Geometric Correction-lab Report
This document was uploaded by user and they confirmed that they have the permission to share it. If you are author or own the copyright of this book, please report to us by using this DMCA report form. Report DMCA
Overview
Download & View Geometric Correction-lab Report as PDF for free.
More details
- Words: 1,877
- Pages: 15
SGS 1633 REMOTE SENSING TECHNOLOGY 1
LAB REPORT 1
LECTURER : PROF MOHD IBRAHIM SEENI MOHD
PREPARED BY: MOHD FARID BIN FAUZI 900707-03-5463 AG090090 1-SGS 2009/10
August 5, [INTRODUCTION TO IMAGE PROCESSING] 2009
10 20
TITLE Geometric Correction OBJECTIVES 2.1 To correct the satellite data so that it is true to scale and projection. 2.2 To enhance the data in the correct position based on the real place
or map magnetic direction. 2.3 Analysing all image resampling data using different method. 10
METHODOLOGY 3.1 First, Click PCI Geomatica V9.1 program and select GCP Works.
3.2We have to select 4 different setups for GCP Works. i. In Processing Requirements column, select Full Processing. ii. In Mathematical Model column, select Polynomial. iii. In Source of GCP column, select User Entered Coordinates. iv. Click on accept button at last and the PCI GCP Works V9.1.0 window will be shown. 3.2Then in PCI GCPWorks V9.1.0 window, i. Click on Select Uncorrected Image.
Page
2 Figure 1: The uncorrected satellite image ii. Choose the uncorrected image which we need to correct. iii. In the Database Channels, choose the best combination of channels which enable us to get a clear image. iv. After select the channels, click Load & Close. Page |2
August 5, [INTRODUCTION TO IMAGE PROCESSING] 2009 3.2 Next, click on Define Goereferencing Units. i. Change Meter to Long/Lat. ii. Click on Earth Model button. iii. Select D085 - Kertau 1948 [West Malaysia & Singapore] for Datum. iv. Then select E012 – WGS 84 for ellipsoid. v. After select the datum and ellipsoid, click on accept button. 3.2 After that, click on Collect GCPs. i. Select a coordinate on the image and key in the coordinate on the GCP menu. ii. Key in the Longitude and Latitude on that point and click on Accept as GCP button to accept it as control point. iii. Repeat this 2 process for others coordinates until the last coordinate. iv. Save and close the GCP after finished key in all coordinates.
3 4
LATITUD
3
2
LONGITU D 103˚36’29. 86” 103˚40’13. 33” 103˚39’09. 58” 103˚48’05. 17”
Page
GCP s 1
1˚39’40.8 1’’ 1˚38’14.6 0’’ 1˚36’19.2 0’’ 1˚36’25.6 3’’
Page |3
August 5, [INTRODUCTION TO IMAGE PROCESSING] 2009 5 6 7 8 9 10 11 12 13 14 15 16 17
103˚48’53. 37” 103˚54’46. 03” 103˚37’25. 11” 103˚39’33. 76” 103˚41’05. 57” 103˚43’05. 50” 103˚46’11. 30” 103˚49’14. 46” 103˚52’04. 82” 103˚35’29. 15” 103˚39’14. 35” 103˚46’05. 28” 103˚46’46. 60”
1˚34’13.0 5’’ 1˚33’16.1 6’’ 1˚31’17.0 9’’ 1˚32’53.7 1’’ 1˚30’00.1 9’’ 1˚29’29.8 6’’ 1˚31’27.9 8’’ 1˚31’15.5 9’’ 1˚28’02.7 3’’ 1˚26’51.1 3’’ 1˚25’41.8 2 1˚27’26.2 7’’ 1˚27’58.3 6’’
4
Page |4
Page
3.2 Then, click on Perform Registration to Disk to start the re-sampling of the image. i. Click on New Output File and type a name for the file. ii. In New File Type Selector window, select PCIDSK(.pix) as file format. iii. Under Number of Channels column, change to 7 Channels with 8-bit signed. iv. Select Band Interleaved as Format Options. v. Under Geo-Referencing Information, change Use pixel/lines and bounds to Used pixel/lines and resolution. vi. Change the pixel size to 0d00’01.000” for X and Y column. vii.Click on Create. viii.Select Default and drag the Memory (MB) slide to 32.0. ix. Click on Perform Registration button. x. Close all the related windows.
