See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/369090969

A Report on Drone Image Processing Using Agisoft.

Technical Report · February 2023

DOI: 10.13140/RG.2.2.25216.51202

CITATIONS READS
0 1,367

1 author:

Sankalpa Dhakal
Tribhuvan University
14 PUBLICATIONS 0 CITATIONS

SEE PROFILE

All content following this page was uploaded by Sankalpa Dhakal on 09 March 2023.

The user has requested enhancement of the downloaded file.

 Tribhuvan University
Institute of Engineering
Pashchimanchal Campus
Lamachaur, Pokhara
Department of Geomatics Engineering

A Report on Drone Image Processing Using Agisoft.

Prepared by: Sankalpa Dhakal (PAS075BGE040)

i
Date: Feb 2023
Abstract
This technical report presents the methodology and results of the preparation
of a digital terrain model (DTM) of Paschimanchal Campus Pokhara, using a
Phantom 4 drone and processing through the Agisoft software. The report
outlines the steps taken to plan and execute the drone flights, collect aerial
images, and process the data using photogrammetric techniques to generate a
high-quality DTM of the study area. The report also includes a detailed
description of the Agisoft application and its features, as well as a discussion
of the challenges encountered during the data processing and analysis. The
final DTM produced by this study provides accurate and detailed
information on the topography and features of the study area, which can be
useful for various applications such as land-use planning, urban
development, and natural resource management. The report concludes with
recommendations for future research and improvements to the data
collection and processing methodology to further enhance the accuracy and
usefulness of the DTM.

i
Contents
Abstract ............................................................................................................ i
List of abbreviations ...................................................................................... iii
List of figures:................................................................................................ iv
Introduction......................................................................................................1
Literature review ..............................................................................................2
Study Area .......................................................................................................4
Objectives ........................................................................................................4
Methodology ....................................................................................................5
Drone Flight Planning and Execution .........................................................5
Image Collection..........................................................................................5
Data Processing ...........................................................................................5
Data Analysis .............................................................................................12
Results and Discussion ..................................................................................12
Conclusion .....................................................................................................15
References......................................................................................................16
Appendex ........................................................................................................ A
A) Processing Parameters ........................................................................... A
B) GENERAL ............................................................................................. C
C) Camera Calibration ................................................................................ C

ii
List of abbreviations

DEM Digital Elevation Models

DSM Digital Surface Models
DTM Digital Terrain Models
GCPS Ground Control Points
GIS Geographic Information Systems
LiDAR Light Detection and Ranging
TIN Triangular Irregular Networks

iii
List of figures:
Figure 1: Google Earth Image of Study Area ..................................................4
Figure 2 : Camera locations and image overlap. .............................................5
Figure 3 : Adding Photo on Single Camera Data Layout................................6
Figure 4: Aligning Photos...............................................................................7
Figure 5: Camera locations on Study area .......................................................7
Figure 6: Generating Dense Cloud ..................................................................8
Figure 7: Generated Dense cloud of Study Area .............................................8
Figure 8: Mesh cloud Model of Study Area ....................................................9
Figure 9: Generating DTM ............................................................................10
Figure 10: Generated DEM of Study Area ....................................................10
Figure 11: Creating Orthophoto ....................................................................11
Figure 12: Generated Orthophoto of Study Area ..........................................11
Figure 13: Dem of Paschimanchal Campus ..................................................12
Figure 14: Slope of Paschimanchal Campus .................................................14
Figure 15: Orthomoisaic of Paschimanchal Campus ....................................15

