Journal of Advances in Management, Engineering and Science (JAMES)

Volume 01, Issue 01, March 2023

A COMPREHENSIVE STUDY OF ADVANCED TECHNIQUES FOR

PLANT HEALTH MONITORING USING DEEP MACHINE LEARNING
TECHNIQUES

Jawahar M
Associate Professor
Department of Computer Science and Engineering,
Malla Reddy College of Engineering, Secunderabad
Email: [email protected]

Gladson Britto
Associate Professor
Department of Computer Science and Engineering,
Malla Reddy College of Engineering, Secunderabad
Email: [email protected]

Abstract: Plant diseases cause significant production and economic losses in

agricultural industries worldwide. Health monitoring and disease detection of plants and trees
is important for sustainable agriculture. This review describes current technology that can be
used to develop soil sensor systems to help monitor plant health and disease under field
conditions. There are several leaf diseases responsible for the damaging the crop and leaf rust
is one of the major diseases which can destroy the entire wheat crop. Leaf diseases in crops
can be predicted more quickly and accurately, which could lead to the development of an
early treatment method and a significant reduction in economic losses. Advancement in
computer vision technologies has offered new dimensions in the direction of developing a
decision support system that can assist in identification and detection of plant diseases in
early stage. The use of computer vision and pattern recognition to detect disease has been
studied in an effort to reduce losses and achieve intelligent healthy farming. Deep Learning
models are widely used for the identification of plant leaf disease detection. However, still
researchers, experts and farmers face challenges when trying to deploy the trained deep
learning models in agriculture filed, in order to tune the model for achieving better results on
huge datasets. Majority of the deep learning models provide poor testing results if there is
some variation (rotation, tiling, and other abnormal image orientations) in scanned images.

Keywords: Classification, Detection, Image Processing, Plant Disease.

40
Journal of Advances in Management, Engineering and Science (JAMES)
Volume 01, Issue 01, March 2023

1. INTRODUCTION
Agriculture will need to produce approximately 50% more food by 2050 as a result of
the increase in the world wide population. India is primarily an agriculturally oriented nation,
with agriculture employing seventy percent of the people. Plant diseases have become
a concern because they can cause a significant reduction in the quality and quantity of
agricultural products, which affects the economies of countries that depend on agriculture for
their capital. get. All disease symptoms attract attention and cause concern because of
their effect on grain or grass and quality.

The development of computer systems in recent years, especially those with

integrated graphics processing units (GPUs), has led to a rapidly increasing number
of artificial intelligence applications related to machine learning, leading to the development
of new, developing methods and models. now a new group known as deep learning has been
formed[1]. When it comes to artificial neural network architecture, deep learning refers to
those with a large number of processing structures, unlike the "swallow" architecture used by
traditional neural network architecture. . Deep learning models, which can now be
computed, have changed areas such as image recognition [2], speech recognition and other
complex methods that consider the analysis of large data, and -give a big boost. applications
that use these techniques, such as autonomous vehicles, machine translation and
interpretation, and other similar techniques. The application of these deep
learning methods in agriculture [3], and especially in the field of plant disease diagnosis, has
only recently started, and is very limited, compared to other areas. Deep learning methods are
increasingly being used to solve machine vision challenges. Various researchers have
previously investigated the plant and leaf identification using various methodologies.
Initially, color information was employed to identify the plant from the soil in these
difficulties[4]. Deep learning is a relatively recent approach to plant and leaf identification.

Despite their best efforts, these artificial neural networks fall far short of the
capabilities of the human brain when it comes to "learning" from vast amounts of data. It's
possible to improve accuracy by using more than one hidden layer of neural network. In
many AI apps and services, deep learning is used to improve automation by performing
analytical and physical tasks without the need for human intervention. One of the machine

41
Journal of Advances in Management, Engineering and Science (JAMES)
Volume 01, Issue 01, March 2023

learning methods is deep learning. Many layers of data are used in deep learning to extract
useful features from the raw data, which is then used to recognize various input data
elements. Deep learning methods include convolutional networks, recurrent neural networks,
and deep neural networks.

Figure 1. Application of deep learning

2. LITERATURE REVIEW
This disease can also spread to other plants. The bacterial infection is more difficult to
find than the fungal infection because bacteria are so small that it is hard to see them[5]. The
first and most accurate diagnosis of plant diseases using artificial neural networks (ANN) and
image processing techniques[6] demonstrated. To accomplish the best outcomes, the
proposed technique depends on fake brain organization (ANN) for arrangement and Gabor
channel for highlight extraction. Different plant illnesses are characterized by ANN
characterization that utilizes a mix of surface, variety and different highlights to recognize
these infections.

