Scribd
Upload
25 views

Lab3 Geometric Correction

correction

Uploaded by

Syiekh Mohd. Amirul Hafizi bin Roslan Jeffery
Copyright
© © All Rights Reserved
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
25 views

Lab3 Geometric Correction

correction

Uploaded by

Syiekh Mohd. Amirul Hafizi bin Roslan Jeffery
Copyright
© © All Rights Reserved
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
You are on page 1/ 23

Lab 2:

Geometric Correction
Laboratory Practical
GSS 612
REMOTE SENSING AND APPLICATION

Nor Aizam Adnan, Ph.D


Ainon Nisa Othman, Ph.D
Nurul Amirah Isa
Universiti Teknologi MARA
What is Geometric
Correction?
• Geometric Correction is a process prior to image processing
to eliminate/minimize the distortions contained in a satellite
images.

• Other than that, the process can be used to project the image
onto a planimetric coordinate system.

• This process is performed during pre-processing stage in


digital image processing workflow.

Two types:
1) Image to Image rectification
2) Image to Map registration
The Importance

• In order to perform highly-accurate analysis, the input digital


map should be free of significant geometric distortion and
conforms to the projection and coordinate system of the
original map.

• Raw images commonly contain internal geometric distortions


that result from the image acquisition process.

• As a result the raw images do not have a simple “map-like”


geometry, and accurate map relationships cannot be derived
from them.
Today’s Objective
• To perform geometric correction (image to image
rectification) on an image which missing its spatial
reference.

• At the end of the laboratory session, the students


should be able to perform the geometric correction
using “Multipoint Geometric Correction” package.

• Students should be able to:


i. Part 1: Examine the image to be corrected
ii. Part 2: Collect the GCPs
iii. Part 3: Perform resampling process
iv. Part 4: Verify the output image
How to do
Datasets to be used

Landsat Images (sensor by US):


Landsat 5 TM

Image to be corrected:
Pkl150700.img

Reference image:
Pkl210291rso.img
How to do (Part 1)
Launch Erdas Imagine
• Launch Erdas Image and add the image to be corrected onto the screen
How to do (Part 1)
Examine the spatial reference of the data
• View the metadata of the image. The Projection Info is “unknown” since the image is
missing its spatial reference.
How to do (Part 2)
Start GCP Tool
• To open the GCP tool, go to Multispectral tab and click on Control Points icon. Since the
image to be corrected is already on the screen, the image is automatically chosen to be
geometrically corrected.

• Select “Polynomial” from Model List.


How to do (Part 2)
Start GCP Tool
• Select the image to be the reference image for the geometric process.
How to do (Part 2)
Start GCP Tool
• Take the default setting for the “Polynomial Order”. Click Close on “Polynomial Model
Properties”.
How to do (Part 2)
Collect the GCPs
• Two-sided window will appear on your screen. The image to be corrected is shown on
the left side of the window whereas the reference image is shown on the right of the
image.

Image to be corrected Reference image

GCPs Table
How to do (Part 2)
Collect the GCPs
• Using icon, start collect the GCP on the image. Click once on the image to be
corrected to add the GCP. Then, click on the same location on the reference image to
add the GCP. Please ensure that the features selected as the GCPs are on the same
locations for both image.

same location
How to do (Part 2)
Collect the GCPs

• Select at least four (4) GCPs to enable the software to calculate the RMS Error in the
GCPs Table.

• This is because, the model used; Polynomial (1st Order), needs at least four points
selected to compute the RMS Error.

• The GCPs selected should be randomly distributed on the image. Please avoid collecting
the GCPs on the beach, rivers or any features that can have major changes due to
natural phenomenon.

• Collect at least 15 GCPs for this practical session.

• The tolerance of the RMS Error is less than 1/2 pixel.


How to do (Part 2)
Collect the GCPs – RMS Error = < 1/2 pixel
How to do (Part 3)
Resample
• Resampling method is used to transform the image into the new coordinate system.
• Click on icon. In the new dialog window appeared, please specify the following
information.
How to do (Part 3)
Resample
• “Process List” window appears indicates that the resampling process is in progress.
• Click close when the process is completed.
How to do (Part 3)
Save the files
• Save the “Input GCPs” and “Reference GCPs”. You can continue you work by loading the
files back into GCP Tool window.
How to do (Part 3)
Save the files

• Load Input GCPs… = To load the GCPs collected/saved during the previous session.

• Save Input As… = To save the input GCPs (the GCPs which are selected on the image to
be corrected.)

• Load Reference GCPs… = To load the reference GCPs collected/saved during the
previous session.

• Save Reference As… = To save the reference GCPs (the GCPs which are selected on the
reference image.)
How to do (Part 4)
Examine the Output Image
• Add the output image on a new 2D viewer.
• Examine the changes.
How to do (Part 4)
Verification of the Output
• One way to verify that the input image has been correctly rectified is to compare the
coordinate values of the same locations on output image with the reference image.
• Open both images in seperated 2D viewer.

Output image Reference image


How to do (Part 4)
Verification of the Output
• Go to Home – Link Views – Link Views to link both viewers.

• Then, go to Home – Link Views – Sync Views to synchronize the viewers whenever you
move around the images.

• Click on to view the images in equal scale.


How to do (Part 4)
Verification of the Output
• Use Inquire cursor to go to specific location (based on x, y coordinate). Examine the
locations.
Homework and
Questions
1) First use 4 GCPs and rectify the uncorrected image. Make your RMS error is less than
1/2 pixel. Compare coordinates of any 9 well defined and scattered points on the
image (between corrected and reference image) and note down the difference in
reading between those points.

2) Then, identify 15 GCPs and run the rectification. Once again compare the coordinates
readings between the rectified and reference image.

3) Note the dataset you are using for ground truth. What kind of errors could be
associated with your approach to GCP acquisition/collection?

4) If you are using higher spatial resolution image (as compared to the image you used in
this lab), would the georeferencing process be easier? Give your opinion.

PLEASE USE JENGKA DATA FROM UFUTURE


Uncorrected image- lanetm_2009_subset
Corrected image -landsat_2019_subset

You might also like