Beruflich Dokumente
Kultur Dokumente
Assignment 3
Geometric Correction, Orthorectification,
and Mosaicking
Matt Reaume
NORTHERN GEOSPATIAL SOLUTIONS
Page |1
Technical Memorandum
Project:
Client:
Subject:
Date:
1. Introduction
Aerial remotely sensed image data can be misrepresented on the Earths surface due to the senor
and the nature of the scene that the sensor is viewing, which can cause a great number of distortions
(Finlay, 2015). Systematic Distortions include: Earth rotation, Earth curvature, Platform velocity,
Mirror scan velocity, Scanskew, and Panoramic distortion. Non-Systematic Distortions include:
altitude and attitude. Terrain Distortions include: displacement caused by topographic relief.
Systematic Distortions are predictable and are removed by preprogrammed data processors, NonSystematic Distortions are unpredictable and are removed by geometric correction, and Terrain
Distortions are corrected by ground elevation and/or DEM are needed to remove this error (Finlay,
2015). Geometric correction is used to remove these distortions by measuring known control
points and transforming the image coordinates into ground coordinates. This rectification process
is accomplished when mosaicking images; therefore, the objective of this assignment is to gain a
familiarity with the fundamentals of orthorectification and mosaic processes, which the technical
memorandum explains in detail (Finlay, 2015).
2. Purpose
The orthorectification process will be completed on three different aerial photographs. These aerial
photographs will be mosaicked to produce one image over the whole area using two different
mosaicking methods, which are MosaicPro and Mosaic Express. In order to show the advantages
of the orthorectification process, a geometric correction will be performed on the images and the
mosaicked results will be compared (Finlay, 2015).
Page |2
3. Methodology
Geometric correction removes distortions by measuring the image coordinates of ground control
points and then computing the coefficients, which is used to transform the image coordinates (pixel
values) into ground reference coordinates in metres (Finlay, 2015). Three aerial photographs are
geometrically corrected (using the polynomial model) to the provided subset image in this
assignment. Each photograph has ten ground control points for the polynomial geometric
correction. Upon applying ten ground control points (GCPs), the three aerial photographs are
clipped accordingly to eliminate the fiducials and text on each photograph. The mosaicking
process comes last using MosaicPro and Mosaic Express to output a mosaicked image using two
different methods. These mosaicked outputs will be compared in Section 4, Mosaic Analysis.
3.1 Prediction Process after Entering 4 or more GCPs
After entering four or more Ground Control Points (GCPs) the prediction process was resulting in
close to accurate localization of the GCPs in most cases, along with GCPs that were very far off
from one another. The prediction process is based upon the 1st degree polynomial mathematical
equation of polynomial transformation, which is explained by the pixel coordinates (x, y) and
the map coordinates (x, y). Each parameter (highlighted red below in the Mathematical Equation)
is effected by shift and scale in both the x and y direction, along with rotation. It can be said that
the prediction process is based upon this mathematic equation, and ERDAS Imagines ability to
match the GCPs based upon reflectance values.
Mathematical Equation: x = A1 x+ B1 y + C1 xy+ D1
y = A2 x+ B2 y + C2 xy + D2
Shown below in Figure 1 is a Ground Control Point that has been predicted after four GCPs were
entered. As you can see there is a high precision and accuracy, and this process was repeated
throughout the three aerial photographs that portrayed close to matching GCPs.
Page |3
The result of the three aerial photographs are shown below in Figure 3 4, and 5 that illustrates all
ten of the Ground Control Points.
Page |4
The RMS Error is the distance between the actual output location for a GCP and the desired output
location for the same point. This means that this is the difference between the actual source location
and the output control point for the same point. As mentioned above, Figure 5 displays the Total
Control Point Error for the third aerial photograph; Table 1 below, displays all three of the Total
Control Point Errors.
Table 1: Total Control Point Error for the 3 Photos
Aerial Photograph
1.
2.
3.
Page |5
During the creation process of the resampled images, the output cell size shown in Figures above
indicate the pixel size of all the three corrected photos are the same at 0.5 when rounded. The
output cell sizes for all three photographs are: photo #1 (X: 0.5027254608, Y: 0.5027254608),
photo #2 (X: 0.5013567361, Y: 0.5013567361), photo #3 (X: 0.5001135909, Y: 0.5001135909).
