Corona KH-4B - Initial stereo got stuck at 2%--
43 views
Skip to first unread message
Suresh Das
Feb 14, 2024, 4:47:22 AMFeb 14
to Ames Stereo Pipeline Support
Hi,
I am interested in generating DEM form Corona images for mountainous region.
I have mosaic the images and generated camera files. Now when I try to run initial stereo, it got stuck at 2%. And after waiting for 5 hours, it sometime improves to 40% progress.
I am not using any map projected images. I am just trying to generate initial stereo with mosaic raw images.
Please suggest how to overcome this.
Regards
Suresh
Input info
***************************************************
# Input images and camera models (output from preprocessing script)
img1 = os.path.realpath("/home/das/Documents/corona_dem_asp/corona_post/for.tif")
img2 = os.path.realpath("/home/das/Documents/corona_dem_asp/corona_post/aft.tif")
cam1 = os.path.realpath("/home/das/Documents/corona_dem_asp/corona_post/for.tsai")
cam2 = os.path.realpath("/home/das/Documents/corona_dem_asp/corona_post/aft.tsai")
# Reference DEM
refdem = os.path.realpath("/home/das/Documents/corona_dem_asp/corona_post/ref_dem.tif")
outdir = os.path.realpath("/home/das/Documents/corona_dem_asp/corona_post/corona_post")
# EPSG code for output projection
epsg = "EPSG:32643"
# Coarse DEM resolution for intermediate steps
coarse_tr = 120
# Fine DEM grid resolution for final products
fine_tr = 30
img1 = os.path.realpath("/home/das/Documents/corona_dem_asp/corona_post/for.tif")
img2 = os.path.realpath("/home/das/Documents/corona_dem_asp/corona_post/aft.tif")
cam1 = os.path.realpath("/home/das/Documents/corona_dem_asp/corona_post/for.tsai")
cam2 = os.path.realpath("/home/das/Documents/corona_dem_asp/corona_post/aft.tsai")
# Reference DEM
refdem = os.path.realpath("/home/das/Documents/corona_dem_asp/corona_post/ref_dem.tif")
outdir = os.path.realpath("/home/das/Documents/corona_dem_asp/corona_post/corona_post")
# EPSG code for output projection
epsg = "EPSG:32643"
# Coarse DEM resolution for intermediate steps
coarse_tr = 120
# Fine DEM grid resolution for final products
fine_tr = 30
*********************************************************************
# Run stereo with initial camera models
init_stereo_dir = os.path.join(outdir, "init_stereo")
os.makedirs(init_stereo_dir, exist_ok=True)
stereo_cmd = f"stereo {img1} {img2} {cam1} {cam2} --alignment-method affineepipolar --unalign-disparity {os.path.join(init_stereo_dir, 'run')}"
os.system(stereo_cmd)
[ 2024-Feb-14 14:46:43 ] : Stage 0 --> PREPROCESSING
--> Setting number of processing threads to: 8
Stereo file ./stereo.default could not be found. Will use default settings and command line options only.
Writing log info to: /home/das/Documents/corona_dem_asp/corona_post/corona_post/init_stereo/run-log-stereo_pprc-02-14-1446-13941.txt
Using session: opticalbar
Loading camera model: /home/das/Documents/corona_dem_asp/corona_post/for.tif /home/das/Documents/corona_dem_asp/corona_post/for.tsai
Loading camera model: /home/das/Documents/corona_dem_asp/corona_post/aft.tif /home/das/Documents/corona_dem_asp/corona_post/aft.tsai
Distance between camera centers in meters: 6684.81.
Using image files: /home/das/Documents/corona_dem_asp/corona_post/for.tif, /home/das/Documents/corona_dem_asp/corona_post/aft.tif
Using camera files: /home/das/Documents/corona_dem_asp/corona_post/for.tsai, /home/das/Documents/corona_dem_asp/corona_post/aft.tsai
Computing statistics for left
Using downsample scale: 28
stats: [**************************************************] Complete!
Writing stats file: /home/das/Documents/corona_dem_asp/corona_post/corona_post/init_stereo/run-for-stats.tif
left: [ lo: 6 hi: 255 mean: 123.066 std_dev: 59.5897 ]
Left image stats time: 4.25426
Computing statistics for right
Using downsample scale: 28
stats: [**************************************************] Complete!