iv
Introduction
A Digital Elevation Model (DEM) is a representation of the bare ground
(bare earth) topographic surface of the Earth excluding trees, buildings, and
any other surface objects which is often used in geographic information
systems (GIS) and other geospatial analysis applications such as land use
planning, environmental assessment, disaster management, and urban
planning.
DTMs are created by analyzing and processing a variety of data sources,
such as aerial photographs, satellite imagery, LiDAR (Light Detection and
Ranging) data, and other remote sensing data. Creating a digital elevation
model (DEM) from drone imagery is a useful technique for mapping terrain
and creating 3D models of terrain surface of the Earth. However, the
processing of drone imagery can be a complex and time-consuming task,
requiring specialized software and technical expertise.
Agisoft Metashape is powerful photogrammetry software that is commonly
used for processing drone imagery and generating high-quality 3D models of
the terrain. This software can be used to create a digital terrain model
(DTM), digital surface model (DSM), and orthophotoplan - orthomosaic and
has many advantages, including the ability to generate high-quality and
accurate 3D models of the terrain surface, the flexibility to control the
density and accuracy of the point cloud, and the ability to evaluate the
accuracy and quality of the resulting DEM.
In this technical report, we present the methodology and results of preparing
a DTM of Paschimanchal Campus, Pokhara, using a Phantom 4 drone and
processing through Agisoft software. The report outlines the process of
drone flight planning and execution, image collection, data processing, and
analysis.

1
Literature review
Orthomosaic: An orthomosaic is a large, high-resolution aerial image that
is geometrically corrected, meaning that it has been adjusted for lens
distortion, camera tilt, and topography to create a planimetrically accurate
image. Orthomosaic images are typically used as base maps for various
applications, including creating DTMs.
Ground Control Points (GCPs): GCPs are points on the ground that are
accurately surveyed and marked so that they can be identified in drone
images. GCPs are used to georeference the images and align them with real-
world coordinates.
Point Cloud: A point cloud is a set of 3D points in space that represent the
surface of an object or terrain. Point clouds are created from drone images
using SfM algorithms, and they are used to generate DTMs.
Digital Surface Model (DSM): A DSM is a digital representation of the
Earth's surface, including all objects on it, such as trees, buildings, and other
structures. DSMs are created by processing drone images and generating a
point cloud that includes all surface features.
Digital Elevation Model (DEM): A DEM is a digital representation of the
Earth's surface that shows only the terrain or ground surface, excluding all
objects on it. DEMs are generated by subtracting the DSM from the DTM,
leaving only the elevation data of the terrain.
Triangulated Irregular Network (TIN): A TIN is a mathematical model
used to represent the surface of the terrain as a series of interconnected
triangles. TINs are used to generate DTMs, as they allow for the
interpolation of elevation data between points in the point cloud.
Orthophoto: A geometrically corrected aerial photograph in which the
image is transformed to have a uniform scale, allowing for accurate
measurements of distances and areas. Orthophotos are often used as a base
layer for DTM generation.
Photogrammetry: The art of capturing high-resolution photographs to
recreate a survey area. These images are processed and stitched together

2
using sophisticated software to create realistic, geo-referenced, and
measurable 3D models of the real world.
Agisoft Metashape: It is a powerful software tool for processing drone
imagery and generating a high-resolution digital elevation model.
Nadir: The point directly below the drone, which is the ideal angle for
capturing images for DTM generation.
Camera Calibration: The process of determining the intrinsic and
extrinsic parameters of a camera to improve the accuracy of the imagery.
Bundle Adjustment: The process of refining the estimated camera
positions and orientations to improve the accuracy of the final DTM.
Triangulation: The process of determining the position of points in space
by measuring the angles between lines that connect the point to two or more
other points.
Image Overlap: The amount of overlap between adjacent images, which is
necessary to generate accurate DTMs.
GeoTIFF: A standard file format used to store georeferenced raster images,
which is commonly used to export DTMs created with Agisoft.

3
Study Area
The study area of our project is Paschimanchal Campus located at Pokhara
Metropolitan city, Kaski, Nepal.

Figure 1: Google Earth Image of Study Area

Objectives

The purpose of this article is to accomplish a Digital Terrain Model, Digital

Surface Model and an orthophotoplan - orthomosaic of an area within the
Paschimanchal Campus, Pokhara.