August 5, [INTRODUCTION TO IMAGE PROCESSING] 2009
3.2 Lastly, on Geomatica Toolbar V9.1.0, select Image Works Configuration. i. Click Use Image File, select the new image file. ii. Under File Information, change the Y dimension to maximum scale . iii. Change also Image Planes to 7. iv. Click Accept & Load to view the corrected image.
Page
5 Page |5
August 5, [INTRODUCTION TO IMAGE PROCESSING] 2009
10
RESULT
Page
Page |6
6
The above figure shown all Ground Control Points(GCPs) and the following figure below shows the coefficients which calculated using 1st Degree Polynomial.
August 5, [INTRODUCTION TO IMAGE PROCESSING] 2009
This is the coefficient that being calculated using 2nd Degree Polynomial. Forward x‘= -373990.1 +3.283271y2 y’ = 40196.67 +2.07147y2
+3528.7x
+783.9292y
-2.107199xy
-168.303x
+3490.923y
+0.4506205xy
+0.8769021x2 -1.572458x2
Backward: x = 103.5835 +0.0002637396x -4.143648e-005y +3.163749e-011xy -1.616482e-011x2 -6.710729e-011y2 y = -1.667018 +3.696235e-005x +0.0002774233y -2.649279e-011xy +2.662779e-011x2 -5.227675e-011y2
This is the coefficient that being calculated using 3rd Degree Polynomial.
Page |7
7
y’ = -1.667008 +3.697817e-005x +0.0002771613y +6.523898e-010xy -1.108128e-010x2 +2.803846e-010y -1.611679e-013x2 y -6.387019e-013xy2 +7.149368e-014x3 -5.703968e3 014y
Page
Forward: x‘= -229743.1 +1485.46x +147592.2y -2842.357xy +0x2 2 2 -404.3998y +13.6907x y +0.7899071xy2 +0.06887933x3 -70.67193y3 y’ = 36915.28 -320.471x -26997.54y +279.8739xy +0x2 2 2 -10428.3y +0.1489505x y +101.1137xy2 +0.0019725x 3 +11.24569y3 Backward: x’ = 103.5835 +0.0002636583x -4.119007e-005y -1.000802e-010xy +7.989206e-011x2 -7.064835e-010y2 -1.413018e-013x2 y +4.411238e-013xy2 -6.844629e-015x3 +3.558562e3 013y
August 5, [INTRODUCTION TO IMAGE PROCESSING] 2009 After all the points have been entered, the image then corrected by 2nd degree polynomial and nearest interpolation. The figure below is the image overview after Geometrical Correction.
Page
8 Page |8
August 5, [INTRODUCTION TO IMAGE PROCESSING] 2009
Page
This is the overview of the corrected image with bilinear interpolation and 2nd degree polynomial after Geometric Correction.
9 Page |9
August 5, [INTRODUCTION TO IMAGE PROCESSING] 2009
The figure below is the image overview after Geometrical Correction by using 2nd degree polynomial and nearest interpolation.
Page
10 P a g e | 10
August 5, [INTRODUCTION TO IMAGE PROCESSING] 2009
20
ANALYSING DATA For this practical, the satellite image data from Landsat TM have been used. This data being used as to done Geometric Correction. This raw image is a distorted image. All images taken from satellite contain distortion and error. This distortion was cause by the perspective of the sensor optics; the motion of the scanning system; the motion of the platform; the platform altitude, attitude and velocity change; the terrain relief and the curvature and rotation of the earth. There many type of distortion such as radial distortion, tangential distortion, stepwise distortion, scale error, projection distortion, skew, along track scale error and scan line scale error. Geometric correction is done in order to achieve as close as possible between geometry representation of image and the real place on the earth. The first aspect that we should consider while doing this practical is to achieve a low Root Mean Square (RMS) error. Refer to my report, I had achieve a RMS error of 2.90 and 2.77 for the coordinates X and Y respectively in First Degree Polynomial. While in the Second Degree Polynomial, the RMS error of coordinates X and Y are 0.71 and 0.92. Theoretically, the RMS error for Second Degree Polynomial should be less than First Degree Polynomial due to Second Degree Polynomial is used to achieve more accurate result. In professional use of the image, the particular image should contain RMS error less than 2.00 to give a precise analysis.