Writing stats file: /home/das/Documents/corona_dem_asp/corona_post/corona_post/init_stereo/run-aft-stats.tif
right: [ lo: 3 hi: 255 mean: 165.02 std_dev: 65.6873 ]
Right image stats time: 3.82478
--> Applying alignment method: affineepipolar
--> Matching interest points in StereoSession.
Using epipolar threshold = 7151.04
IP uniqueness threshold = 0.8
Datum: Geodetic Datum --> Name: WGS_1984 Spheroid: WGS 84 Semi-major axis: 6378137 Semi-minor axis: 6356752.3142451793 Meridian: Greenwich at 0 Proj4 Str: +proj=longlat +datum=WGS84 +no_defs
Using rough homography.
Rough homography--> [****************************************] Complete!
Projected 19507 rays for rough homography.
Rough homography inlier threshold: 10726.6
RANSAC iterations: 100
Estimating rough homography using RANSAC with 100 iterations.
Number of inliers: 19507.
RANSAC time: 10.5321 seconds.
Rough homography fit elapsed time: 11.8125 s.
Aligning right to left for IP capture using rough homography: Matrix3x3((0.967173,-0.0515867,0)(-0.00333539,0.980833,0)(-3.14486e-07,8.58098e-08,1))
Looking for IP in left image.
Using 50 interest points per tile (1024^2 px).
Detecting IP
Building descriptors
Found interest points: 37000
Recording interest points to file: /home/das/Documents/corona_dem_asp/corona_post/corona_post/init_stereo/run-for.vwip
Looking for IP in right image.
Using 50 interest points per tile (1024^2 px).
Detecting IP
Building descriptors
Found interest points: 38318
Elapsed time in ip detection: 117.154 s.
--> Matching interest points using the epipolar line.
Uniqueness threshold: 0.8
Epipolar threshold: 7151.04
Matching forward
---> Obtained 37000 matches.
Matching backward
---> Obtained 38318 matches.
Matched 620 points.
Unable to find inlier cluster, keeping best 70% of points.
Removed 23 points in stddev filtering.
Reduced matches to 411
Elapsed time in ip matching: 25.19 s.
Computing homography transform.
Homography rectification inlier threshold: 10726.6
RANSAC iterations: 100
Homography matrix:
Matrix3x3((1.11495,0.391284,-3813.5)(0.0313604,1.0823,-2645.09)(1.0867e-06,6.89247e-06,1))
Number of inliers: 391.
Elapsed time in homography computation: 0.173545 s.
* Writing match file: /home/das/Documents/corona_dem_asp/corona_post/corona_post/init_stereo/run-for__aft.match
Computing the epipolar rectification using RANSAC with 1000 iterations and inlier threshold 10.
RANSAC was unable to find a fit that matched the supplied data.
Attempting RANSAC with 136 inliers.
RANSAC was unable to find a fit that matched the supplied data.
Attempting RANSAC with 90 inliers.
RANSAC was unable to find a fit that matched the supplied data.
Attempting RANSAC with 60 inliers.
Found 99 / 411 inliers.
Maximum absolute difference of y components of aligned inlier interest points is 9.53316 pixels.
--> Aligning left and right images using affine matrices:
Matrix2x3((0.031938,-0.999737,5327)(0.999953,0.00969494,0))
Matrix2x3((0.0173951,-0.953017,6898.06)(0.997107,0.0414458,-529.051))
--> Normalizing globally to: [3.88657 296.394]
--> Writing pre-aligned images.
--> Writing: /home/das/Documents/corona_dem_asp/corona_post/corona_post/init_stereo/run-L.tif.
L: [******************************************************] Complete!
Time to write left image: 43.0865
--> Writing: /home/das/Documents/corona_dem_asp/corona_post/corona_post/init_stereo/run-R.tif.
R: [******************************************************] Complete!
Time to write right image: 44.232
--> Generating image masks.
Writing masks: /home/das/Documents/corona_dem_asp/corona_post/corona_post/init_stereo/run-lMask.tif /home/das/Documents/corona_dem_asp/corona_post/corona_post/init_stereo/run-rMask.tif.
