search

dicengine / dice

Digital Image Correlation Engine (DICe): a stereo DIC application that runs on Mac, Windows, and Linux
Other
342 stars 125 forks source link

High epipolar error and RMS error #158

Open KartikaAhuja26 opened 4 years ago

KartikaAhuja26 commented 4 years ago

Untitled I was trying to perform a stereo DIC calibration. I have captured 30 images of the calibration target in different positions. When I perform the calibration, I get a high value of RMS error (4.5). The resolution of images taken is 1280 *800. What can I do to reduce epipolar and RMS error?

dicengine commented 4 years ago

Have you tried adjusting the cal options? Playing with some of the options might reduce the RMS error.

KartikaAhuja26 commented 4 years ago

Thanks, I tried adjusting the threshold and stereo calibration is working fine now. image

But the model displacement value I am getting is not coming correct. I displaced my object by 5mm but the displacement coming is 3.2 mm. How can I improve accuracy? I have used 30 images for calibration. image

dicengine commented 4 years ago

Could it be that you have the wrong pattern spacing? It seems that rather than 1.0mm, it would be 1.563mm.

KartikaAhuja26 commented 4 years ago

DICeCalTarget.pdf I have used this file as a calibration pattern and pattern spacing(center to center) distance is 26 mm approximately. Untitled Still, But the model displacement value I am getting is not coming correct. I displaced my object by 5mm but the displacement coming is 3.2 mm. What is the minimum stereo displacement that can be captured accurately?

l13r commented 4 years ago

I used a similar calibration target with a different spacing and I had the same problems with epipolar and RMS error. Then, I decided to change the calibration target and I used a checkerboard target. This type of patter improved the error in my case. Regarding your problem, I think the number of images is insufficient. You have 30 images and most of them are skipped. This can produce that the stereo calibration is wrong and the displacements are not measured correctly. I think you should take a larger number of images (amore than 50).

KartikaAhuja26 commented 4 years ago

Thanks a lot. This time, I used a checkerboard for stereo calibration and it reduced the values of epipolar and RMS error to around 0.3. Now, I am getting accurate values for displacement.