(San and Suzen, 2005 Hohle, 2009) (Kervyn, 2001) (Favey Et Al., 2003) (Schenk, 1996) (Taud Et Al., 1999)
(San and Suzen, 2005 Hohle, 2009) (Kervyn, 2001) (Favey Et Al., 2003) (Schenk, 1996) (Taud Et Al., 1999)
INTRODUCTION
1.1 Introduction
Digital Elevation Model (DEM) is a quantitative representation of the Earth’s surface
providing basic information about the terrain relief (Guth, 2006). DEM and its derived
attributes (slope, aspect, drainage area and network, curvature, topographic index, etc.) are
important parameters for information extraction or assessment of any process using terrain
analysis (Wolock and Price, 1994). These are prerequisite in different applications such as
modelling water flow (Jain and Singh, 2003), estimating runoff (Cai and Wang, 2006; Chappell
et al., 2006), flood simulation and management (Honghai and Altinakar, 2011; Ramlal and
Baban, 2008), route modelling (Romanowicz et al., 2008), mass movement (Iwahashi et al.,
2003), landform analysis (Weibel and Heller, 1990), creation of relief maps (Fraser et al.,
2002), volcanic hazards (Vassilopouloua et al., 2002), terrain visualization and mapping
(Spark and Williams, 1996), climate and meteorological studies (Thornton et al., 1997). The
outcomes of the models depend on the accuracy of DEM (Zhang and Montgomery, 1994;
Januchowski et al., 2010; Gómez-Gutiérrez et al., 2011).
DEM is generated using different techniques such as photogrammetric method using stereo
data (San and Suzen, 2005; Hohle, 2009), interferometry (Kervyn, 2001), airborne laser
scanning (Favey et al., 2003), aerial stereo photograph (Schenk, 1996) and topographic surveys
(Wilson and Gallant, 2000) using interpolation of contours maps (Taud et al., 1999). Like any
spatial dataset, DEM is subject to different type of errors such as gross error during data
collection (Rodgriguez et al., 2006), deficient orientation of stereo images (systematic error)
with photogrammetrically determined elevation values (Mukherjee et al., 2011) and unknown
combinations of errors (random error) which cannot be avoided. These errors vary
geographically depending on terrain conditions (Holmes et al., 2000). The other issues related
to DEM accuracy are grid spacing and interpolation techniques (Mukherjee et al., 2011).
The quality of DEM depends on the density of input data, on the accuracy of the surveying
technique that was used to collect the data and on the selection of the interpolation method
used to generate the DEM. DEMs are used as elevation data sources in various geospatial
studies and applications, such as topography, geomorphology, plant cover research, tsunami
assessments, urban studies, archaeology, and glacier observations. Contour lines, topographic
maps, global positioning system (GPS) measurements, photogrammetry techniques, radar
interferometry, stereo satellite images, and laser scanning are the main data sources that
produce DEMs. These data sources can be evaluated in four different aspects: cost, accuracy,
resolution, and pre-processing. Moreover, each of these techniques has both advantages and
disadvantages. (Alganci, Besol, & Sertel, 2018)
The UAV photogrammetry is a relatively young offspring from photogrammetry and
computer vision. Very high-resolution georeferenced ortho-images, point clouds, height
models, etc. are the base for a multitude of geospatial information. DEMs are also one of the
deliverables of photogrammetry which can be used for various civil engineering purposes such
as hydrological modelling, dam construction etc. The accuracy of these elevation model plays
a very important role in the successful execution of these crucial projects. These days with the
advent of UAVs in the market of photogrammetry and remote sensing-based application can
be used. The main objective of this study focuses on the comparison of the existing free or
open source DEMs with the one created by UAV based imageries in various aspects of hilly
and flat terrain. UAVs as a platform for data acquisition has been much efficient and cost-
effective in comparison with other platforms. In the discussion regarding the accuracies of
DEM, this study will evaluate the quality parameters and standards for the other DEMs with
the comparison with UAV generated DEM.
LITERATURE STUDY
2.1 Introduction
2.1.1 Digital Elevation Model (DEM)
Digital Elevation Model (DEM) is the digital representation of the land surface
elevation with respect to any reference datum. DEM is frequently used to refer to any digital
representation of a topographic surface. DEM is the simplest form of digital representation of
topography. DEMs are used to determine terrain attributes such as elevation at any point, slope
and aspect. Terrain features like drainage basins and channel networks can also be identified
from the DEM. (Scales, 2002)
DEMs represent elevation data and are the principal digital data source for slope and aspect
map coverages used in Geographic Information System (GIS) analysis for resource
management. Elevation data can be represented digitally in many ways including a gridded
model (where elevation is estimated for each cell in a regular grid), a triangular irregular
network, and contours. There are two main approaches for generating DEM: Interpolating the
regular grid from an irregularly distributed elevation data set, or generating the grid directly
using photogrammetric techniques.
Digital Elevation Model (DEM) is an important topographic product and essential DEM and
for many applications. Traditional methods for creating DEM are very costly and time-
consuming because of land surveying. In time, Photogrammetry has become one of the major
methods to generate DEM. Recently, airborne Light Detection and Ranging (LiDAR) system
have become a powerful way to produce a DEM due to the advantage of collecting three-
dimensional information very effectively over a large area by means of precision and time.
