0% found this document useful (0 votes)
85 views73 pages

(San and Suzen, 2005 Hohle, 2009) (Kervyn, 2001) (Favey Et Al., 2003) (Schenk, 1996) (Taud Et Al., 1999)

This document discusses digital elevation models (DEMs) and their generation using unmanned aerial vehicles (UAVs). It provides background on DEMs, including what they are, how they are typically generated from data sources like topographic surveys, photogrammetry, satellite imagery, and airborne laser scanning, and some common uses of DEMs. It then discusses UAVs and how their use is increasing for applications like DEM generation due to their low cost and efficiency compared to other data collection platforms. The objectives of the study described are to generate DEMs of study areas using a UAV and analyze the accuracy of the UAV-derived DEMs compared to openly available DEMs, both with and without ground control points.

Uploaded by

Anup
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
85 views73 pages

(San and Suzen, 2005 Hohle, 2009) (Kervyn, 2001) (Favey Et Al., 2003) (Schenk, 1996) (Taud Et Al., 1999)

This document discusses digital elevation models (DEMs) and their generation using unmanned aerial vehicles (UAVs). It provides background on DEMs, including what they are, how they are typically generated from data sources like topographic surveys, photogrammetry, satellite imagery, and airborne laser scanning, and some common uses of DEMs. It then discusses UAVs and how their use is increasing for applications like DEM generation due to their low cost and efficiency compared to other data collection platforms. The objectives of the study described are to generate DEMs of study areas using a UAV and analyze the accuracy of the UAV-derived DEMs compared to openly available DEMs, both with and without ground control points.

Uploaded by

Anup
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
You are on page 1/ 73

CHAPTER 1

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.

The sub-objectives of the study are as follows:

 UAV platform is used to capture aerial images of study areas GEOMATICS


BUILDING, OLD ROORKEE and KANLOG in the purpose of generating the digital
elevation model.
 Analysis and comparison of UAV derived DEM accuracy obtained with GCPs, without
GCPs and RTK mode.
 Analysis and comparison of the accuracy of UAV derived DEM with openly available
DEMs on hilly and flat terrain.
CHAPTER 2

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)

Digital Elevation Model (DEM) is a digital representation of Terrain as a raster (a grid of


squares) of the earth's surface that Stores Earth’s elevation information. DEMs represent a
Convenient way of storing elevation information and of Making such information available to
applications programs Such as GIS. Most frequently the term is used to refer to a set of
elevation data. Hence due to its expanding utilization and Importance, many national
cartographic organizations are Putting their efforts to generate DEMs of different
Characteristics. (Baral, S. S et al.,2010)

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

2.1.2 Unmanned Aerial Systems (UAS)


Systems comprising of a remotely controlled aerial segment also known as unmanned aircraft
vehicle or UAV loaded with various sensors to capture high-resolution data of terrain or any
surface with a ground controlling station can be concisely defined as UAS or unmanned aerial
system, which can further be used for various application of collected data using
photogrammetrically approaches.

An unmanned aerial system basically consists of mainly three components:

1) Unmanned Aerial Vehicle, 2) Communication Datalink, 3) Ground Control Station.


Figure 2: Unmanned Aerial System (Source: Paul Fahlstrom and Thomas Gleason, 2012
Introduction to UAV Systems, the 4th Edition)

2.1.3 Unmanned aerial vehicles (UAV)


An unmanned aerial vehicle, commonly known as a drone, is an aircraft without a
human pilot aboard. UAVs are a component of an unmanned aircraft system; which include a
UAV, a ground-based controller, and a system of communications between the two.