Techniques for leaf-based disease detection are proposed[7]. Random forest is the
algorithm used in this proposed work to identify healthy and infected leaves in images. Data
identification, feature extraction from leaf images, data extraction of function and classifier
and classification are the proposed phases of work. Random forest algorithm and grouping of
infected and healthy videos are used to create datasets of infected and stable or healthy
leaves. In order to extract useful features from images, the histogram-oriented gradient
(HOG) was used. The SVM algorithm is a recent addition to the classifier in the neural
network approach to problems involving detection and accuracy in classification.

42
Journal of Advances in Management, Engineering and Science (JAMES)
Volume 01, Issue 01, March 2023

Augmentation Process. The fundamental reason for applying expansion is to build the
dataset and bring less twisting into the picture, which assists with diminishing the size during
the preparation interaction. In AI, and in measurements, restrictions emerge when factual
strategies depict commotion or arbitrary mistake as opposed to the hidden relationship [8].
Picture joining comprises of one of a few change strategies, including relative change,
viewpoint change, and straightforward picture change.

3. ROLE OF DEEP LEARNING IN AGRICULTURE AND PLANT DISEASE

DETECTION

Deep learning methods are increasingly being used to solve machine vision
challenges. Various researchers have previously investigated the plant and leaf identification
using various methodologies. Initially, color information was employed to identify the plant
from the soil in these difficulties [9]. The morphology of veins was used in some
investigations[10]. The leaf veins include a variety of texture and form features that can help
with plant identification using eyesight. The shape information was also utilized in certain
leaf-based research. Several researchers[11] integrated the shape and texture information
retrieved from the leaf, while others used color and texture information [12]. The main goal
of hyperspectral analysis of plant diseases is from cellular to regional. In general, there are
two main aspects involved in the hyperspectral detection of plant diseases: (I) to distinguish
infected plants from healthy plants and (II) to detect specific diseases among others. The first
condition affects only one type of plant and one virus and the second condition focuses only
on one virus, classifying all other possibilities as areas of indifference. For non-graphical
hyperspectral data, the representation of healthy and diseased parts of the body must be
obtained separately, and the analysis usually spans leaf, plant, and field scales. For the
selection of the classification algorithm, according to previous studies in the diagnosis of
plant diseases using hyperspectral imaging methods, the simplest and most commonly used is
the segmentation of the image is the boundary segmentation.

Plant sickness conclusion incorporates distinguishing proof, grouping and

characterization, and is one of them Picture division is utilized as a pre-handling procedure,
normally performed prior to checking an example to separate an objective item from the

43
Journal of Advances in Management, Engineering and Science (JAMES)
Volume 01, Issue 01, March 2023

foundation or make a cover for district of interest (return for capital invested) handling for
recovering extra data. Highlight extraction is a strategy for decreasing the quantity of
elements in a dataset by extricating new ones from existing elements (and erasing the first
elements).

Figure 2.1. workflow of hyperspectral imaging plant disease.

The main list of capabilities will actually want to sum up the vast majority of the data
in the new list of capabilities. From the summation of the first elements, a synopsis form of
the first highlights can be created. It is one of the mainstays of the ID and characterization of
articles by hyperspectral imaging [13-17]. In this way, in the characterization of plant
illnesses in view of hyperspectral pictures, the principal highlights incorporate the
appearance, yet in addition spatial elements, text highlights and other data that can be gotten
with the assistance of picture information.

Image processing and plant disease diagnosis based on hyperspectral images include
techniques which separates the data into health groups and different types of pathogens or
diseases. The above-mentioned image segmentation is often done in production methods to
improve the performance of data analysis, which is not always necessary in the diagnosis of
plant diseases.

In artificial intelligence (AI), and notably in machine learning, it is permissible to use

information and experience obtained from solving one problem to assess and handle other
related (and even unrelated) problems and difficulties. This is an AI and ML phrase referred

44
Journal of Advances in Management, Engineering and Science (JAMES)
Volume 01, Issue 01, March 2023

to as "transfer learning" [18-21]. To make research more effective in a setting where

computational resources are limited, this practice is becoming increasingly prevalent. Models
that have already been trained can be re-used to fill in the gaps.
Numerous techniques are right now utilized for plant illness finding utilizing PC
vision. One of them is sickness analysis by separating variety highlights as shown by the
creators [22]. In this article, the YcbCr, HSI, and CIELB tones are utilized in the review;
thus, the illness spot is recognized well and isn't impacted by clamor from various sources,
for example, camera streak.