However, the main disadvantage of aerial manned platforms such as aeroplanes is being
expensive, especially for small study areas. During the last decades, low-cost Unmanned Aerial
Vehicles (UAVs) are used to pass this handicap. Nowadays, the use of UAVs is increasing day
by day due to its advantages at cost, inspection, surveillance, reconnaissance, and
mapping.(Uysal, Toprak, & Polat, 2015)
Digital Elevation Models (DEMs) are important datasets used in many different fields. Due to
the mishaps (time, resolution, accuracy, extend) faced in DEM generation by classical methods,
DEM production has become an important issue with similar new technologies. Therefore
LIDAR (Light Detection and Ranging) data and UAV photogrammetry are commonly used in
recent studies. Since LIDAR technology is still rather expensive, aerial photo-generated
mapping and orthophoto production is preferred to be the main sources of DEM production in
many countries. When similar studies in the literature are examined, it is readily obvious that
the high-resolution DEMs obtained through UAV aerial images appear to provide relatively
high accuracy. However, when investigated in detail, it became apparent that the settings on
where UAV imagery was acquired, had been ideal enough to produce those nearly perfect
results.
DEMs are considered as very significant geospatial datasets due to the versatile possibilities
for using them. They are used for ortho-rectification of aerial photographs, cartographic
representations, 3D visualizations, hydrological analyses and models, water management,
analysis of landscape dynamics, climate and climate impact studies, geological applications,
agriculture and forestry applications, road and dam planning, cut-drainage automatic basin
analysis, and-fill delineation, flood risk analysis, planning of telecommunication networks, and
geophysical modelling. On the other hand, DSM represents the Earth’s surface and includes all
objects on it, and DTM is a topographic model of the bare soil. (Joseph, 1984)
Figure 1: Digital Elevation Model
“UAVs are to be understood as uninhabited and reusable motorized aerial vehicles.” States are
remotely controlled, semi-autonomous, van Blyenburgh, 1999. These vehicles autonomous, or
have a combination of these capabilities. Comparing UAV to manned aircraft, it is obvious that
the main difference between the two systems is that on the UAV no pilot is physically present
in the aircraft. This does not necessarily imply that UAV flies by itself autonomously. In many
cases, the crew (operator, backup-pilot etc.) responsible for a UAV is larger than that of a
conventional aircraft (Everaerts, 2008). The term UAV is commonly used in the Computer
Science, Robotics and Artificial Intelligence, as well as the Photogrammetry and Remote
Sensing communities. Additionally, synonyms like Remotely Piloted Vehicle (RPV),
Remotely Operated Aircraft (ROA) or can also (UVS) and Unmanned Vehicle Systems
Remotely Piloted Aircraft (RPA) infrequently be found in the literature. RPV is a term to
describe a robotic aircraft flown by a pilot using a ground control station. The first use of this
term may be addressed to the United States (U.S.) Department of Defence during the1970s and
1980’s. The terms ROA and RPA have been used by the National Aeronautics and Space
Administration (NASA) and Federal the U.S. in place of UAV. Furthermore, the term Aviation
Administration (FAA) in the Unmanned Aircraft System (UAS) is also being used (Colomina,
et al., 2008). The FAA has adopted the generic class UAS, which was originally introduced by
the U.S. Navy. The common understanding is that the terminology UAS stands for the whole
system, including the Unmanned Aircraft (UA) and the Ground Control Station
(GCS).(Eisenbeiss, H., Grün, A., & Eisenbeiß, 2009)
Unmanned aerial vehicles (UAV), alternatively called drones, unmanned aerial systems
(UAS), and remotely piloted aircraft (RPA) are vehicles that can fly according to the
aerodynamic principles without a pilot. UAVs were initially developed and used for military
purposes around a century ago. During World War I and II, Korean War, Vietnam War and
Cold War, flying platforms were first utilized for surveillance, then used for bombing. The use
of this technology in mapping has begun as early as the beginning of the 21st century. Early
studies have proven successful enough results for this approach to be suitable for
photogrammetric and geometric necessities.(Akturk & Altunel, 2019)
Today drones are often used instruments in many disciplines because they provide flexibility
and many advantages in areas where direct human intervention is difficult and risky. In
addition, UAV imagery was reported to provide much better spatial and temporal resolution
than the satellite image. Thus, the benefits in terms of time, labour and cost, compared to
classical methods have increased the use of UAVs. Nowadays, forestry, agricultural,
archaeological, environmental, and three-dimensional topographical studies are generally
performed with this technology. The effectiveness and the precision provided by the
implementation of UAVs in various disciplines and industries have greatly increased the DEM
and for them, and the sales figures soared all around the world. UAV technology has become
cheaper over time, justifying the significant increase in sales.(Akturk & Altunel, 2019)
The term UAV is used commonly in the computer science and artificial intelligence
community, but terms like Remotely Piloted Vehicle (RPV), Remotely Operated Aircraft
(ROA), Remote Controlled (RC) Helicopter, Unmanned Vehicle Systems (UVS), and Model
Helicopter are often in use. The RC and Model Helicopters are clearly defined by the
Unmanned Vehicle Systems International Associations as mini, close, short, and medium-
range UAVs depending on their size, endurance, range, and flying altitude. UAVs are to be
understood as uninhabited and reusable motorized aerial vehicles which are remotely
controlled, semi-autonomous or have a combination of these capabilities, and that can carry
various types of payloads, making them capable of performing specific tasks within the earth’s
atmosphere, or beyond, for a duration, which is related to their missions. The UAVs system
was firstly applied for military goals but in 1979 Przybilla and Wester-Ebbinghaus, applied
one of the earliest experiments with fixed-wing UAVs for geomatic applications. UAVs are
established as a serious alternative for traditional data capture, especially while aiming at
mapping application with high spatial and temporal resolution and introduces also a low-cost
alternative to the classical manned aerial Photogrammetry. Data collection just requires small,
light UAV platform with a control system and standard consumer-grade digital camera.