Mask L: [***************************************************] Complete!
Mask R: [***************************************************] Complete!
--> Creating previews. Subsampling by 0.0491556 by using a tile of size 128 and 8 threads.
Sub L: [*************************************************] Complete!
Sub R: [*************************************************] Complete!
Sub L Mask: [********************************************] Complete!
Sub R Mask: [********************************************] Complete!
Convergence angle percentiles (in degrees) based on interest point matches:
25% 1.5065, 50% 1.63485, 75% 1.74922.
[ 2024-Feb-14 14:55:29 ] : PREPROCESSING FINISHED
--> Setting number of processing threads to: 8
Stereo file ./stereo.default could not be found. Will use default settings and command line options only.
Writing log info to: /home/das/Documents/corona_dem_asp/corona_post/corona_post/init_stereo/run-log-stereo_corr-02-14-1455-17538.txt
Using session: opticalbar
Loading camera model: /home/das/Documents/corona_dem_asp/corona_post/for.tif /home/das/Documents/corona_dem_asp/corona_post/for.tsai
Loading camera model: /home/das/Documents/corona_dem_asp/corona_post/aft.tif /home/das/Documents/corona_dem_asp/corona_post/aft.tsai
Distance between camera centers in meters: 6684.81.
--> Using LOG pre-processing filter with 1.5 sigma blur.
[ 2024-Feb-14 14:55:30 ] : Stage 1 --> CORRELATION
[ 2024-Feb-14 14:55:30 ] : Stage 1 --> LOW-RESOLUTION CORRELATION
Cached IP match file found: /home/das/Documents/corona_dem_asp/corona_post/corona_post/init_stereo/run-for__aft.match
--> Using interest points to determine search window.
* Loading match file: /home/das/Documents/corona_dem_asp/corona_post/corona_post/init_stereo/run-for__aft.match
Removed 2 outliers based on percentiles of differences of interest points with --outlier-removal-params.
Refined search range using a histogram and --outlier-removal-params: (Origin: (-1612, -4098) width: 4946 height: 12843)
Search width of 4946 is greater than desired limit of 4000, retrying with more aggressive IP filter.
Refined search range using a histogram and --outlier-removal-params: (Origin: (-644, -758) width: 1670 height: 2163)
D_sub search range: (Origin: (-42.375, -51.5) width: 103.75 height: 135) px
Writing: /home/das/Documents/corona_dem_asp/corona_post/corona_post/init_stereo/run-D_sub.tif
--> Low-resolution disparity:[*******************************] Complete!
Filtering outliers in D_sub based on --outlier-removal-params.
Inlier range based on x coordinate of disparity: -175 210.
Inlier range based on y coordinate of disparity: -186 185.
Number (and fraction) of removed outliers by disparity values in x and y: 0 (0).
Height above datum inlier range: -280971 294245.
Number (and fraction) of removed outliers by the height check: 0 (0).
Triangulation error inlier range: -21256.4 25137.2.
Number (and fraction) of removed outliers by the triangulation error check: 0 (0).
Writing filtered D_sub: /home/das/Documents/corona_dem_asp/corona_post/corona_post/init_stereo/run-D_sub.tif
D_sub: [****************************************************] Complete!
Writing triangulated point cloud based on D_sub: /home/das/Documents/corona_dem_asp/corona_post/corona_post/init_stereo/run-PC_sub.tif
PC_sub: [***************************************************] Complete!
--> Full-res search range based on D_sub: (Origin: (-814, -1038) width: 2035 height: 2300)
[ 2024-Feb-14 14:58:00 ] : LOW-RESOLUTION CORRELATION FINISHED
[ 2024-Feb-14 14:58:00 ] : CORRELATION FINISHED
--> Setting number of processing threads to: 8
Stereo file ./stereo.default could not be found. Will use default settings and command line options only.
Writing log info to: /home/das/Documents/corona_dem_asp/corona_post/corona_post/init_stereo/run-log-stereo_corr-02-14-1458-17725.txt
Using session: opticalbar
Loading camera model: /home/das/Documents/corona_dem_asp/corona_post/for.tif /home/das/Documents/corona_dem_asp/corona_post/for.tsai
Loading camera model: /home/das/Documents/corona_dem_asp/corona_post/aft.tif /home/das/Documents/corona_dem_asp/corona_post/aft.tsai
Distance between camera centers in meters: 6684.81.