Plant illnesses can be analyzed by fostering a progression of side effects. Patil and
Bodhe applied this technique for sickness determination in sugarcane passes on where they
utilized edge division to recognize leaf region and three-sided limit for sore region,
accomplishing a typical exactness of 98.60% in the last examination.

ANNs as a method for a programmed discovery and grouping of plant sicknesses was
utilized related to implies as a bunching system proposed by the creators in [23]. By and by,
there is a business arrangement, which involves visual acknowledgment to distinguish tree
species from their leaves' pictures however as the organization introduced in this paper is
ordering the plant illnesses rather than kinds of plant, Leafsnap was not utilized for
examination of the accomplished outcomes. Atomic strategies for plant illness location have
been widely settled[23-26]. The awareness of the atomic techniques alludes to the most
reduced amount of microorganism that can be distinguished in the example. The for the most
part involved sub-nuclear methods for contamination acknowledgment integrate ELISA and
PCR (PCR and consistent PCR) (PCR and steady PCR). Other sub-nuclear procedures
consolidate immunofluorescence (IF), stream cytometry, fluorescence in situ hybridization
(FISH), and DNA microarrays[27]. In the ELISA-based disease acknowledgment, the
microbial protein (antigen) associated with a plant disorder is imbued into an animal that
makes antibodies against the antigen. These antibodies are assembled from the animal's body
and utilized for antigen conspicuous evidence using a fluorescent tone and mixtures.

Fluorescence of citrus leaves was assessed for 60 days under four specific conditions:
leaves with no strain, leaves with mechanical tension, leaves with contamination, and passes

45
Journal of Advances in Management, Engineering and Science (JAMES)
Volume 01, Issue 01, March 2023

on with ailment notwithstanding mechanical strain. The assessments revealed the capacity of
fluorescence spectroscopy for illness end and detachment between the mechanical and
hypochondriac strain.

The ML models help to set up the model and give us incredible results. The photos
dealt with into the model will be ready by the components isolated and it will be ready. The
model used in the assessment have shown basic progression and results.

To portray the plant prosperity, the plant have relatively few component like length of
the leaf, size of the leaf and the assortment reflected from the plant. In this assessment we
have taken out the assortments bunches from the image that would be valuable to recognize
the sufficiency of the plant[28-29].

With regards to hyperspectral information imaging, environment handling,

mathematical amendment, and unearthly handling are the three fundamental parts of
hyperspectral information handling. Environmental change centers around surface reflectivity
blunders of air dissipating and retention applied to satellite information investigation;
Mathematical rectification centers around mathematical mutilations brought about by soil
changes, stage slant, and so forth. It is likewise utilized in field, low-level, airborne, and
satellite information examination.

This large number of physiological and biochemical markers are dynamic changes all
through the entire development time of yields. The extraordinary development time frame
and exceptional status have their unique attributes. Accordingly, SVIs are viewed as the lists
that mirror the development phase of plants in remote detecting. Direct model fitting and
game plan are the earliest and most intuitionistic approaches in quantitative end. To analysis
the illness seriousness by direct model fitting and grouping, measurable strategies and AI
calculations are generally utilized. Since measurable techniques are easy to see hypothetically
and yield noticeable outcomes, they are for the most part utilized in sickness seriousness
measurement

46
Journal of Advances in Management, Engineering and Science (JAMES)
Volume 01, Issue 01, March 2023

4. CONCLUSIONS

Plant diseases cause critical monetary and post-gather misfortunes in horticultural

regions. Creation area on the planet, particularly affected by environmental change age. This
article reviews and summarizes some of the methods used to diagnose plant diseases. The
review shows that these diagnostic methods have good potential to detect plant
diseases effectively. Studies show that these symptomatic techniques have huge likely in the
capacity to analyze plant sicknesses precisely. Spectroscopic and imaging innovation can be
incorporated into an independent farming vehicle for solid, continuous plant sickness
identification to accomplish progressed plant infection the executives. There are numerous
strategies for programmed or fake vision location in the administration of plant sicknesses,
however this exploration region is as yet uncommon. The utilization of profound learning
strategies to naturally extricate and recognize plant sicknesses from leaf pictures was
explored. The expansion of this study will zero in on the assortment of pictures to further
develop information capacity and work on the precision of the model utilizing different
handling and streamlining techniques. Additionally, future work will build the utilization of
the model via preparing it to perceive plant sicknesses across the land, by joining aeronautical
photographs of plantations and grape plantations caught by rambles with convolutional neural
network disclosure.