Basically, UAVs applications can be classified as forestry and agriculture, archaeology and
cultural heritage, environmental surveying, traffic monitoring, 3D reconstruction. There are
various UAVs based civilian applications at the scientific literature.(Uysal et al., 2015)
Most of the (non-)commercially available UAV systems on the market focus on low-cost
systems, and thus a major advantage of using UAVs is also the cost factor, as UAVs are less
expensive and have lower operating costs than manned aircraft have. But, sometimes as
mentioned in the previous section - depending on the application - the cost can be similar to
manned systems. As for small-scale applications, the expenses for manned aircraft are not
maintainable, projects are quite often not feasible or terrestrial systems have to be used as
alternative systems, recognizing not all project requirements are met. Thus, UAVs can be seen
as a supplement or replacement to terrestrial photogrammetry in a certain area of applications.
In the case of a combination of terrestrial and UAV photogrammetry, it is even possible to use
the same camera system and having the same distance to the object, which simplifies the
combined data processing.
In addition to these advantages, the UAV-images can be also used for the high-resolution
texture mapping on existing DSMs and 3D-models, as well as for image rectification. The
rectified images and derivates, like image mosaics, maps and drawings, can be used for image
In addition to these drawbacks, UAVs do not benefit from the sensing and intelligent features
of human beings. Thus, UAVs cannot react like human beings in unexpected situations, e.g.
the unexpected appearance of an obstacle. In general, there are no sufficient regulations for
UAVs given by the civil and security authorities (Colomina, et al., 2008). Low-cost UAVs are
not equipped with air traffic communication equipment and collision avoidance systems, like
manned aircraft. Therefore, due to the lack of communication with the air traffic authorities,
UAVs are restricted to the flight in line-of-sight and to operate with a back-up pilot. The flight
range of the UAV is also, in addition to the line-of-sight regulation, dependant on the skill of
the pilot to detect and follow the orientation of the UAV-system. To take full advantage of the
impressive flying capabilities of UAVs, like the fully automated operating rotary wing UAVs,
there needs to be a well-trained pilot, due to security issues. The pilot should be able to interact
with the system at any time and manoeuvres. Based on the communication and steering unit of
UAVs, we can state that the operation distance depends on the range of the radio link for rotary
and fixed-wing UAVs, which is equivalent to the length of the rope for kites and balloon
systems used in the past. In addition, the radio frequencies (35 and 40MHz in Switzerland) may
be subject to interferences caused by other systems (remote controlled cars and model aircraft,
as well as citizens’ band radio), which use the same frequencies or may suffer from signal
jamming. Thus, depending on the local situation of the area of interest, the frequency for the
communication between GCS and UAV has to be selected carefully. Nowadays, UAVs are
also controlled via a 2.4GHZ radio connection, while the video link has to be shifted to 5GHz.
Table 1 features of aerial, close-ranging and UAV photogrammetry(Eisenbeiss, H., Grün, A., & Eisenbeiß, 2009)
UAV photogrammetry has opened a variety of new applications in the field of close-range
photogrammetry by combining aerial and terrestrial photogrammetry techniques to offer the
new applications in the field of close-range photogrammetry by combining aerial and terrestrial
photogrammetry techniques to offer the advantages of both. Thus, the application of UAV
photogrammetry in the field of civil engineering can be situated between techniques using the
advantages of both. Thus, the application of UAV photogrammetry in the field of civil
engineering can be situated between techniques using classic terrestrial systems and techniques
based on photogrammetry from images taken from conventional aircraft, representing an
economically viable alternative. Many such cases, UAVs are more competitive because they
require less time for data acquisition and reduce costs compared to the use of classical manned
aircraft (Aber et al., 2010).
1. Images are scanned for characteristic points, such as, for example, marks created in the
ground specifically to support the survey or manholes. If Ground Control Points (GCPs)
are used to geo-reference the model, they are usually labelled in the images before this
step.
2. Based on the characteristic points, image geo-information and the known camera
parameters, a sparse point cloud model is derived with a so-called bundle block
adjustment algorithm (Triggs et al., 2000). It is sparse since formed only of the
characteristic points from step 1.
3. Based on the sparse point cloud, dense image matching is performed to increase the
spatial resolution of the point cloud model and the 3-D elevation model generated.
(Leitão, Moy De Vitry, Scheidegger, & Rieckermann, 2016)
Figure 3: This model of a house has high relative accuracy as it's correctly
reconstructed, but low absolute accuracy, as it's in the wrong place
Ground sampling distance (GSD) is the distance between the centre of two consecutive
pixels measured on the ground.
For example, a GSD of 5 centimetres means one pixel in the image represents 5 linear
centimetres on the ground. The same pixel will cover 25 square centimetres (5 x 5 centimetres).
Projects with a high GSD will have less visible details. In extreme cases, your project may look
like an old video game.
Figure 4: The image on the left has a GSD of 5cm, and the image on the right has a GSD
of 30cm
Ground sampling distance is affected by the height you fly at and the camera specifications. A
lower height, with the same camera, means lower ground sample distance and more detailed
outputs.
Figure 5: Altitude affects ground sample distance. Lower height means a lower ground
sample distance.