--> Using LOG pre-processing filter with 1.5 sigma blur.
[ 2024-Feb-14 14:58:02 ] : Stage 1 --> CORRELATION
--------------------------------------------------
Kernel size: Vector2(21,21)
Search range: (Origin: (-814, -1038) width: 2035 height: 2300)
Cost mode: 2
--------------------------------------------------
Writing: /home/das/Documents/corona_dem_asp/corona_post/corona_post/init_stereo/run-D.tif
--> Correlation :[*...............................................] 2%
init_stereo_dir = os.path.join(outdir, "init_stereo")
os.makedirs(init_stereo_dir, exist_ok=True)
stereo_cmd = f"stereo {img1} {img2} {cam1} {cam2} --alignment-method affineepipolar --unalign-disparity {os.path.join(init_stereo_dir, 'run')}"
os.system(stereo_cmd)
[ 2024-Feb-14 14:46:43 ] : Stage 0 --> PREPROCESSING
--> Setting number of processing threads to: 8
Stereo file ./stereo.default could not be found. Will use default settings and command line options only.
Writing log info to: /home/das/Documents/corona_dem_asp/corona_post/corona_post/init_stereo/run-log-stereo_pprc-02-14-1446-13941.txt
Using session: opticalbar
Loading camera model: /home/das/Documents/corona_dem_asp/corona_post/for.tif /home/das/Documents/corona_dem_asp/corona_post/for.tsai
Loading camera model: /home/das/Documents/corona_dem_asp/corona_post/aft.tif /home/das/Documents/corona_dem_asp/corona_post/aft.tsai
Distance between camera centers in meters: 6684.81.
Using image files: /home/das/Documents/corona_dem_asp/corona_post/for.tif, /home/das/Documents/corona_dem_asp/corona_post/aft.tif
Using camera files: /home/das/Documents/corona_dem_asp/corona_post/for.tsai, /home/das/Documents/corona_dem_asp/corona_post/aft.tsai
Computing statistics for left
Using downsample scale: 28
stats: [**************************************************] Complete!
Writing stats file: /home/das/Documents/corona_dem_asp/corona_post/corona_post/init_stereo/run-for-stats.tif
left: [ lo: 6 hi: 255 mean: 123.066 std_dev: 59.5897 ]
Left image stats time: 4.25426
Computing statistics for right
Using downsample scale: 28
stats: [**************************************************] Complete!
Writing stats file: /home/das/Documents/corona_dem_asp/corona_post/corona_post/init_stereo/run-aft-stats.tif
right: [ lo: 3 hi: 255 mean: 165.02 std_dev: 65.6873 ]
Right image stats time: 3.82478
--> Applying alignment method: affineepipolar
--> Matching interest points in StereoSession.
Using epipolar threshold = 7151.04
IP uniqueness threshold = 0.8
Datum: Geodetic Datum --> Name: WGS_1984 Spheroid: WGS 84 Semi-major axis: 6378137 Semi-minor axis: 6356752.3142451793 Meridian: Greenwich at 0 Proj4 Str: +proj=longlat +datum=WGS84 +no_defs
Using rough homography.
Rough homography--> [****************************************] Complete!
Projected 19507 rays for rough homography.
Rough homography inlier threshold: 10726.6
RANSAC iterations: 100
Estimating rough homography using RANSAC with 100 iterations.
Number of inliers: 19507.
RANSAC time: 10.5321 seconds.
Rough homography fit elapsed time: 11.8125 s.
Aligning right to left for IP capture using rough homography: Matrix3x3((0.967173,-0.0515867,0)(-0.00333539,0.980833,0)(-3.14486e-07,8.58098e-08,1))
Looking for IP in left image.
Using 50 interest points per tile (1024^2 px).
Detecting IP
Building descriptors
Found interest points: 37000
Recording interest points to file: /home/das/Documents/corona_dem_asp/corona_post/corona_post/init_stereo/run-for.vwip
Looking for IP in right image.