P a g e | 11
11
The point that measured also must be exactly the same with the point on the ground. Make sure that the point is determined correctly. Do not ever put the point in different pixel if want to get better result. As prevention, zoom the image as large as there will not be a mistake done.
Page
As to get better result, we must consider the points keyed-in the software. These GCPs should key in with well define and well distributed points in the image. The meaning of well define point is the location which will not be effect by any changes over a long time. The example of well define point is the road junction. The GCPs have to well distribute too on the image. Don’t put the all GCPs in nearest place or focus in certain place only. We must spread all the points to get the actual result.
August 5, [INTRODUCTION TO IMAGE PROCESSING] 2009
Geometric correction was done by using polynomial function. This polynomial function is important to gain the corrected image as well as on the ground. We can use either 1 st, 2nd, 3rd, 4th degree or so on but mostly people use 2nd and 3rd degree polynomials. •
First Degree Polynomial X1 = a0 + a1x + a2y Y1 = b0 + b1X + b2y
•
Second Degree Polynomial X1 = a0 + a1x + a2y + a3xy + a4x2 + a5y2 Y1 = b0 +b1x + b2y + b3xy + b4x2 + b5y2
•
Third Degree Polynomial X1 = a0 + a1x + a2y + a3xy + a4x2 + a5y2 + a6x2y + a7xy2 + a8x3 + a9y3 Y1 = b0 + b1x + b2y + b3xy + b4x2 + b5y2 + b6x2y + b7xy2 + b8x3 + b9y3
wher e:
X1,and Y1 are coordinate points from the corrected image, x and y are coordinates from the original image. n≥3 n≥6
Third Degree Polynomial
n ≥ 10
Page
First Degree Polynomial Second Degree Polynomial
1st degree polynomial is the simplest geometric correction. As we using higher degree of polynomials such as 2nd, 3rd, 4th, 5th and so on, we can get almost accurate result. But people always use whether 2nd 0r 3rd degree because it is most accurate than first degree and does not need more GCPs. P a g e | 12
12
3, 6 and 10 is the minimum GCPs that you should have if using 1st, 2nd and 3rd degree polynomial but do not use exactly those number of GCPs because you may also get result (with no residual) but you cannot ensure whether your points are right or wrong.
August 5, [INTRODUCTION TO IMAGE PROCESSING] 2009
The last part of geometric correction is resampling. This process is function as moving the pixel of the original image to the corrected image based on the algorithms above. This process will perform an image after the transformation. There are 3 types of resampling method named Nearest Neighbour, Bilinear Interpolation, and Cubic Convolution. Technique Transfer of output pixel to the input pixel which nearest with it.
Advantages - Minimum lost of the extremes and subtleties of the image. - Fast computation.
Disadvantages - Resampling to smaller grid size. - “Stair stepped” effect occur around diagonal lines and curves. - Edges are smoothed and some extremes of the data file are lost.
Bilinear Interpolation
First-order interpolation between four adjacent input pixel.
Cubic Convolution
Cubic function based on 16 input pixel.
- Spatially more accurate than nearest neighbour. - Result are smoother, without “stair stepped” effect that occur in nearest neighbour. -Most accurate - Slow computation. resampling method. - Most expensive - Effecting the image by method. sharpen the image and smooth out noise.
Page
Method Nearest Neighbour
13 P a g e | 13
August 5, [INTRODUCTION TO IMAGE PROCESSING] 2009
This is the effect if we put GCPs wrongly. This is resampling image overview by bad distribution of GCPs.
Page
14 P a g e | 14
August 5, [INTRODUCTION TO IMAGE PROCESSING] 2009
10
CONCLUSION 1.1 Every image taken from satellite sensor has geometric error. 1.2 Geometric correction done to make the image true to scale and true in direction which exactly the same with the ground. 1.3 We must put the GCPs point carefully to reduced the residual. 1.4 We must have more than minimum GCPs for each degree polynomial to prevent undefined residual or error. 1.5 Resampling image must not folded.