However, flying lower means taking more time (and batteries) to capture the same area.
Differences in terrain and changes in the camera angle mean that images in the same project
may have different GSDs.
What ground sampling distance you need depends on the needs of your project. If you are
modelling a complex structure, you need a low GSD to capture as many details as possible.
Before taking off, decide the right GSD and define the flight height. The required height that
is needed to obtain a given GSD depends on the camera focal length, sensor width and image
width.
Figure 6: Calculating the right flight height involves the relationship between the sensor
width, focal length and image width.
Accuracy is one factor influencing the overall quality of a data set. According to Foote and
Huebner (1997) describes how similar a data set is to the real world or true values. Error is a
specific measurement of the difference between a value is a data set and the corresponding true
values. Goodchild et al. (1994) define accuracy as the difference between values recorded in a
special data set and modelled or assumed values. They define an error as the difference between
the data set values and true values. When dealing with continuous phenomena and their
representation in GIS as surfaces, it is impossible to measure all true values and hence calculate
all errors. So, the true value must be modelled or estimated. So, in the case of continuous
phenomena, one deals with accuracy indices derived from a limited number of error
measurements. (Ravibabu, M. V., & Jain, 2008)
Accuracy analysis was performed with the sample point surveying in the study area and
comparing them with generated DEM.
The RMSE method is based on the removal of errors using the 3-sigma rule. The 3-sigma rule
removes all errors greater than 3 standard deviations until either, the 5% of data has been
removed, or there are no more errors greater than 3-sigma. Once these errors have been
removed, the vertical accuracy is reported at 68% confidence level as direct value of the RMSE
and at 95% confidence level as per the formula:
PALSAR was one of three instruments on the Advanced Land Observing Satellite-1 (ALOS),
also known as DAICHI, developed to contribute to the fields of mapping, precise regional land-
coverage observation, disaster monitoring, and resource surveying. ALOS was a mission of the
Japan Aerospace Exploration Agency (JAXA).
STUDY AREA
3.1.1 STUDY AREA 1: GEOMATICS ENGINEERING DEPARTMENT
The Area of study is Geomatics Engineering department building and surroundings inside the
campus of Indian Institute of Technology Roorkee.
3.1.2 STUDY AREA 2: KANLOG (HILLY TERRAIN)
The study area Kanlog is located in Shimla Himachal Pradesh lies between longitude 77.00"
and 78.19" East and latitude 30.45" and 31.44" north, with its headquarters in Shimla. It is
surrounded by Mandi and Kullu in the north, Kinnaur in the east, Uttarakhand in the
southeast, Solan to the southwest and Sirmaur in the south. The elevation of the district ranges
from 300 metres (984 ft) to 6,000 metres (19,685 ft).
3.1.3 STUDY AREA 3: OLD ROORKEE (FLAT TERRAIN)
Roorkee is located at 29.87°N 77.88°E. It has an average elevation of 268 metres (879 feet).
Roorkee is 165 kilometres (103 mi) north of the Indian capital, New Delhi, between the
rivers Ganges and Yamuna, close to the foothills of the Himalayas. Before the creation
of Uttarakhand on 9 November 2000, Roorkee was a part of the state of Uttar Pradesh.
CHAPTER 4
METHODOLOGY
UAV derived
DEM with
GCPs
UAV
derived
DEM
UAV derived
UAV derived
DEM
DEM with
without
RTK
GCPs
•Alos palsar DEM
(12.5 m)
•Cartosat DEM (30 m)
UAV derived openly •Aster DEM (30m)
DEM availabe DEM •SRTM DEM (30 m)
Figure 7 Flowchart
3.2 DATA COLLECTION
3.2.1 OPENLY AVAILABLE DEMS
In this project we are using the ASTER, SRTM, ALOS PALSAR and CARTOSAT
DEMs as data sets. Nowadays many websites are available for free satellite data Such as USGS
Earth Explorer, ESA, VERTEX, BHUVAN etc.
Data sets which are used in this project downloaded from the following websites-
selecting their Open hub option. It is easy to collect data from that website simply register on
the website and then select your region, satellite and time period for which you required the
data set and click on search.
The DEMs is download for study area 2 (Kanlog-hilly terrain) and study area 3 (Old Roorkee-
flat terrain) and clip the relevant area from DEM using ArcMap. The resolution of Aster, SRTM
and Cartosat are 30m and Alos palsar is 12.5 m.
Specifications of the Camera model FC6310 are as follows: It has a 1” CMOS sensor with
effective pixels of 20 MP and image size varies from 5472 × 3648, 4864 × 3648 and 5472 ×
3078. The lens is having a FOV 84° 8.8 mm/24 mm (35 mm format equivalent) f/2.8 - f/11
autofocus at 1 m - ∞.
Flight Planning
Depending upon the area of surveying an overlap of 80% and side-lap of 70% was
selected, for which images were taken from a various height for the different study area. The
flight plan was designed using Pix4D capture app for the desired area. There are various
applications and algorithm available in the market. For analysis purpose, we have done the
same planning using three different applications to check the usability and versatility for
mapping and surveying platform, in which we have chosen Pix4D mapper as a better alternative
than the DJI’s autopilot software.
Various options that we can perform in Flight planning software are as follows:
These are the various parameters that are available in the application for designing a flight plan
for a project. These are derived from the basic photogrammetric principals that were initially
used at the time of aerial photogrammetric projects using aeroplanes mounted with the cameras
facing nadir or downward direction.