REFERENCES

1. Mirik, M.; Ansley, R.J.; Price, J.A.; Workneh, F.; Rush, C.M. Remote monitoring of
wheat streak mosaic progression using sub-pixel classification of Landsat 5 TM imagery
for site specific disease management in winter wheat. Adv. Remote Sens. 2013, 2, 16–
28.

2. Suomalainen, J.; Anders, N.; Iqbal, S.; Roerink, G.; Franke, J.;Wenting, P.; Hünniger,
D.; Bartholomeus, H.; Becker, R.; Kooistra, L. A Lightweight hyperspectral mapping
system and photogrammetric processing chain for unmanned aerial vehicles. Remote
Sens. 2014, 6, 11013–11030.

47
Journal of Advances in Management, Engineering and Science (JAMES)
Volume 01, Issue 01, March 2023

3. Jay, S.; Bendoula, R.; Hadoux, X.; Féret, J.-B.; Gorretta, N. A physically-based model
for retrieving foliar biochemistry and leaf orientation using close-range imaging
spectroscopy. Remote Sens. Environ. 2016, 177, 220–236.

4. Zhang, N.; Pan, Y.; Feng, H.; Zhao, X.; Yang, X.; Ding, C.; Yang, G. Development of
Fusarium head blight classification index using hyperspectral microscopy images of
winter wheat spikelets. Biosyst. Eng. 2019,186, 83–99.

5. Moshou, D.; Bravo, C.;West, J.;Wahlen, S.; McCartney, A.; Ramon, H. Automatic
detection of ‘yellow rust’ in wheat using reflectance measurements and neural networks.
Comput. Electron. Agric. 2004, 44, 173–188.

6. Singh, C.B.; Jayas, D.S.; Paliwal, J.; White, N.D.G. Detection of midge-damaged wheat
kernels using short-wave near-infrared hyperspectral and digital colour imaging. Biosyst.
Eng. 2010, 105, 380–387.

7. Shahin, M.A.; Symons, S.J. Detection of Fusarium damaged kernels in CanadaWestern

Red Spring wheat using visible/near-infrared hyperspectral imaging and principal
component analysis. Comput. Electron. Agric. 2011, 75, 107–112.

8. Mirik, M.; Jones, D.C.; Price, J.A.; Workneh, F.; Ansley, R.J.; Rush, C.M. Satellite
remote sensing of wheat infected by wheat streak mosaic virus. Plant Dis. 2011, 95, 4–
12.

9. Nagasubramanian, K.; Jones, S.; Singh, A.K.; Singh, A.; Ganapathysubramanian, B.;
Sarkar, S. Explaining hyperspectral imaging based plant disease identification: 3D CNN
and saliency maps. In Proceedings of the 31st Conference on Neural Information
Processing Systems (NIPS 2017), Long Beach, CA, USA, 4–9 December 2017.

10. Mishra, P.; Asaari, M.S.M.; Herrero-Langreo, A.; Lohumi, S.; Diezma, B.; Scheunders,
P. Close range hyperspectral imaging of plants: A review. Biosyst. Eng. 2017, 164, 49–
67.

11. Pandey, P.; Ge, Y.; Stoerger, V.; Schnable, J.C. High throughput In vivo analysis of
plant leaf chemical properties using hyperspectral imaging. Front. Plant Sci. 2017, 8,
1348–1359.

12. Tekle, S.; Mage, I.; Segtnan, V.H.; Bjørnstad, A. Near-infrared hyperspectral imaging of
Fusarium-damaged oats (Avena sativa L.). Cereal Chem. 2015, 92, 73–80.

48
Journal of Advances in Management, Engineering and Science (JAMES)
Volume 01, Issue 01, March 2023

13. Li, J.; Zhang, R.; Li, J.;Wang, Z.; Zhang, H.; Zhan, B.; Jiang, Y. Detection of early
decayed oranges based on multispectral principal component image combining both bi-
dimensional empirical mode decomposition and watershed segmentation method.
Postharvest Biol. Technol. 2019, 158, 110986–110996.