4
Methodology
Drone Flight Planning and Execution
We used a Phantom 4 drone to capture aerial images of the study area.
Before the flight, we planned the drone's path to ensure maximum coverage
of the study area while maintaining a consistent altitude and image overlap.
The drone was flown at a height of 49 meters with an overlap of 70%
between the images. The flight was conducted during good weather
conditions to ensure clear and high-quality images.

Figure 2 : Camera locations and image overlap.

Image Collection
The images were captured in the form of JPGs and saved on a high-capacity
SD card in the drone. A total of 1500 images were collected for the study
area of Paschimanchal Campus.

Data Processing
The collected images were processed through the Agisoft software.
Processing drone images in Agisoft software to create a Digital Terrain
Model (DTM) and Orthophoto involves involves importing images, aligning
photos, creating a dense point cloud, generating a mesh, creating a DSM and
DTM, and finally, creating an orthophoto., which are explained in detail
below:

5
  Import Images into Agisoft
The first step is to import the images into Agisoft software. We opened the
software and created a new project. Then Clicked on Workflow and select
all photos in a folder that is to be processed selecting single camera data
layout.

Figure 3 : Adding Photo on Single Camera Data Layout

 Aligning the images using the "Align Photos" tool

Once the images are imported, the next step is to align them. This can be
done by going to Workflow > Align Photos. Metashape will use the
overlapping features in the images to calculate the camera positions and
orientation, which are necessary for subsequent processing. The software
will detect the common features in each photo and match them to create a
3D model of the study area.

6
 Figure 4: Aligning Photos

 Optimizing Cameras
During the initial photo alignment process, Metashape estimates the internal
and external orientation parameters, however, optimizing the cameras using
the imported camera geolocations to reduce or minimize the error resulting
from coordinate misalignments and reprojection errors.

Figure 5: Camera locations on Study area

 Generating a dense point cloud using the "Build Dense

Cloud" tool
After aligning the photos, the next step is to create a dense point cloud. This
is done by selecting the "Build Dense Cloud" option. The software will use
the common features detected in the photos to create a dense 3D point cloud
of the study area. Metashape will use the camera positions and orientation
calculated in the previous step to generate a dense point cloud from the
images.
7
 Figure 6: Generating Dense Cloud

Figure 7: Generated Dense cloud of Study Area

 Creating a mesh model using the "Build Mesh" tool

Once the dense point cloud is generated, the next step is to create a mesh of
the study area. This can be done by going to Workflow > Build Mesh.

8
Metashape will use the dense point cloud to generate a 3D mesh. The
software will create a textured 3D model of the study area.

Figure 8: Mesh cloud Model of Study Area

 Generating DSM and DTM using the "Build DEM" tool

The next step is to generate a Digital Surface Model (DSM) and a Digital
Terrain Model (DTM). To generate these models, select the "Build DEM"
option. The software will generate a DSM and DTM of the study area. The
DSM represents the surface of the ground and includes all features such as
trees, buildings, and other structures. The DTM, on the other hand,
represents the bare earth terrain and excludes all above-ground features.

9
 Figure 9: Generating DTM

Figure 10: Generated DEM of Study Area

 Creating an Orthophoto
Finally, to create an orthophoto, select the "Orthomosaic Generation" option.
The software will stitch together the images and create an orthophoto of the
study area. An orthophoto is a georeferenced image that is corrected for

10
perspective and has a uniform scale, which makes it an accurate
representation of the study area.

Figure 11: Creating Orthophoto

Figure 12: Generated Orthophoto of Study Area

11
Data Analysis
The generated DTM was analysed using ArcGIS software to identify the
elevations and contours of the study area. The elevation information was
used to calculate the slope of the terrain to identify the high and low points
of the study area. According to the DEM data, the elevation for campus
premises falls between 918 and 954m.