Page
15 P a g e | 15
LAB REPORT 1
LECTURER : PROF MOHD IBRAHIM SEENI MOHD
PREPARED BY: MOHD FARID BIN FAUZI 900707-03-5463 AG090090 1-SGS 2009/10
August 5, [INTRODUCTION TO IMAGE PROCESSING] 2009
10 20
TITLE Geometric Correction OBJECTIVES 2.1 To correct the satellite data so that it is true to scale and projection. 2.2 To enhance the data in the correct position based on the real place
or map magnetic direction. 2.3 Analysing all image resampling data using different method. 10
METHODOLOGY 3.1 First, Click PCI Geomatica V9.1 program and select GCP Works.
3.2We have to select 4 different setups for GCP Works. i. In Processing Requirements column, select Full Processing. ii. In Mathematical Model column, select Polynomial. iii. In Source of GCP column, select User Entered Coordinates. iv. Click on accept button at last and the PCI GCP Works V9.1.0 window will be shown. 3.2Then in PCI GCPWorks V9.1.0 window, i. Click on Select Uncorrected Image.
Page
2 Figure 1: The uncorrected satellite image ii. Choose the uncorrected image which we need to correct. iii. In the Database Channels, choose the best combination of channels which enable us to get a clear image. iv. After select the channels, click Load & Close. Page |2
August 5, [INTRODUCTION TO IMAGE PROCESSING] 2009 3.2 Next, click on Define Goereferencing Units. i. Change Meter to Long/Lat. ii. Click on Earth Model button. iii. Select D085 - Kertau 1948 [West Malaysia & Singapore] for Datum. iv. Then select E012 – WGS 84 for ellipsoid. v. After select the datum and ellipsoid, click on accept button. 3.2 After that, click on Collect GCPs. i. Select a coordinate on the image and key in the coordinate on the GCP menu. ii. Key in the Longitude and Latitude on that point and click on Accept as GCP button to accept it as control point. iii. Repeat this 2 process for others coordinates until the last coordinate. iv. Save and close the GCP after finished key in all coordinates.
3 4
LATITUD
3
2
LONGITU D 103˚36’29. 86” 103˚40’13. 33” 103˚39’09. 58” 103˚48’05. 17”
Page
GCP s 1
1˚39’40.8 1’’ 1˚38’14.6 0’’ 1˚36’19.2 0’’ 1˚36’25.6 3’’
Page |3
August 5, [INTRODUCTION TO IMAGE PROCESSING] 2009 5 6 7 8 9 10 11 12 13 14 15 16 17
103˚48’53. 37” 103˚54’46. 03” 103˚37’25. 11” 103˚39’33. 76” 103˚41’05. 57” 103˚43’05. 50” 103˚46’11. 30” 103˚49’14. 46” 103˚52’04. 82” 103˚35’29. 15” 103˚39’14. 35” 103˚46’05. 28” 103˚46’46. 60”
1˚34’13.0 5’’ 1˚33’16.1 6’’ 1˚31’17.0 9’’ 1˚32’53.7 1’’ 1˚30’00.1 9’’ 1˚29’29.8 6’’ 1˚31’27.9 8’’ 1˚31’15.5 9’’ 1˚28’02.7 3’’ 1˚26’51.1 3’’ 1˚25’41.8 2 1˚27’26.2 7’’ 1˚27’58.3 6’’
4
Page |4
Page
3.2 Then, click on Perform Registration to Disk to start the re-sampling of the image. i. Click on New Output File and type a name for the file. ii. In New File Type Selector window, select PCIDSK(.pix) as file format. iii. Under Number of Channels column, change to 7 Channels with 8-bit signed. iv. Select Band Interleaved as Format Options. v. Under Geo-Referencing Information, change Use pixel/lines and bounds to Used pixel/lines and resolution. vi. Change the pixel size to 0d00’01.000” for X and Y column. vii.Click on Create. viii.Select Default and drag the Memory (MB) slide to 32.0. ix. Click on Perform Registration button. x. Close all the related windows.
August 5, [INTRODUCTION TO IMAGE PROCESSING] 2009
3.2 Lastly, on Geomatica Toolbar V9.1.0, select Image Works Configuration. i. Click Use Image File, select the new image file. ii. Under File Information, change the Y dimension to maximum scale . iii. Change also Image Planes to 7. iv. Click Accept & Load to view the corrected image.