14. Liu, J.; Chiang, C.; Chen, S. Image-based plant recognition by fusion of multimodal
information. In Proceedings of the 10th International Conference on Innovative Mobile
and Internet Services in Ubiquitous Computing (IMIS), Fukuoka, Japan, 6–8 July 2016;
pp. 5–11.

15. Liu, J.; Lin, T. Location and image-based plant recognition and recording system. J. Inf.
Hiding Multimed. Signal Process. 2015, 6, 898–910.

16. Williams, D.; Britten, A.; McCallum, S.; Jones, H.; Aitkenhead, M.; Karley, A.; Loades,
K.; Prashar, A.; Graham, J. A method for automatic segmentation and splitting of
hyperspectral images of raspberry plants collected in field conditions. Plant Methods
2017, 13, 74–85.

17. Barbedo, J.G.A. A review on the main challenges in automatic plant disease
identification based on visible range images. Biosyst. Eng. 2016, 144, 52–60

18. Ali, M.M.; Bachik, N.A.; Muhadi, N.A.; Yusof, T.N.T.; Gomes, C. Non-destructive
techniques of detecting plant diseases: A review. Physiol. Mol. Plant Pathol. 2019, 108,
101426–101437.

19. Golhani, K.; Balasundram, S.K.; Vadamalai, G.; Pradhan, B. A review of neural
networks in plant disease detection using hyperspectral data. Inf. Process. Agric. 2018, 5,
354–371.

20. Dhingra, G.; Kumar, V.; Joshi, H.D. Study of digital image processing techniques for
leaf disease detection and classification. Multimed. Tools Appl. 2017, 77, 19951–20000.

21. Wahabzada,M.;Mahlein, A.-K.; Bauckhage, C.; Steiner, U.; Oerke, E.-C.; Kersting, K.
Plant phenotyping using probabilistic topic models: Uncovering the hyperspectral
language of plants. Sci. Rep. 2016, 6, 22482–22492.

22. Zhou, W.; Zhang, J.; Zou, M.; Liu, X.; Du, X.; Wang, Q.; Liu, Y.; Liu, Y.; Li, J.
Prediction of cadmium concentration in brown rice before harvest by hyperspectral
remote sensing. Environ. Sci. Pollut. Res. Int.2019, 26, 1848–1856.

49
Journal of Advances in Management, Engineering and Science (JAMES)
Volume 01, Issue 01, March 2023

23. Thomas, S.; Kuska, M.T.; Bohnenkamp, D.; Brugger, A.; Alisaac, E.; Wahabzada, M.;
Behmann, J.; Mahlein, A.-K. Benefits of hyperspectral imaging for plant disease
detection and plant protection: A technical perspective. J. Plant Dis. Prot. 2017, 125, 5–
20.

24. Virlet, N.; Sabermanesh, K.; Sadeghi-Tehran, P.; Hawkesford, M.J. Field Scanalyzer: An
automated robotic field phenotyping platform for detailed crop monitoring. Funct. Plant
Biol. 2017, 44, 143–153.

25. Shurtle, M.C.; Pelczar, M.J.; Kelman, A.; Pelczar, R.M. Plant disease. In Plant
Pathology; Encyclopedia Britannica: Chicago, IL, USA, 2020; Available online:
https://www.britannica.com/science/plant-disease (accessed on 1 April 2020).

26. Zhang, J.; Huang, Y.; Pu, R.; Gonzalez-Moreno, P.; Yuan, L.;Wu, K.; Huang,W.
Monitoring plant diseases and pests through remote sensing technology: A review.
Comput. Electron. Agric. 2019, 165, 104943–104956.

27. Dhaware, Chaitali G., and K. H. Wanjale. “A Modern Approach for Plant Leaf Disease
Classification Which Depends on Leaf Image Processing.” 2017 International
Conference on Computer Communication and Informatics (ICCCI), 2017.

28. Ramcharan, Amanda, Kelsee Baranowski, Peter McCloskey, Babuali Ahmed, James
Legg, and David P. Hughes. “Deep Learning for Image- Based Cassava Disease
Detection.” Frontiers in Plant Science 8 (2017).

29. Shafi, Uferah, Rafia Mumtaz, Naveed Iqbal, Syed Mohammad Zaidi, Syed Ali Zaidi,
Imtiaz Hussain, and Zahid Mahmood. “A Multi- Modal Approach for Crop Health
Mapping Using Low Altitude Remote Sensing, Internet of Things (IOT) and Machine
Learning.” IEEE Access (2020).

50