The reason these pixel sizes are slightly off from one another is because of the resample method
chosen (Cubic Convolution). In this method, the transferred synthetic pixel values are determined
by evaluating the block of 16 pixels in the input matrix that surrounds each output pixel. It also
avoids the disjointed appearance of the nearest neighbor method and provides a slightly sharper
image than the bilinear interpolation method (Lillesand, Kiefer, and Chipman, 2008, pp. 488). The
matching output cell size ensures that the three aerial photographs can be matched together when
mosaicked, which is discussed in Section 3.4.
Page |6
As shown below in Figure 11 and Figure 12, the road network and buildings vector layer are
accurately place on top of the aerial photographs. This is caused by having a lot of GCPs in the
area, as well as a low RMS error. These specific areas had more GCPs because it was easier to
select a GCP on the aerial photograph and precisely match the same GCP to the subset image.
Page |7
4. Mosaic Analysis
4.1 Method 1: Mosaic Pro
MosaicPro from 2D View is an application in ERDAS Imagine for combining multiple images
into single (or tiled) color balanced and compressed ortho-mosaic output files. This option
automatically loads all of the image data present in the active 2D View. Before the files can be
mosaicked, the images must be clipped to remove the fiducials and test surrounding the image. As
seen in Figure 13, there is text clearly visible around the images.
The drawing tool and subset and chip tool was used to eliminate the text from the images. The
final results are shown in Figure 14, which illustrates the same images without the fiducials and
text.
Page |8
Figure 14: Clipped Aerial Photographs
Once these files are loaded into MosaicPro window, an outline of the images is displayed as shown
in Figure 15. For further viewing purposes these images can be visible or not, such as in this case
where the images are outlined showing the exact points of overlap.
After the mosaicking process was complete using MosaicPro, the end result can be seen below on
the following page in Figure 16. There is also noticeable overlap on the mosaicked output which
is caused by the images being placed directly on top of each other. The Mosaic Express method
will eliminate the overlap as shown and discussed in Section 4.2.
Page |9
P a g e | 10
After the mosaicking process was complete using Mosaic Express, the end result can be seen on
the following page in Figure 19. The only lines that are displayed are the fiducials and text that do
not interfere with the overall image. There is also no overlap on the mosaic output as opposed to
the output in MosaicPro.
P a g e | 11
P a g e | 12
Figure 20: MosaicPro Farmland Shift and Overlap (left) and Mosaic Express Farmland No Shift and Visible Cutline (right)
The road network also produced a different result between the two methods. MosaicPro and
Mosaic Express illustrated a shift, but MosaicPro produced a slightly larger shift than Mosaic
Express method. This output can be seen in Figure 21.
P a g e | 13
Figure 21: MosaicPro Road Network Large Shift (left) and Mosaic Express Road Network Small Shift (right)
5. Conclusion
This assignment provided the fundamentals needed to produce a mosaic output of two methods,
which were MosiacPro and Mosaic Express. Both of these methods proved be reliable, but with
slightly different results. MosaicPro successfully merged the three orthorectified aerial
photographs, but with some overlap and shift between the road network and building vector layer
on top of the aerial photographs. The Mosaic Express method merged the aerial photographs
successfully as well, but with no overlap and slightly less shift. Upon completing this assignment,
the Mosaic Express method served to be the most effective due to minimal shift and overlap, which
resulted in a realistic depiction of the ground surface.
P a g e | 14
6. References
Finlay, J. (2015). Assignment_Number_3_2015. GISC9216 Digital Image Processing.
Assignment #3, Geometric Correction, Orthorectification, and Mosaicking. Niagara College,
Niagara-on-the-Lake.
Lillesand, T. M., Kiefer, R. W., & Chipman, J. W. (2008). Remote Sensing and Image
Interpretation (Sixth Edition). Hoboken, New Jersey, United States of America: John
Wiley & Sons, Inc.
Niagara College. (n.d.). From: X:\GIS Resources\GIS - Second Semester\GISC9216
DIP\Week6 Geometric Correction. Niagara-On-The-Lake, Ontario, Canada: Niagara
College Canada.