“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)

2.1.4 Advantages of UAVs


Major advantages of UAVs compared to manned aircraft systems are that UAVs can be used
in high-risk situations without endangering human life and inaccessible areas, at low altitude
and at flight profiles close to the objects where manned systems cannot be flown. These regions
are for example natural disaster sites, e.g. mountainous and volcanic areas, flood plains,
earthquake and desert areas and scenes of accidents. In areas where access is difficult and where
no manned aircraft is available or even no flight permission is given, UAVs are sometimes the
only practical alternative. Furthermore, in cloudy and drizzly weather conditions, the data
acquisition with UAVs is still possible, when the distance to the object permits flying below
the clouds. Such weather conditions do not allow the data acquisition with large format cameras
integrated into manned aircraft due to required larger flight altitude above ground. In addition,
one fundamental advantage of using UAVs is that they are not burdened with the physiological
limitations and economic expenses of human pilots. Moreover, supplementary advantages are
the real-time capability and the ability for fast data acquisition, while transmitting the image,
video and orientation data in real time to the ground control station.

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

interpretation. The implementation of GPS/INS systems, as well as the stabilization and


navigation units, allow precise flights, guaranteeing, on the one hand, sufficient image
coverage and overlap and on the other hand, enabling the user to estimate the expected product
accuracy pre-flight. Looking at rotary wing UAVs, the platform allows vertical take-off and
landing vanishing the need for an available runway. Furthermore, the use of VTOL (Vertical
take-off and landing) systems permit the image acquisition on a hovering point, while the
camera is turning in the vertical and horizontal direction.

(Eisenbeiss, H., Grün, A., & Eisenbeiß, 2009)

2.1.4 Limitations in the use of UAVs


UAVs, especially low-cost UAVs, limit the sensor payload in weight and dimension so that
often low weight sensors like small or medium format amateur cameras are selected. Therefore,
in comparison to large format cameras, UAVs have to acquire a higher number of images in
order to obtain the same image coverage and comparable image resolution. Moreover, low-
cost sensors are normally less stable than high-end sensors, which results in reduced image
quality. In addition, these payload limitations require the use of low weight navigation units,
which implies less accurate results for the orientation of the sensors. Furthermore, low-cost
UAVs are normally equipped with less powerful engines, limiting the reachable altitude.
Existing commercial software packages applied for photogrammetric data processing are rarely
set up to support UAV images, as though no standardized workflows and sensor models are
being implemented.

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.

(Eisenbeiss, H., Grün, A., & Eisenbeiß, 2009)

2.1.5 UAV PHOTOGRAMMETRY


UAVs are unmanned aerial vehicles. Hence, UAV photogrammetry can be understood
as a new photogrammetric measurement tool. UAV photogrammetry opens various new
applications in the close-range domain, combining aerial and terrestrial photogrammetry, but
also introduces low-cost alternatives to the classical manned aerial photogrammetry. UAV
Photogrammetry describes photogrammetric measurement platforms, which operate as either
remotely controlled, semi-autonomously, or autonomously, all without a pilot sitting in the
platform, and the photogrammetric processing of UAV images.

The new terminology UAV photogrammetry (Eisenbeiss, 2008c) describes a operates


remotely controlled, semi-photogrammetric measurement platform, which autonomously, or
autonomously, without a pilot sitting in the vehicle. The platform is equipped with a
photogrammetric measurement system, including, but not limited to a small or medium size
still-video or video camera, thermal or infrared camera systems, airborne LiDAR system, or a
combination thereof. Current standard UAVs allow the registration and tracking of the position
and orientation of the implemented sensors in a local or global coordinate system. Hence, UAV
photogrammetry can be understood as a new photogrammetric measurement tool. UAV
photogrammetry opens various new applications in the close-range domain, combining aerial
and terrestrial photogrammetry, but also introduces new (near-) real-time application and low-
cost alternatives to the classical manned aerial photogrammetry. (Eisenbeiss, H., Grün, A., &
Eisenbeiß, 2009)

Table 1 features of aerial, close-ranging and UAV photogrammetry(Eisenbeiss, H., Grün, A., & Eisenbeiß, 2009)