Page
5 Page |5
August 5, [INTRODUCTION TO IMAGE PROCESSING] 2009
10
RESULT
Page
Page |6
6
The above figure shown all Ground Control Points(GCPs) and the following figure below shows the coefficients which calculated using 1st Degree Polynomial.
August 5, [INTRODUCTION TO IMAGE PROCESSING] 2009
This is the coefficient that being calculated using 2nd Degree Polynomial. Forward x‘= -373990.1 +3.283271y2 y’ = 40196.67 +2.07147y2
+3528.7x
+783.9292y
-2.107199xy
-168.303x
+3490.923y
+0.4506205xy
+0.8769021x2 -1.572458x2
Backward: x = 103.5835 +0.0002637396x -4.143648e-005y +3.163749e-011xy -1.616482e-011x2 -6.710729e-011y2 y = -1.667018 +3.696235e-005x +0.0002774233y -2.649279e-011xy +2.662779e-011x2 -5.227675e-011y2
This is the coefficient that being calculated using 3rd Degree Polynomial.
Page |7
7
y’ = -1.667008 +3.697817e-005x +0.0002771613y +6.523898e-010xy -1.108128e-010x2 +2.803846e-010y -1.611679e-013x2 y -6.387019e-013xy2 +7.149368e-014x3 -5.703968e3 014y
Page
Forward: x‘= -229743.1 +1485.46x +147592.2y -2842.357xy +0x2 2 2 -404.3998y +13.6907x y +0.7899071xy2 +0.06887933x3 -70.67193y3 y’ = 36915.28 -320.471x -26997.54y +279.8739xy +0x2 2 2 -10428.3y +0.1489505x y +101.1137xy2 +0.0019725x 3 +11.24569y3 Backward: x’ = 103.5835 +0.0002636583x -4.119007e-005y -1.000802e-010xy +7.989206e-011x2 -7.064835e-010y2 -1.413018e-013x2 y +4.411238e-013xy2 -6.844629e-015x3 +3.558562e3 013y
August 5, [INTRODUCTION TO IMAGE PROCESSING] 2009 After all the points have been entered, the image then corrected by 2nd degree polynomial and nearest interpolation. The figure below is the image overview after Geometrical Correction.
Page
8 Page |8
August 5, [INTRODUCTION TO IMAGE PROCESSING] 2009
Page
This is the overview of the corrected image with bilinear interpolation and 2nd degree polynomial after Geometric Correction.
9 Page |9
August 5, [INTRODUCTION TO IMAGE PROCESSING] 2009
The figure below is the image overview after Geometrical Correction by using 2nd degree polynomial and nearest interpolation.
Page
10 P a g e | 10
August 5, [INTRODUCTION TO IMAGE PROCESSING] 2009
20
ANALYSING DATA For this practical, the satellite image data from Landsat TM have been used. This data being used as to done Geometric Correction. This raw image is a distorted image. All images taken from satellite contain distortion and error. This distortion was cause by the perspective of the sensor optics; the motion of the scanning system; the motion of the platform; the platform altitude, attitude and velocity change; the terrain relief and the curvature and rotation of the earth. There many type of distortion such as radial distortion, tangential distortion, stepwise distortion, scale error, projection distortion, skew, along track scale error and scan line scale error. Geometric correction is done in order to achieve as close as possible between geometry representation of image and the real place on the earth. The first aspect that we should consider while doing this practical is to achieve a low Root Mean Square (RMS) error. Refer to my report, I had achieve a RMS error of 2.90 and 2.77 for the coordinates X and Y respectively in First Degree Polynomial. While in the Second Degree Polynomial, the RMS error of coordinates X and Y are 0.71 and 0.92. Theoretically, the RMS error for Second Degree Polynomial should be less than First Degree Polynomial due to Second Degree Polynomial is used to achieve more accurate result. In professional use of the image, the particular image should contain RMS error less than 2.00 to give a precise analysis.
P a g e | 11
11
The point that measured also must be exactly the same with the point on the ground. Make sure that the point is determined correctly. Do not ever put the point in different pixel if want to get better result. As prevention, zoom the image as large as there will not be a mistake done.
