul-Henry Cournède, A Regularized Particle Filter
EM Algorithm Based on Gaussian Randomization with an Application to Plant Growth
Modeling, Springer, Methodology and Computing in Applied Probability March 2015.

[13] Baey, C, Didier, A, Lemaire, S, Maupas, F, Cournède, P-H (2013) Parametrization

of five classical plant growth models applied to sugar beet and comparison of their
predictive capacities on root yield and total biomass. Ecol Model 290: pp. 11-20.

Appendix
A) Implementation code

function varargout = PlantGrowth(varargin)

gui_Singleton = 1;
gui_State = struct('gui_Name', mfilename, ...
'gui_Singleton', gui_Singleton, ...
'gui_OutputFcn', @PlantGrowth_OutputFcn, ...
'gui_LayoutFcn', [] , ...
'gui_Callback', []);
if nargin && ischar(varargin{1})
gui_State.gui_Callback = str2func(varargin{1});
end

if nargout
[varargout{1:nargout}] = gui_mainfcn(gui_State, varargin{:});
else
gui_mainfcn(gui_State, varargin{:});
end

function PlantGrowth_OpeningFcn(hObject, eventdata, handles, varargin)

handles.output = hObject;
guidata(hObject, handles);

68
function varargout = PlantGrowth_OutputFcn(hObject, eventdata, handles)
varargout{1} = handles.output;

function edit1_Callback(hObject, eventdata, handles)

function edit1_CreateFcn(hObject, eventdata, handles)
if ispc && isequal(get(hObject,'BackgroundColor'),
get(0,'defaultUicontrolBackgroundColor'))
set(hObject,'BackgroundColor','white');
end

function pushbutton1_Callback(hObject, eventdata, handles)

[fn pn] = uigetfile('4.jpg','Select Image File');

complete = strcat(pn,fn);
filename = fn;
fname = complete;
set(handles.edit1,'string',complete);
axes(handles.axes1);
imshow(Im);
end

function pushbutton2_Callback(hObject, eventdata, handles)

fname = get(handles.edit1,'String');
Im = imread(fname);
Im(1,:)=histeq(Im(1,:),64);
Im(2,:)=histeq(Im(2,:),64);
Im(1,:)=histeq(Im(3,:),64);
axes(handles.axes2);
imshow(Im);

Im(2,:)
x=size(Im(2,:));
count=0;
x
mx=Im(2,:);
for i=1:x(1,1)
for j=1:x(1,2)
if mx(i,j)>230
count=count+1;
end
end
end
count
pd=(count/(x(1,1)*x(1,2)))*100;
pd
mx = rgb2gray(Im);
y=size(mx);
count=1;
count1=1;
for i=1:y(1,1)
for j=1:y(1,1)
if mx(i,j)<30
my(1,count)=mx(i,j);
count=count+1;
else

69
 my1(1,count1)=mx(i,j);
count1=count1+1;
end
end
end
my
ht=hist(my);
ht

x=size(mx);
count=0;
for i=1:x(1,1)
for j=1:x(1,1)
if mx(i,j)>230
count=count+1;
end
end
end
wf=(count/(x(1,1)*x(1,2)))*100;
wf
rs=std(my);
rs

rs1=std(my1);
rs1
Ps=zeros(2,2,4);
Ps(1,:,:)=Training('Class1');
Ps(2,:,:)=Training('Class2');

Ps
simi = zeros(1,2);
for i=1:2
Sim=0;
for j=1:2
Sim=Sim+((dist(rs1,Ps(i,j,1))+dist(rs,Ps(i,j,2))
+dist(pd,Ps(i,j,3))+dist(wf,Ps(i,j,4)))/4);

end
Sim=Sim/2;
simi(1,i)=Sim;
end
xs=0;
index=1;
for i=1:2
if simi(1,i)>xs
xs=simi(1,i);
index=i;
end
end
if index==1
set(handles.edit2,'String','Class1');
else
set(handles.edit2,'String','Class2');
end
set(handles.edit3,'String',xs);

axes(handles.axes3);

70
imshow(mx);
function edit2_Callback(hObject, eventdata, handles)
if ispc && isequal(get(hObject,'BackgroundColor'),
get(0,'defaultUicontrolBackgroundColor'))
set(hObject,'BackgroundColor','white');
end

function edit3_Callback(hObject, eventdata, handles)

if ispc && isequal(get(hObject,'BackgroundColor'),
get(0,'defaultUicontrolBackgroundColor'))
set(hObject,'BackgroundColor','white');
end

71