Aerial Close-range UAV


Planning (Semi-)automatic Manual Automatic-manual
Data acquisition/ Assisted/manual Auto/assisted/manual Auto/assisted/manual
flight
Size of the area km2 mm2-m2 m2-km2
Image resolution/ cm-m mm-dm mm-m
GSD
Distance to the 100m-10km cm-~300m m-km
object
Orientation Normal case, Normal/ oblique Normal/ oblique
recently also
oblique
Absolute accuracy cm-dm mm-m cm-10m
of the initial
orientation values
Image block size/ 10-1000 1-500 1-1000
number of scans
Large scale areas Small-scale areas Small and large-scale
(mapping, forestry, and objects areas and objects
glaciology, 3D-city (archaeological (archaeological
Special modelling) documentation, 3D documentation,
applications modelling of monitoring of hazards,
(examples) and buildings) 3D modelling of
features buildings and objects)
Aerial view Terrestrial view Aerial view
Real-time
applications(monitoring)
UAV platforms are nowadays a valuable source of data for inspection, surveillance, mapping
and 3D modelling issues. New applications in the short- and close-range domain are
introduced, being the UAVs a low-cost alternative to the classical manned aerial
photogrammetry. Rotary or fixed-wing UAVs, capable of performing the photogrammetric
data acquisition with amateur or SLR digital cameras, can fly in manual, semi-automated and
autonomous modes. With a typical photogrammetric pipeline, 3D results like DSM/DTM,
contour lines, textured 3D models, vector data, etc. can be produced, in a reasonable automated
way.

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).

Photogrammetry is often the preferred methodology when collecting three-dimensional (3-D)


data using UAVs. Photogrammetry produces 3-D point clouds based on overlapping images.
Other useful by-products can be derived, such as urban façade textures (Leberl et al., 2010).
Specific applications of UAV photogrammetry are presented in (Remondino et al. 2011). The
main photogrammetry steps to generate 3-D elevation models from overlapping images are
presented in Kung et al. (2011):

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)

2.1.6 ACCURACY IN AERIAL MAPPING


2.1.6.1 Relative accuracy
Relative accuracy is accuracy comparing features within a reconstruction. For example,
if a model of a building shows windows two meters apart, and this is the same distance as on
the actual building, the model has high relative accuracy.

2.1.6.2 Absolute accuracy


Absolute accuracy is the accuracy of the reconstruction in relation to its true position
on the planet. If the same model of the building is not in the correct place on the base map, it
has low absolute accuracy, even if it has high relative accuracy.

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

2.1.6.3 Expected accuracy


Are your reconstructions as accurate as possible? The accuracy depends on the quality
of the project which depends on a lot of different things - such as the overlap between images,
the visual content of the images and more.
Generally, in a project which has been correctly reconstructed, you can expect the relative
accuracy to be within one to three times the ground sampling distance (GSD). The absolute
accuracy of a correctly reconstructed project is one to two GSD horizontally (X and Y) and
one to three GSD vertically (Z).

Ground sampling distance (GSD)

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.

The right flight height for the right GSD

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.

To calculate flight height:

H [m] = (ImW * GSD * F) / (SW * 100)

2.2 ACCURACY ASSESSMENT OF DEM


No DEM is absolutely correct, just like every map, elevation models contain errors.
Error all correction should be applied in order to eliminate the error sources. The differences
of elevations on test points are calculated to obtain the statistical definition of their accuracy.
(Joseph, 1984)

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:

Accuracy = RMSE × 1.9600.

Accuracy measures for DEMs presenting the normal distribution of errors.

2.3 Openly Available DEM Characteristic and Accuracy


2.3.1 Cartosat-DEM Characteristics and Accuracy
Indian Space Research Organization (ISRO) has launched Cartosat-1 on May 5, 2005,
with the prime objective of delivering high-resolution satellite data of 2.5 m in-track stereo.
One of its mission goals is to generate a DEM and corresponding ortho-image for the entire
country to facilitate large scale mapping and terrain modelling applications. Cartosat-1 has
completed nine years of its operation successfully and has acquired images over India and
across the globe. Cartosat-1 satellite has two panchromatic cameras with 2.5 m spatial
resolution, to acquire two images simultaneously, one forward-looking (Fore) at +26 degrees
and another rear looking (Aft) -5 degrees for near instantaneous stereo data. The time difference
between the acquisitions of the same scene by two cameras is about 52 seconds. The spacecraft
body is steerable to compensate the earth rotation effect and to force both Fore and Aft cameras
to look at the same ground strip when operated in stereo mode. Simultaneous stereo pair
acquisitions are of an immense advantage since the radiometric parameters of the images are
identical. The stereo pairs have a swath of 26 km and a fixed B/H ratio of 0.62. The satellite
covers the same area in a specified interval of 126 days. The roll and tilt capability of Cartosat-
1 can be used to increase the viewing frequency, which varies with latitude. The revisit
capability at the equator is 5 days.

