Improvement of Absolute Accuracy by processing images together with gcp_list.txt file


#1

Thanks to @pierotofy, processing 100 images together with gcp_list.txt file on WebODM updated was successfully complete.

This is orthophoto tif file generated by processing on WebODM:
https://drive.google.com/file/d/1AoMmEDpJGxHzM_Ain92EGKbRfY7llNf_/view?usp=sharing

I checked whether absolute accuracy produced by processing images together with gcp file having 5 GCP points has been improved or not, by comparing coordinates of GCP points on the “Orthophoto” on QGIS program with “Actual” coordinates measured by GPS station. These are comparison of 2 GCP points among 5 points:

Coordinate System: WGS84 UTM54N
gcp_01…Actual…4154527.79…312559.16…6.89
…Orhophoto…4154524.93…312558.77…?

gcp_02…Actual…4154602.92…312471.89…7.85
…Orhophoto…4154601.12…312471.00…?

As mentioned above, processing was successfully complete. I wonder whether the coordinates of GCPs on gcp_list.txt improved the absolute accuracy of orthophoto or not by processing on WebODM.

This is gcp_list.txt file processed together with 100 images on WebODM:
https://drive.google.com/file/d/1cW8zZM8cRD0bkyqjS69CvAUimSJ5IJXz/view?usp=sharing

For reference, the absolute accuracy of survey grade shall be less than 10 centimeters on X and Y axis here in UTM 54N.

Any suggestions and recommendations on this issue will be highly appreciated.


#2

I processed 100 images without gcp_list.txt file on WebODM and compared the coordinates of orthophoto generated with those of orthophoto produced from processing images with gcp_list file. The coordinates were same in both cases, which means gcp_list file didn’t improve absolute accuracy at all.

What is wrong with this? How can absolute accuracy be improved to survey grade?


#3

I am also unsure about the GCPs. In my case, they did improve the precision, but the GCPs on the ground do not match the ortho created.

As I mentioned on the other thread, by specifying the photo coords (pixels), I would assume that the collected GCPs would be right on top of the painted location in the generated ortho.

I do notice in the log that there are multiple “mean georeference error” messages, but I’m not sure that’s related.