Figure 13: Dem of Paschimanchal Campus

Results and Discussion

Digital Elevation Model (DEM) is a raster data that can be generated solely
from the aerial images themselves, however, to produce a higher accuracy
product, it is important to place Ground Control Point (GCP) with accurate
coordinates. A total of 1500 images were processed using Agisoft
Metashape software to produce an orthomosaic photo along with Digital
Elevation Model (DEM). The DTM accurately represents the topography

12
and features of the study area, providing accurate elevation data and detailed
contour information.
All the processing took place without using any Ground Control Points
(GCPs). Thus, there is a lot of error accumulation in the obtained data as
shown in the table below
X error Y error Z error XY error Total
(m) (m) (m) (m) error (m)
1.46828 1.75142 20.9087 2.28546 21.0332
X=Longitude Y=Latitude Z=Altitude

In order to minimize such errors, it is recommended to use Ground Control

Points (GCPs). It will be better if those GCPs are taken from different methods
like DGPS, RTK or any other precise technique.

The use of a drone allowed for efficient data collection and reduced
fieldwork time, while the Agisoft software provided accurate and reliable
data processing. However, we encountered challenges during the data
processing stage due to the high volume of images and the need to manage
the processing time.
DEM, Orthomoisaic , Slope and contour map were prepared using the drone
image which are shown in the report.

13
Figure 14: Slope of Paschimanchal Campus

14
 Figure 15: Orthomoisaic of Paschimanchal Campus

Conclusion
The use of a Phantom drone and Agisoft software provides an efficient and
effective way to prepare a high-quality DTM. Further research can explore
ways to improve the data processing methodology and integrate other
datasets to enhance the accuracy and usefulness of the DTM.

15
References
F Arif et al 2018 IOP Conf. Ser.: Earth Environ. Sci. 169 012093
Gonçalves J A, Henriques R. 2015. UAV photogrammetry for topographic
monitoring of coastal areas. ISPRS Journal of Photogrammetry and Remote
Sensing, 104:101–11
https://geodetics.com/dem-dsm-dtm-digital-elevation-models/
https://www.usgs.gov/search?keywords=DEM
Nex F, Fabio R. 2014. UAV for 3D mapping applications: a review. Applied
geomatics, 6(1):1- 15.
Wolf P R, Dewitt B A. 2000 Elements of photogrammetry: with applications
in GIS Vol. 3 (New York: McGraw-Hill).

16
Appendex

A) Processing Parameters

General
Cameras 1500
Aligned cameras 1493
Shapes
Polygon 1
Coordinate system WGS 84 (EPSG::4326)
Rotation angles Yaw, Pitch, Roll
Point Cloud
Points 563,470 of 920,325
RMS reprojection error 0.207211 (0.588118 pix)
Max reprojection error 1.01213 (5.77344 pix)
Mean key point size 2.82163 pix
Point colors 3 bands, uint8
Key points No
Average tie point multiplicity 6.83061
Alignment parameters
Accuracy High
Generic preselection Yes
Reference preselection Source
Key point limit 40,000
Key point limit per Mpx 1,000
Tie point limit 4,000
Exclude stationary tie points Yes
Guided image matching No
Adaptive camera model fitting No
Matching time 14 minutes 11 seconds
Matching memory usage 3.40 GB
Alignment time 17 minutes 56 seconds
Alignment memory usage 1.22 GB
Optimization parameters
Parameters f, b1, b2, cx, cy, k1-k4, p1, p2
Adaptive camera model fitting Yes
Optimization time 21 seconds
Date created 2023:01:27 11:07:48
Software version 1.7.6.13524
File size 111.65 MB
Depth Maps
Count 1493
Depth maps generation parameters
Quality High
Filtering mode Moderate
Max neighbors 40
Processing time 1 hours 33 minutes
Memory usage 4.28 GB
Date created 2023:01:27 13:25:04
Software version 1.7.6.13524
File size 2.92 GB
Dense Point Cloud
Points 94,891,254
Point colors 3 bands, uint8
Page 6
Depth maps generation parameters
Quality High
Filtering mode Moderate
Max neighbors 40
Processing time 1 hours 33 minutes
Memory usage 4.28 GB
Dense cloud generation parameters