2.3.2 SRTM-DEM Characteristics and Accuracy


The SRTM-DEM (30m) has been acquired from the Consultative Group on
International Agricultural Research and Consortium for Spatial Information. The Shuttle Radar
Topography Mission (SRTM) 1-arc second DEM (approx. 30m resolution) is the result of a
collaborative effort by the National Aeronautics and Space Administration (NASA), the
National Imagery and Mapping Agency (NIMA), the German Aerospace Centre (DLR), and
Italian Space Agency (ASI) (Foni et al.,2004, Jenson et al.,1998 and Sharma et al.,2010). The
mission was launched on 11 February 2000 aboard the Space Shuttle Endeavour. Using radar
interferometry, a 3-arc second (SRTM-3) DEM was produced covering almost 80% globe
excluding polar regions (Table 1). Initially, a 1-arc second data product was also produced but
was not available for all countries. However, since January 2015 NASA is providing the 1-arc
second data freely for many countries including India. The data currently is being distributed
by NASA/USGS (finished product) contains ‘no-data’ termed as voids where water or heavy
shadow prevented the quantification of elevation. These are generally small holes, which
nevertheless render the ‘no- data’ especially in fields of hydrological modelling. Later, through
further processing, the original DEMs have filled in these no-data voids. Data were collected
using two interferometers, C-band (American) and X-band (German) systems, at 1-arc second
(30 m) (Foni et al.,2004, Jenson et al.,1998 and Sharma et al.,2010).

2.3.3 ASTER-GDEM Characteristics and Accuracy


The Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER)
instrument was launched onboard NASA’s Terra spacecraft in December 1999. The ASTER
GDEM, made available in June 2009, was generated using stereo- pair image collected by the
ASTER instrument onboard Terra. ASTER GDEM coverage spans from 83° north latitude to
83° south, encompassing 99% of Earth's landmass. It has an along-track stereoscopic capability
using its near-infrared spectral band and its nadir-viewing and backward viewing telescope to
acquire stereo image data with a base-to-height ratio of 0.6. One nadir-looking ASTER visible
and near-infrared (VNIR) scene consists of 4,100 samples by 4,200 lines corresponding about
60 km x 60 km ground area. The accuracies estimated for the ASTER- GDEM is 20m for the
elevation values and 30m for horizontal positioning. This assessment concluded that the actual
elevation accuracy of GDEM tiles is within or close to the stated accuracy of 20 meters at 95%
confidence.
2.3.4 ALOS PALSAR-DEM Characteristics and Accuracy
From 2006 to 2011, PALSAR's L-band synthetic aperture radar (SAR) yielded detailed,
all-weather, day-and-night observation, as well as repeat-pass interferometry. PALSAR data
are from multiple observation modes with variable polarization, resolution, swath width, and
off-nadir angle.

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).

2.4 LITERATURE REVIEW


CHAPTER 3

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

DEM deriving from UAV photogrammetry

Data Data Data


Collection Processing Presentation
• Hardware • Pix4dmapper • Digital
• Flight planning software Elevation
• GCPs Model (DEM)

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)

Clipping relevant portion of DEM (Arcmap)

Extracting DEM orthometric height after overlaying


the selected points for comparision on them
(ArcMap)

Difference of openly available DEMs with UAV


derived DEM

statical analysis for vertical accuracy

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-

ASTER AND SRTM DEMs- USGS Earth Explorer (https://earthexplorer.usgs.gov/)

ALOS PALSAR DEM-VERTEX (https://vertex.daac.asf.alaska.edu/?tour=granule)

CARTOSAT DEM- BHUVAN (http://bhuvan.nrsc.gov.in/bhuvan_links.php)

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.