Using 50 interest points per tile (1024^2 px).
Detecting IP
Building descriptors
Found interest points: 38318
Elapsed time in ip detection: 117.154 s.
--> Matching interest points using the epipolar line.
Uniqueness threshold: 0.8
Epipolar threshold: 7151.04
Matching forward
---> Obtained 37000 matches.
Matching backward
---> Obtained 38318 matches.
Matched 620 points.
Unable to find inlier cluster, keeping best 70% of points.
Removed 23 points in stddev filtering.
Reduced matches to 411
Elapsed time in ip matching: 25.19 s.
Computing homography transform.
Homography rectification inlier threshold: 10726.6
RANSAC iterations: 100
Homography matrix:
Matrix3x3((1.11495,0.391284,-3813.5)(0.0313604,1.0823,-2645.09)(1.0867e-06,6.89247e-06,1))
Number of inliers: 391.
Elapsed time in homography computation: 0.173545 s.
* Writing match file: /home/das/Documents/corona_dem_asp/corona_post/corona_post/init_stereo/run-for__aft.match
Computing the epipolar rectification using RANSAC with 1000 iterations and inlier threshold 10.
RANSAC was unable to find a fit that matched the supplied data.
Attempting RANSAC with 136 inliers.
RANSAC was unable to find a fit that matched the supplied data.
Attempting RANSAC with 90 inliers.
RANSAC was unable to find a fit that matched the supplied data.
Attempting RANSAC with 60 inliers.
Found 99 / 411 inliers.
Maximum absolute difference of y components of aligned inlier interest points is 9.53316 pixels.
--> Aligning left and right images using affine matrices:
Matrix2x3((0.031938,-0.999737,5327)(0.999953,0.00969494,0))
Matrix2x3((0.0173951,-0.953017,6898.06)(0.997107,0.0414458,-529.051))
--> Normalizing globally to: [3.88657 296.394]
--> Writing pre-aligned images.
--> Writing: /home/das/Documents/corona_dem_asp/corona_post/corona_post/init_stereo/run-L.tif.
L: [******************************************************] Complete!
Time to write left image: 43.0865
--> Writing: /home/das/Documents/corona_dem_asp/corona_post/corona_post/init_stereo/run-R.tif.
R: [******************************************************] Complete!
Time to write right image: 44.232
--> Generating image masks.
Writing masks: /home/das/Documents/corona_dem_asp/corona_post/corona_post/init_stereo/run-lMask.tif /home/das/Documents/corona_dem_asp/corona_post/corona_post/init_stereo/run-rMask.tif.
Mask L: [***************************************************] Complete!
Mask R: [***************************************************] Complete!
--> Creating previews. Subsampling by 0.0491556 by using a tile of size 128 and 8 threads.
Sub L: [*************************************************] Complete!
Sub R: [*************************************************] Complete!
Sub L Mask: [********************************************] Complete!
Sub R Mask: [********************************************] Complete!
Convergence angle percentiles (in degrees) based on interest point matches:
25% 1.5065, 50% 1.63485, 75% 1.74922.
[ 2024-Feb-14 14:55:29 ] : PREPROCESSING FINISHED
--> Setting number of processing threads to: 8
Stereo file ./stereo.default could not be found. Will use default settings and command line options only.
Writing log info to: /home/das/Documents/corona_dem_asp/corona_post/corona_post/init_stereo/run-log-stereo_corr-02-14-1455-17538.txt
Using session: opticalbar
Loading camera model: /home/das/Documents/corona_dem_asp/corona_post/for.tif /home/das/Documents/corona_dem_asp/corona_post/for.tsai
Loading camera model: /home/das/Documents/corona_dem_asp/corona_post/aft.tif /home/das/Documents/corona_dem_asp/corona_post/aft.tsai
Distance between camera centers in meters: 6684.81.
--> Using LOG pre-processing filter with 1.5 sigma blur.
[ 2024-Feb-14 14:55:30 ] : Stage 1 --> CORRELATION
[ 2024-Feb-14 14:55:30 ] : Stage 1 --> LOW-RESOLUTION CORRELATION
Cached IP match file found: /home/das/Documents/corona_dem_asp/corona_post/corona_post/init_stereo/run-for__aft.match
--> Using interest points to determine search window.