Figure 11: Various parameters required for flight planning
GCP Acquisition
As we have discussed earlier that, in our case UAV’s GPS/IMU systems are not
particularly powerful, which gives inaccurate EOPs causing serious inaccuracy so we need
GCPs to remove those inaccuracies.
So, GPS points are required for the exterior orientation of the model. We have used Spectra
SP60 for our GPS surveying of our GCPs or Ground Control Points. The spectra Precision
SP60 GNSS receiver works with Z-Blade GNSS-centric technology for providing a fast and
reliable GNSS positioning exploiting all available GNSS signals to produce the optimal
solution even in challenging environments. Satellite delivered corrections from Trimble RTX
services are available over L-band and IP.
Pix4dMAPPER
This is also one of the leading platforms for 3D reconstruction and mapping especially using
images captured from unmanned aerial vehicle or UAVs.
Input + Process
Use any camera and lens, from any angle, and include geotags and GCPs for precise
geo-referencing. Choose your processing options and assess the quality and completeness of
your input images at all stages, even while still in the field. Processing is made easy with the
fully automatic workflow.
Assess + Edit
Keep full control of your projects with these integrated editing tools:
The rayCloud editor combines the 3D point cloud with the original images and is a ground-
breaking concept that offers you a completely new viewing and annotating experience. Use the
rayCloud editor to view, assess, interpret and improve the quality and accuracy of your results.
And measure, vectorize, classify and perform GIS and CAD functionalities directly in
the rayCloud editor.
The Mosaic editor is your answer to beautiful maps with a few clicks. Improve the quality of
the automatically generated orthomosaic with the easiest seamline editing tool available in
the market.
The Index calculator enables you to create and customize index maps, using your multi-
spectral imagery with radiometric accuracy. Application maps can then be produced by
integrating the results, such as NDVI, into all major AG management software packages.
Output + Share Choose from a variety of formats for your geo-referenced orthomosaics,
DSMs, DTMs and point clouds, texturized 3D models and simplified CAD models. Customize
index maps at any resolution, classify terrain and objects automatically, create objects and
animations directly in the software. Then import generated results and vector objects
seamlessly into a wide variety of GIS, CAD and traditional photogrammetry software
packages, including industry-specific software. (Pix4D, n.d.)
Pix4dmapper processing
Add geotagged UAV Add and mark 3 GCPs Mark other GCPs
images using the rayCloud using the rayCloud
Select images
Run step 1 (initial
properties and Reoptimize
processing)
camera model
ArcMap
ArcMap is the main component of Esri's ArcGIS suite of geospatial processing programs and
is used primarily to view, edit, create, and analyse geospatial data. ArcMap allows the user to
explore data within a data set, symbolize features accordingly, and create maps.
Dataset 1 (Geomatics):
Images captured from UAV are processed in pix4dmapper with GCPs, without GCPs
and with RTK mode. The average ground sampling distance (GSD) is 2.06 cm. total of 54
numbers of geotagged images is used for processing and clip the relevant area from generated
DEM. The DEM resolution is10.3 cm (5×GSD).
Dataset 2 (Kanlog):
Images captured from UAV are processed in pix4dmapper. The average ground
sampling distance (GSD) is 10.65 cm. the total 400 number of geotagged images are used for
processing and clip the relevant area from generated DEM. The DEM resolution is 53.25 cm
(5×GSD).
Images captured from UAV are processed in pix4dmapper with GCPs. The
average ground sampling distance (GSD) is 5.39 cm. the total 758 number of geotagged images
are used for processing and clip the relevant area from generated DEM. The DEM resolution
is 26.95 cm (5×GSD).
DEM adjustments and error estimation
Prior to the comparison of DEMs, it is important to remove the vertical biases in the
DEMs and vertically adjust them accordingly. The adjustment of vertical shift is crucial before
evaluating the accuracy of DEMs (Berthier et al. 2014; Rott et al. 2014). Without removing the
vertical offsets and biases, the accurate measurement of DEMs would be erroneous and non-
representative. To remove the vertical offsets between GPS and DEMs elevations, the ALOS
PALSAR, CARTOSAT, SRTM and ASTER GDEM DEMs were vertically adjusted by the
mean elevation difference between GPS elevation points and the respective DEM’s elevations.