3.2.2 UAV DERIVED DEM


Hardware
UAV used for the purpose of Digital surface model creation was DJI Phantom 4 pro
which is one of the states of the art in drone technology was provided by the supervisor, as
shown below are three main components of the drone i.e. Aircraft body, remote control and
propellers.

Figure 8: DJI Phantom 4 pro components


DJI ‘Phantom 4 Pro’ unmanned aerial vehicle (UAV) is used to capture the area of interest.
This profession model is equipped with a Gimbel HD camera, IMU data logger (GPS + IMU)
and collision avoidance system.

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 - ∞.

Figure 9: DJI Phantom 4 pro Toolkit

Table 2. DJI Phantom 4 Pro Specifications


GPS instruments with better accuracy are required for Absolute Orientation or Geo-referencing
of the relatively recreated geometry by deriving Ground Control Points or GCPs. With
sophisticated RTK and Trimble RTX technology Spectra Precision SP60 is one of the best
GNSS receiver offering a high level of flexibility for any desired project requirements. With a
combination of Z-Blade GNSS-centric technology and L-band capability with satellite-derived
Trimble RTX correction services, GPS receiver can give us the most reliable measurements
with high accuracy.

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:

Figure 10: Flight planning options

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.

Figure 12 The Spectra Precision SP60 GNSS receiver


3.3 DATA PROCESSING
Mainly UAV images processing is done by Pix4Dmapper and further reduction of all
datasets are done by ArcMap software.

Pix4dMAPPER

Pix4Dmapper is a complete mapping and modelling solution to convert thousands of images


into geo-referenced 2D mosaics and 3D models.

This is also one of the leading platforms for 3D reconstruction and mapping especially using
images captured from unmanned aerial vehicle or UAVs.

Figure 13 Basic workflow of Pix4D mapper

Pix4Dmapper automatically processes terrestrial and aerial imagery acquired by light-weight


UAV or aircraft using its innovative technology based purely on image content. This desktop
software converts images into highly accurate, precise, customizable and timely orthomosaics,
surface models, point clouds, textured 3D and simplified CAD models.

Based purely on image content, Pix4Dmapper uses an automated workflow to process


terrestrial and aerial imagery and convert it into highly precise, customizable results for a wide
range of GIS and CAD applications.

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

Open new project Add other GCPs using


Reoptimize the GCP/MTP
On Pix4dmapper manager

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

Run steps 2 (point


Select output Select processing cloud and mesh)and 3
coordinate system option template (DSM, orthomosaic
and index)

Figure 14 flowchart of processing from Pix4dmapper sowtware

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.

The following tools of ArcMap are used:-

 Creating boundary shapefile for all study area.


 Clipping of relevant area of DEM is done by option “extract by mask” in Arc Toolbox.
 Creating point shapefile using the “fishnet” option in Arc Toolbox.
 Extracting DEM orthometric height after overlaying the points on them using “extract
values to the points” option in Arc Toolbox.
 Difference DEMs are generated from selected points.
 Contour is generated for DEMs for all datasets.
 Map preparation is also done in ArcMap.

Datasets deriving from UAV photogrammetry

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).

Dataset 3 (Old Roorkee):

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

UAV DERIVED DEM,


GPS ELEVATIONS

Bias and Offset Removal

Vertically adjusted ALOS


PALSAR, CARTOSAT,
SRTM, ASTER GDEM w.r.t
DGPS and UAV derived
DEM

Statistical Analysis

Error of in terms of
RMSE

Figure 15 the methodology adopted to compute the error and accuracy of DEMs
CHAPTER 5

RESULTS AND DISCUSSION


Analysis and comparison of UAV derived DEM accuracy obtained with GCPs, without GCPs
and RTK mode
4.1 Geomatics
GEOMATICS
310
305
300
295
290
ELEVTION(m)

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)

UAV derived DEM_WO_GCPs UAV derived DEM_W_RTK


UAV derived DEM_W_GCPs
GEOMATICS
320.00
310.00
ELEVATION 300.00
290.00
280.00
270.00
260.00
250.00
240.00