A
Processing time 2 hours 48 minutes
Memory usage 4.80 GB
Ground points classification parameters
Max angle (°) 12.5
Max distance (m) 1.3
Cell size (m) 55
Classification time 22 minutes 49 seconds
Classification memory usage 4.86 GB
Date created 2023:01:27 16:13:20
Software version 1.7.6.13524
File size 2.97 GB
Model
Faces 18,846,662
Vertices 9,431,257
Vertex colors 3 bands, uint8
Depth maps generation parameters
Quality High
Filtering mode Moderate
Max neighbors 40
Processing time 1 hours 33 minutes
Memory usage 4.28 GB
Reconstruction parameters
Surface type Height field
Source data Dense cloud
Interpolation Enabled
Strict volumetric masks No
Processing time 10 minutes 42 seconds
Memory usage 3.73 GB
Date created 2023:01:28 05:47:12
Software version 1.7.6.13524
File size 431.49 MB
DEM
Size 25,511 x 23,522
Coordinate system WGS 84 (EPSG::4326)
Reconstruction parameters
Source data Dense cloud
Interpolation Enabled
Processing time 1 minutes 43 seconds
Memory usage 316.96 MB
Date created 2023:01:28 04:55:29
Software version 1.7.6.13524
File size 592.12 MB
Orthomosaic
Size 39,568 x 33,708
Coordinate system WGS 84 / UTM zone 44N (EPSG::32644)
Colors 3 bands, uint8
Reconstruction parameters
Blending mode Mosaic
Surface Mesh

Enable hole filling Yes

Enable ghosting filter No
Processing time 43 minutes 18 seconds
Memory usage 3.41 GB
Date created 2023:01:28 06:34:18
Software version 1.7.6.13524
File size 17.53 GB

System
Softwarename Agisoft Metashape Professional
Software version 1.7.6 build 13524
OS Windows 64 bit
RAM 7.88 GB
CPU Intel(R) Core(TM) i5-6200U CPU @ 2.30GHz
GPU(s) NVIDIA GeForce 940MX

B
 B) GENERAL
Number of
1,500
images:
49.4 m
Flying altitude:
1.84
Ground
cm/pix
resolution:
0.176 km²
Coverage area:

Camera stations: 1,493

Tie points: 563,470
Projections: 3,045,64
Reprojection 6
error: 0.588 pix

Focal Precalibrate
Camera Model Resolution Pixel Size
Length d
FC300S 4000 x 1.56 x 1.56
3.61 mm No
(3.61mm) 3000 μm
Table . Cameras.

C) Camera Calibration
FC300S (3.61mm)
1500 images

Type Resolution Focal Length Pixel Size

Frame 4000 x 3000 3.61 mm 1.56 x 1.56 μm
Value Error F Cx Cy B1 B2 K1 K2 K3 K4 P1 P2
F 2319.68 0.079 1.00 -0.10 -0.83 -0.11 -0.05 0.05 0.02 0.04 -0.07 -0.02 -0.22
Cx -0.161449 0.0091 1.00 0.08 0.03 0.06 -0.01 0.00 -0.01 0.01 0.63 0.01
Cy 14.4501 0.014 1.00 0.06 0.05 -0.08 0.03 -0.08 0.10 0.01 0.40
B1 -5.03751 0.0052 1.00 -0.02 -0.03 -0.01 0.00 0.00 0.00 0.08
B2 0.0618618 0.0046 1.00 0.00 -0.00 -0.00 0.00 0.01 -0.00
K1 -0.00408822 2.8e-05 1.00 -0.92 0.88 -0.83 -0.02 -0.25
K2 -0.0242938 9.1e-05 1.00 -0.99 0.96 -0.00 0.02
K3 0.0553575 0.00012 1.00 -0.99 0.00 -0.03
K4 -0.024357 5.7e-05 1.00 -0.00 0.04
P1 9.33912e-05 1.1e-06 1.00 0.03
P2 0.000362432 1.3e-06 1.00
Table . Calibration coefficients and correlation matrix.

View publication stats