The vertical adjustments of the three DEMs have also been carried out with respect to UAV
derived DEM’s elevation points by identifying the vertical shifts. After removing the vertical
shifts between DEMs and GPS and UAV derived DEM data, the accuracy of the DEMs were
calculated and compared. The methodology adopted has been simplified and presented in
Figure. (Pandey et al., 2017)
ALOS PALSAR, CARTOSAT,
SRTM, ASTER GDEM
Statistical Analysis
Error of in terms of
RMSE
Figure 15 the methodology adopted to compute the error and accuracy of DEMs
CHAPTER 5
285
280
275
270
265
260
255
250
245
0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 160 170 180
DISTANCE(m)
50
141
1
8
15
22
29
36
43
57
64
71
78
85
92
99
106
113
120
127
134
148
155
162
169
POINTS
200.00
150.00
100.00
50.00
0.00
1 2 3 4 5 6
POINTS
GPS DEM WITH GCPs DEM WITH RTK DEM W/O GCPs
Figure 16
Table 3 GEOMATICS
Figure 17 A) Difference DEM B/W DEM with GCPs v/s DEM W/O GCPs B) Difference Contour B/W DEM with GCPs v/s
DEM W/O GCPs
Figure 18 A) Difference DEM B/W DEM with GCPs v/s DEM with RTK B) Difference Contour B/W DEM with GCPs v/s
DEM with RTK
Analysis and comparison of the accuracy of UAV derived DEM with openly available DEMs
on hilly and flat terrain
4.2 DEMs and Contour of KANLOG
1,920
1,900
1,880
1,860
1,840
ELEVATION(m)
1,820
1,800
1,780
1,760
1,740
1,720
1,700
1,680
0 100 200 300 400 500 600 700 800 900 1,000 1,100 1,200 1,300 1,400 1,500 1,600 1,700
DISTANCE(m)
1,920 1,920
1,900 1,900
1,880 1,880
1,860 1,860
1,840 1,840
ELEVATION(m)
ELEVATION(m)
1,820 1,820
1,800
1,800
1,780
1,780
1,760
1,760
1,740
1,740
1,720
1,720
1,700
1,700
1,680
1,680
0 100 200 300 400 500 600 700 800 900 1,000 1,100 1,200 1,300 1,400 1,500 1,600 1,700 0 100 200 300 400 500 600 700 800 900 1,000 1,100 1,200 1,300 1,400 1,500 1,600 1,700
DISTANCE(m) DISTANCE(m)
UAV derived DEM Alos Palsar DEM UAV derived Dem Cartosat DEM
1,840
1,840 1,830
1,820
1,820
1,810
1,800
1,800
1,790
1,780
1,780
1,770
1,760 1,760
1,750
1,740 1,740
1,730
1,720 1,720
1,710
0 100 200 300 400 500 600 700 800 900 1,000 1,100 1,200 1,300 1,400 1,500 1,600 1,700 0 100 200 300 400 500 600 700 800 900 1,000 1,100 1,200 1,300 1,400 1,500 1,600 1,700
DISTANCE (m) DISTANCE(m)
UAV derived DEM Aster DEM UAV derived DEM SRTM DEM
4.3 DEMs and Contour of OLD ROORKEE
OLD ROORKEE
290
280
270
ELEVATION(m)
260
250
240
230
220
210
0 200 400 600 800 1,000 1,200 1,400 1,600 1,800 2,000 2,200 2,400 2,600 2,800 3,000 3,200
DISTANCE(m)
UAV derived DEM Aster DEM Alos Palsar DEM SRTM DEM
Cartosat DEM
OLD ROORKEE OLD ROORKEE
221 225
220 224
223
219 222
218 221
ELEVATION(m)
ELEVATION(m)
220
217 219
216 218
217
215 216
214 215
214
213 213
212 212
211
211
0 200 400 600 800 1,000 1,200 1,400 1,600 1,800 2,000 2,200 2,400 2,600 2,800 3,000 3,200 0 200 400 600 800 1,000 1,200 1,400 1,600 1,800 2,000 2,200 2,400 2,600 2,800 3,000 3,200
DISTANCE(m) DISTANCE(m)
UAV derived DEM Alos Palsar DEM UAV deived DEM Cartosat DEM
ELEVATION(m)
ELEVATION(m)
265 250
260
255 245
250 240
245
235
240
235 230
230 225
225
220
220
215 215
0 200 400 600 800 1,000 1,200 1,400 1,600 1,800 2,000 2,200 2,400 2,600 2,800 3,000 3,200 0 200 400 600 800 1,000 1,200 1,400 1,600 1,800 2,000 2,200 2,400 2,600 2,800 3,000 3,200
DISTANCE(m) DISTANCE(m)
UAV derived DEM Aster DEM UAV derived DEM SRTM DEM
Accuracy assessment
ALOS PALSAR, CARTOSAT, SRTM and ASTER GDEM were checked against the GPS
points and also checked against the selected points on UAV derived DEM. The elevation values
of ALOS PALSAR, CARTOSAT, SRTM and ASTER GDEM were extracted for the
corresponding locations of GPS points and selected points on UAV derived DEM. We assessed
the accuracy of DEM as RMSE relative to GPS points and selected points on UAV derived
DEM. The RMSE is defined as the difference between the elevation values of DEMs to be
assessed and the reference height measured in the field (Erasmi et al. 2014). Before
computation of errors, the vertical adjustments of the DEMs with respect to GPS points as well
as selected points on UAV derived DEM have been carried out to remove the vertical shifts
and biases. We report the error as relative (to the GPS as well as UAV derived DEM) and not
in terms of absolute error.
Old Roorkee
4.50
3.94
4.00
3.50
3.00 2.76
RMSE (m)
2.50
1.87
2.00 1.64
1.50
1.00
0.50 0.34
0.00
alos palsar aster cartosat srtm uav
Figure 31 the RMSE value of the five DEMs with respect to GPS points at various observed locations.
On average for the whole study of Old Roorkee, the RMSE of as against GPS measurement
and that of SRTM is 1.64 m and ALOS PALASAR is 2.76 m, and ASTER is 3.94 m, and
CARTOSAT is 1.87 m. The RMSE of SRTM is closer to CARTOSAT DEM but both deviate
significantly from the result of ASTER GDEM.
Evaluation of ALOS PALSAR, CARTOSAT, SRTM and ASTER GDEM with UAV derived
DEM for both hilly and flat terrain
Figure 32 A) KANLOG: Points Selected for Accuracy Assessment B) Old Roorkee: Points Selected for Accuracy Assessment
To evaluate the accuracy of the openly available DEM, the elevation points only from
study areas were selected. The vertical shifts have been removed by vertical adjustments of the
DEMs.