Page
As to get better result, we must consider the points keyed-in the software. These GCPs should key in with well define and well distributed points in the image. The meaning of well define point is the location which will not be effect by any changes over a long time. The example of well define point is the road junction. The GCPs have to well distribute too on the image. Don’t put the all GCPs in nearest place or focus in certain place only. We must spread all the points to get the actual result.
August 5, [INTRODUCTION TO IMAGE PROCESSING] 2009
Geometric correction was done by using polynomial function. This polynomial function is important to gain the corrected image as well as on the ground. We can use either 1 st, 2nd, 3rd, 4th degree or so on but mostly people use 2nd and 3rd degree polynomials. •
First Degree Polynomial X1 = a0 + a1x + a2y Y1 = b0 + b1X + b2y
•
Second Degree Polynomial X1 = a0 + a1x + a2y + a3xy + a4x2 + a5y2 Y1 = b0 +b1x + b2y + b3xy + b4x2 + b5y2
•
Third Degree Polynomial X1 = a0 + a1x + a2y + a3xy + a4x2 + a5y2 + a6x2y + a7xy2 + a8x3 + a9y3 Y1 = b0 + b1x + b2y + b3xy + b4x2 + b5y2 + b6x2y + b7xy2 + b8x3 + b9y3
wher e:
X1,and Y1 are coordinate points from the corrected image, x and y are coordinates from the original image. n≥3 n≥6
Third Degree Polynomial
n ≥ 10
Page
First Degree Polynomial Second Degree Polynomial
1st degree polynomial is the simplest geometric correction. As we using higher degree of polynomials such as 2nd, 3rd, 4th, 5th and so on, we can get almost accurate result. But people always use whether 2nd 0r 3rd degree because it is most accurate than first degree and does not need more GCPs. P a g e | 12
12
3, 6 and 10 is the minimum GCPs that you should have if using 1st, 2nd and 3rd degree polynomial but do not use exactly those number of GCPs because you may also get result (with no residual) but you cannot ensure whether your points are right or wrong.
August 5, [INTRODUCTION TO IMAGE PROCESSING] 2009
The last part of geometric correction is resampling. This process is function as moving the pixel of the original image to the corrected image based on the algorithms above. This process will perform an image after the transformation. There are 3 types of resampling method named Nearest Neighbour, Bilinear Interpolation, and Cubic Convolution. Technique Transfer of output pixel to the input pixel which nearest with it.
Advantages - Minimum lost of the extremes and subtleties of the image. - Fast computation.
Disadvantages - Resampling to smaller grid size. - “Stair stepped” effect occur around diagonal lines and curves. - Edges are smoothed and some extremes of the data file are lost.
Bilinear Interpolation
First-order interpolation between four adjacent input pixel.
Cubic Convolution
Cubic function based on 16 input pixel.
- Spatially more accurate than nearest neighbour. - Result are smoother, without “stair stepped” effect that occur in nearest neighbour. -Most accurate - Slow computation. resampling method. - Most expensive - Effecting the image by method. sharpen the image and smooth out noise.
Page
Method Nearest Neighbour
13 P a g e | 13
August 5, [INTRODUCTION TO IMAGE PROCESSING] 2009
This is the effect if we put GCPs wrongly. This is resampling image overview by bad distribution of GCPs.
Page
14 P a g e | 14
August 5, [INTRODUCTION TO IMAGE PROCESSING] 2009
10
CONCLUSION 1.1 Every image taken from satellite sensor has geometric error. 1.2 Geometric correction done to make the image true to scale and true in direction which exactly the same with the ground. 1.3 We must put the GCPs point carefully to reduced the residual. 1.4 We must have more than minimum GCPs for each degree polynomial to prevent undefined residual or error. 1.5 Resampling image must not folded.
Page
15 P a g e | 15
Related Documents
Heavy Vehicle Report Geometric
November 2019 22
Geometric Correction-lab Report
June 2020 15
Geometric
June 2020 25
Geometric
June 2020 19
Geometric
November 2019 33
Geometric Addition + Logarithms=geometric Multiplication
November 2019 34More Documents from "Ahsan Shahid"
Perlembagaan Malaysia Dlm Konteks Hubungan Etnik
June 2020 14
Assignment Remote Sensing
June 2020 12
Introduction To Image Processing-lab Report
June 2020 10
Geometric Correction-lab Report
June 2020 15