* Loading match file: /home/das/Documents/corona_dem_asp/corona_post/corona_post/init_stereo/run-for__aft.match
Removed 2 outliers based on percentiles of differences of interest points with --outlier-removal-params.
Refined search range using a histogram and --outlier-removal-params: (Origin: (-1612, -4098) width: 4946 height: 12843)
Search width of 4946 is greater than desired limit of 4000, retrying with more aggressive IP filter.
Refined search range using a histogram and --outlier-removal-params: (Origin: (-644, -758) width: 1670 height: 2163)
D_sub search range: (Origin: (-42.375, -51.5) width: 103.75 height: 135) px
Writing: /home/das/Documents/corona_dem_asp/corona_post/corona_post/init_stereo/run-D_sub.tif
--> Low-resolution disparity:[*******************************] Complete!
Filtering outliers in D_sub based on --outlier-removal-params.
Inlier range based on x coordinate of disparity: -175 210.
Inlier range based on y coordinate of disparity: -186 185.
Number (and fraction) of removed outliers by disparity values in x and y: 0 (0).
Height above datum inlier range: -280971 294245.
Number (and fraction) of removed outliers by the height check: 0 (0).
Triangulation error inlier range: -21256.4 25137.2.
Number (and fraction) of removed outliers by the triangulation error check: 0 (0).
Writing filtered D_sub: /home/das/Documents/corona_dem_asp/corona_post/corona_post/init_stereo/run-D_sub.tif
D_sub: [****************************************************] Complete!
Writing triangulated point cloud based on D_sub: /home/das/Documents/corona_dem_asp/corona_post/corona_post/init_stereo/run-PC_sub.tif
PC_sub: [***************************************************] Complete!
--> Full-res search range based on D_sub: (Origin: (-814, -1038) width: 2035 height: 2300)
[ 2024-Feb-14 14:58:00 ] : LOW-RESOLUTION CORRELATION FINISHED
[ 2024-Feb-14 14:58:00 ] : CORRELATION FINISHED
--> Setting number of processing threads to: 8
Stereo file ./stereo.default could not be found. Will use default settings and command line options only.
Writing log info to: /home/das/Documents/corona_dem_asp/corona_post/corona_post/init_stereo/run-log-stereo_corr-02-14-1458-17725.txt
Using session: opticalbar
Loading camera model: /home/das/Documents/corona_dem_asp/corona_post/for.tif /home/das/Documents/corona_dem_asp/corona_post/for.tsai
Loading camera model: /home/das/Documents/corona_dem_asp/corona_post/aft.tif /home/das/Documents/corona_dem_asp/corona_post/aft.tsai
Distance between camera centers in meters: 6684.81.
--> Using LOG pre-processing filter with 1.5 sigma blur.
[ 2024-Feb-14 14:58:02 ] : Stage 1 --> CORRELATION
--------------------------------------------------
Kernel size: Vector2(21,21)
Search range: (Origin: (-814, -1038) width: 2035 height: 2300)
Cost mode: 2
--------------------------------------------------
Writing: /home/das/Documents/corona_dem_asp/corona_post/corona_post/init_stereo/run-D.tif
--> Correlation :[*...............................................] 2%
Oleg Alexandrov
Feb 14, 2024, 10:37:05 AMFeb 14
to Suresh Das, Ames Stereo Pipeline Support
The problem is here:
Search range: (Origin: (-814, -1038) width: 2035 height: 2300)
This search range is just outrageous. Anything wider than 100 pixels will take forever.
It is suggested to mapproect the images and overlay them ahead of time to see how well they agree. If they don't agree much in places, that may also result in a big search range.
It is very tricky to process such images with ASP as it is hard to get the camera model right.
--
You received this message because you are subscribed to the Google Groups "Ames Stereo Pipeline Support" group.
To unsubscribe from this group and stop receiving emails from it, send an email to ames-stereo-pipeline...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/ames-stereo-pipeline-support/2bf8f7b5-f7de-4dee-9086-85ecec8bdd87n%40googlegroups.com.
Reply all
Reply to author
Forward
0 new messages