In the study region KANLOG, a total of 73 points were used for validation of openly available
DEM. From these points’ elevation values are extracted from individual DEMs are used for
RMSE computation. The computed RMSE of ALOS PALSAR DEM elevation with respect to
UAV derived DEM elevation was 10.5 m. The RMSE of CARTOSAT, ASTER GDEM and
SRTM were 9.6 m, 14.7 m and 10.5 m, respectively. The elevations from ASTER GDEM
show maximum deviation from the elevation values of UAV derived DEM data as evident
from the analysis. The SRTM, CARTOSAT and ALOS PALSAR, however, showed closer
results. From Figure, it is evident that elevation values from CARTOSAT correspond well with
the elevations from UAV derived DEM data, whereas elevation from ASTER GDEM has
maximum variation.
In the study region Old Roorkee, a total of 67 points were used for validation of openly
available DEM. From these points’ elevation values are extracted from individual DEMs are
used for RMSE computation. The computed RMSE of ALOS PALSAR DEM elevation with
respect to UAV derived DEM elevation was 2.3 m. The RMSE of CARTOSAT, ASTER
GDEM and SRTM were 3.4 m, 5.7 m and 2.5 m, respectively. The elevations from ASTER
GDEM show maximum deviation from the elevation values of UAV derived DEM data as
evident from the analysis. The SRTM, CARTOSAT and ALOS PALSAR, however, showed
closer results. From Figure, it is evident that elevation values from ALOS PALSAR and SRTM
correspond well with the elevations from UAV derived DEM data, whereas elevation from
ASTER GDEM has maximum variation.
16.0 14.7
14.0
12.0 10.5 10.5
9.6
RSME (m)
10.0
8.0
5.7
6.0
3.4
4.0 2.3 2.5
2.0
0.0
alos palsar cartosat aster srtm
Figure 33 the RMSE value of the four DEMs with respect to UAV derived DEM at various points
Quality assessment of DEM
Residual (m)
Residual(m)
10
0
0 210 215 220 225
1600 1700 1800 1900 2000 2100 -5
-10
-10
-20
-30 -15
-40 -20
UAV derived DEM data UAV derived DEM data
Figure 34 Comparison of elevations points from the openly available DEMs with UAV derived DEM data.
CARTOSAT
1900 1900
1800 1800
1700 1700
1600 1600
1600 1700 1800 1900 2000 2100 1600 1700 1800 1900 2000 2100
UAV UAV
1900 1900
SRTM
1800 1800
1700 1700
1600 1600
1600 1700 1800 1900 2000 2100 1600 1700 1800 1900 2000 2100
UAV UAV
Figure 35
KANLOG KANLOG
40 40
20 20
Residual (m)
Residual (m)
0 0
1 6 11 16 21 26 31 36 41 46 51 56 61 66 71 76 1 6 11 16 21 26 31 36 41 46 51 56 61 66 71 76
-20 -20
-40 -40
Points
Points
alos palsar cartosat
KANLOG KANLOG
40 40
Residuals (m)
Residuals (m)
20 20
0 0
1 6 11 16 21 26 31 36 41 46 51 56 61 66 71 76 1 6 11 16 21 26 31 36 41 46 51 56 61 66 71 76
-20 -20
-40 -40
Points Points
aster srtm
Figure 36
Old Roorkee y = 0.4504x + 119.78 Old Roorkee y = 0.2127x + 171.55
R² = 0.2443 R² = 0.1116
224 224
222 222
220 220
Alos palsar
Axis Title
218 218
216 216
214 214
212 212
210 210
210 215 220 225 205 210 215 220 225
UAV UAV
216 216
214 214
212 212
210 210
200 210 220 230 240 210 215 220 225
UAV UAV
Old Roorkee Old Roorkee
10 10
5
5
Residuals(m)
Residuals(m)
0
0 -5 1 6 11 16 21 26 31 36 41 46 51 56 61 66
1 6 11 16 21 26 31 36 41 46 51 56 61 66 -10
-5
-15
-10 -20
Points(m) Points
alos cartosat
Residuals(m)
5
0
0
-5 1 6 11 16 21 26 31 36 41 46 51 56 61 66
1 6 11 16 21 26 31 36 41 46 51 56 61 66
-10 -5
-15
-20 -10
Points Points
aster srtm
S.NO. ALOS
NORTHING EASTING GCPS PALSAR CARTOSAT ASTER SRTM UAV
1 777835.500 3306088.535 218.60 214 219 266 265 218.89
2 778056.475 3306896.475 218.35 217 221 264 267 218.37
3 777792.763 3307940.915 218.86 218 225 272 269 218.76
4 777580.510 3308019.940 218.95 219 221 276 268 218.52
5 778078.153 3307451.247 218.55 217 224 274 267 218.60
6 778320.831 3308781.667 212.64 212 217 267 262 212.69
7 777871.378 3308241.395 218.77 220 223 269 271 219.52
RMSE 2.00 4.06 52.05 49.21 0.35
Figure 38 Graph of Relative Elevation Between ALOS PALSAR and UAV Derived DEM
OLD ROORKEE
Figure 39 a) ALOS PALSAR DEM b) UAV Derived DEM c) Difference DEM Of ALOS PALSAR And UAV Derived DEM
Figure 40 a) ALOS PALSAR contour b) UAV Derived contour c) Difference contour Of ALOS PALSAR And UAV Derived
DEM
Table 7 Absolute Elevation Statics Of Kanlog
Figure 41 42 Graph of Absolute Elevation Between ALOS PALSAR and UAV Derived DEM
Figure 43 Graph of Relative Elevation Between ALOS PALSAR and UAV Derived DEM
4.3.2 ASTER DEM vs UAV derived DEM