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

DJI_WO_GCPs DJI_RTK DJI_W_GCPs

MIN MAX AVG STDEV RMSE (m)


DEM WITH GCPs 258.83 270.75 262.67 2.23 0.25
DEM WITHOUT 304.59 315.01 307.86 2.31 44.72
GCPs
DEM WITH RTK 244.60 258.19 249.56 2.95 12.65
Geomatics
350.00
300.00
250.00
ELEVATION

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

True Elevation on GCPs Extracted Elevation from DEM on GCPs

DEM W/O DEM WITH DEM WITH


S.No. Easting Northing Elevation GCPs GCPs RTK

1 780141.75 3307048.06 262.50 307.08 262.59 250.94

2 780181.64 3307076.81 262.13 307.02 261.97 249.33

3 780178.70 3307116.67 262.50 307.22 262.17 248.56

4 780143.86 3307079.18 262.54 307.26 262.58 250.10

5 780177.46 3307088.91 262.52 307.14 262.10 249.01

6 780164.31 3307035.32 262.59 307.40 262.78 251.16

RMSE 44.72 0.25 12.65

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

Figure 19 DEMs and Contour of KANLOG

Figure 20 DEMs and Contour of KANLOG


Figure 21 DEMs and Contour of KANLOG

Figure 22 DEMs and Contour of KANLOG


Figure 23 DEMs and Contour of KANLOG
KANLOG PROFILE
1,940

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)

UAV derived DEM SRTM DEM Cartosat DEM Aster DEM


Alos Palsar DEM

KANLOG PROFILE KANLOG PROFILE

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

KANLOG PROFILE KANLOG PROFILE


1,940 1,930
1,920
1,920 1,910
1,900
1,900 1,890
1,880
1,880 1,870
1,860
1,860 1,850
ELEVATION(m)
ELEVATION(m)

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

Figure 24 DEMs and Contour of Old Roorkee

Figure 25 DEMs and Contour of Old Roorkee


Figure 26 DEMs and Contour of Old Roorkee

Figure 27 DEMs and Contour of Old Roorkee


Figure 28 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

OLD ROORKEE OLD ROORKEE


290 270
285 265
280
260
275
270 255

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

Figure 29 alos_uav diff


Figure 30 aster

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

OLD ROORKEE KANLOG

Figure 33 the RMSE value of the four DEMs with respect to UAV derived DEM at various points
Quality assessment of DEM

KANLOG Old Roorkee


40 15
30 10
20
5

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

aster alos palsar cartosat srtm alos cartosat aster srtm

Figure 34 Comparison of elevations points from the openly available DEMs with UAV derived DEM data.

KANLOG y = 0.9941x + 10.377 KANLOG y = 1.0217x - 40.076


R² = 0.9884 R² = 0.9906
2100 2100
2000 2000
Alos palsar

CARTOSAT

1900 1900
1800 1800
1700 1700
1600 1600
1600 1700 1800 1900 2000 2100 1600 1700 1800 1900 2000 2100
UAV UAV

KANLOG y = 1.0059x - 11.078 KANLOG y = 1.0064x - 11.629


R² = 0.9768 R² = 0.9883
2100 2100
2000 2000
ASTER

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

old roorkee y = 0.1179x + 192.21 Old Roorkee y = 0.3883x + 133.24


R² = 0.1005 R² = 0.1794
224 224
222 222
220 220
218 218
SRTM
Aster

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

Old Roorkee Old Roorkee


15 10
10
5
Residuals(m)

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

Table 4 OLD ROORKEE

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

4.3.1 ALOS PALSAR DEM vs UAV derived DEM


KANLOG
Table 5 Absolute Elevation Statics Of Kanlog

Min Max Mean STDEV RMSE


ALOS PALSAR 1602 2033 1818 97.50
UAV 1635 2067 1835 97.49
DIFFERENCE -55 -9 -37 10.53 38.0
Figure 37 Graph of Absolute Elevation Between ALOS PALSAR and UAV Derived DEM