KANLOG
Figure 44 a) ASTER DEM b) UAV Derived DEM c) Difference DEM Of ASTER and UAV Derived DEM
Figure 45 a) ASTER contour b) UAV Derived contour c) Difference contour Of ASTER and UAV Derived DEM
Table 9 Absolute Elevation Statics Of Kanlog
KANLOG
2100
y = 1.0059x - 20.131
2000
R² = 0.9768
1900
ASTER
1800
1700
1600
1600 1800 2000 2200
UAV
Figure 46 47 Graph of Absolute Elevation Between ASTER and UAV Derived DEM
Figure 48 Graph of Relative Elevation Between ASTER and UAV Derived DEM
OLD ROORKEE
Figure 49 a) ASTER DEM b) UAV Derived DEM c) Difference DEM Of ASTER and UAV Derived DEM
Figure 50 a) ASTER contour b) UAV Derived contour c) Difference contour Of ASTER and UAV Derived DEM
Table 11 Absolute Elevation Statics Of Kanlog
Figure 51 52 Graph of Absolute Elevation Between ASTER and UAV Derived DEM
Figure 53 Graph of Relative Elevation Between ASTER and UAV Derived DEM
KANLOG
Figure 54 a) CARTOSAT DEM b) UAV Derived DEM c) Difference DEM Of CARTOSAT And UAV Derived DEM
Figure 55 a) CARTOSAT contour b) UAV Derived contour c) Difference contour Of CARTOSAT And UAV Derived DEM
Table 13 Absolute Elevation Statics Of Kanlog
Figure 56 57 Graph of Absolute Elevation Between CARTOSAT and UAV Derived DEM
Figure 58 Graph of Relative Elevation Between CARTOSAT and UAV Derived DEM
OLD ROORKEE
Figure 59 a) CARTOSAT DEM b) UAV Derived DEM c) Difference DEM Of CARTOSAT And UAV Derived DEM
Figure 60 a) CARTOSAT contour b) UAV Derived contour c) Difference contour of CARTOSAT And UAV Derived DEM
Table 15 Absolute Elevation Statics Of Old Roorkee
Figure 61 62 Graph of Absolute Elevation Between CARTOSAT and UAV Derived DEM
Figure 63 Graph of Relative Elevation Between CARTOSAT and UAV Derived DEM
4.4.4 SRTM DEM vs UAV derived DEM
KANLOG
Figure 64 a) SRTM DEM b) UAV Derived DEM c) Difference DEM Of SRTM And UAV Derived DEM
Figure 65 a) SRTM contour b) UAV Derived contour c) Difference contour Of SRTM And UAV Derived DEM
Table 17 Absolute Elevation Statics Of Kanlog
KANLOG uav
2100 2100
2000 y = 1.0064x
R² =- 16.661
0.9883
2000
Elevation
1900 R² = 0.9883
SRTM 1900
1800
1700 1800
1600
1700
1 6 11 16 21 26 31 36 41 46 51 56 61 66 71
Points 1600
1600 1700 1800 1900 2000 2100
Figure 66 Graph of Absolute Elevation Between SRTM and UAV Derived DEM
300
200
100
0
1 3 5 7 9 1113151719212325272931333537394143454749515355575961636567697173
Points
srtm_dif uav_dif
Figure 67 Graph of Relative Elevation Between SRTM and UAV Derived DEM
OLD ROORKEE
Figure 68 a) SRTM DEM b) UAV Derived DEM c) Difference DEM Of SRTM And UAV Derived DEM
Figure 69 a) SRTM contour b) UAV Derived contour c) Difference contour Of SRTM and UAV Derived DEM
Conclusions
1. With the application of RTK in UAV the quality of data acquired in improved in terms
of vertical accuracy without using GCPs. As our RTK is an in house developed product
due to which error still persist which can be improved by using high-end devices.
2. Comparison in plain terrain b/w open source and UAV generated DEM shows that
UAV generated DEM is quite accurate and reliable as compare to another dataset. It
may be concluded that with minimum use of GCPs can improve overall accuracy.
3. In the case of hilly terrain, data behaves in a different manner due to undulating terrain
and forest cover in our study area which places a crucial role in DEM extraction. The
visual interpretation and qualitative analysis of the DEMs for undulating terrain clearly
reveal the effectiveness of ALOS PALSAR DEM in representing detailed
topographical characteristics and terrain features such as steep slopes, peaks, horns and
cliffs followed by SRTM DEM.
4. The performance of UAV derived DEM mostly depends upon points sampled from
imageries and the quality of classification over those points as in case of forested areas
UAV generated point clouds are not capable of classifying forest cover from ground
points due to which the error persists in UAV derived DEM.
OLD ROORKEE
280
260
Elevation
240
220
200
1 4 7 10 13 16 19 22 25 28 31 34 37 40 43 46 49 52 55 58 61 64 67
Points
srtm uav
Figure 70 71 Graph of Absolute Elevation Between SRTM and UAV Derived DEM
5
Elevation
0
1 3 5 7 9 1113151719212325272931333537394143454749515355575961636567
-5
-10
Points
srtm_diff uav_diff
Figure 72 Graph of Relative Elevation Between SRTM and UAV Derived DEM
BIBLIOGRAPHY