Table 6 Relative Elevation Statics Of Kanlog

Min Max Mean STDEV 2 RMSE


ALOS PALSAR 0 431 196 97.50
UAV 0 432 200 97.49
DIFFERENCE -22 24 -4 10.53 11.03

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

Count Min Max Mean STDEV RMSE


ALOS PALSAR 68 211 225 217 3
UAV 68 212 223 218 2
DIFFERENCE 68 -8 13 -1 3 3
DIFFERENCE 64 -5 3 -1 2 2.2

Figure 41 42 Graph of Absolute Elevation Between ALOS PALSAR and UAV Derived DEM

Table 8 Relative Elevation Statics Of Kanlog

Min Max Mean STDEV RMSE


ALOS PALSAR -2 12 4 3
UAV 0 11 6 2
DIFFERENCE -12 9 2 3 3

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

Min Max Mean STDEV RMSE


ASTER 1651 2079 1798 95.79
UAV 1635 2067 1835 97.49
DIFFERENCE -21 41 9 14.85 17.43

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

Table 10 Relative Elevation Statics Of Kanlog

Min Max Mean STDEV RMSE


ASTER 0 428 193 95.79
UAV 0 432 200 97.49
DIFFERENCE -37 25 -7 14.85 17.43

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

Min Max Mean STDEV RMSE


ASTER 255 311 270 8
UAV 212 223 218 2
DIFFERENCE 36 99 52 8 53

Figure 51 52 Graph of Absolute Elevation Between ASTER and UAV Derived DEM

Table 12 Relative Elevation Statics Of Kanlog

Min Max Mean STDEV RMSE


ASTER 0 56 15 8
UAV -7 4 -1 2
DIFFERENCE -63 0 -16 8 18

Figure 53 Graph of Relative Elevation Between ASTER and UAV Derived DEM

4.3.3 CARTOSAT DEM vs 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

Min Max Mean STDEV RMSE


CARTOSAT 1605 2026 1792 94.97
UAV 1635 2067 1835 97.49
DIFFERENCE -69 -15 -43 9.66 43.85
DIFFERENCE

Figure 56 57 Graph of Absolute Elevation Between CARTOSAT and UAV Derived DEM

Table 14 Relative Elevation Statics Of Kanlog

Min Max Mean STDEV RMSE


CARTOSAT 0 421 187 94.97
UAV 0 432 200 97.49
DIFFERENCE -39 15 -13 9.66 15.98

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

Min Max Mean STDEV RMSE


CARTOSAT 210 226 220 3
UAV 212 223 218 2
DIFFERENCE -12 10 2 3 4
DIFFERENCE -5 8 2 2.7

Figure 61 62 Graph of Absolute Elevation Between CARTOSAT and UAV Derived DEM

Table 16 Relative Elevation Statics Of Old Roorkee

Min Max Mean STDEV RMSE


CARTOSAT -8 8 2 3
UAV -7 4 -1 2
DIFFERENCE -11 11 -3 3 4

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

Min Max Mean STDEV RMSE


SRTM 1650 2073 1840 96.31
UAV 1635 2067 1835 97.49
DIFFERENCE -19 37 5 11 11.60

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

SRTM uav UAV

Figure 66 Graph of Absolute Elevation Between SRTM and UAV Derived DEM

Table 18 Relative Elevation Statics Of Kanlog

Min Max Mean STDEV RMSE


SRTM 0 423 190 96.31
UAV 0 432 200 97.49
DIFFERENCE -34 22 -10 11 14.52
KANLOG
500
400
Elevation

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.

Table 19 Absolute Elevation Statics Of Old Roorkee

Min Max Mean STDEV RMSE


SRTM 261 276 267 3
UAV 212 223 218 2
DIFFERENCE 42 64 49 3 49

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

Table 20 Relative Elevation Statics Of Old Roorkee

Min Max Mean STDEV RMSE


SRTM -8 7 -2 3
UAV -7 4 -1 2
DIFFERENCE -14 8 1 3 3
OLD ROORKEE
10

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

You might also like