How to test SfM for multiple view?

Hello everyone, I try to test the SfM with my own data. I want to know

(1) in this command, load (fullfile(imageDir,'cameraParams.mat')) , Is it need to do camera calibration to load cameraParams.mat? If need, I already try the camera calibration toolbox in matlab, but it's not ok. So, I do the cameraParams.mat in SfM demo folder is moved to my current own image data folder. And then

(2) i use helperEstimateRelativePose , but they showed this error

Error using helperEstimateRelativePose (line 70)

Unable to compute the Essential matrix

Error in sfM_multi_own (line 39) [relativeOrient, relativeLoc, inlierIdx] = helperEstimateRelativePose(matchedPoints1, matchedPoints2, cameraParams);

I don't know how to do and I'm really confused. Please help me! I want to listen for your suggestion seriously.

[Thida Aung]

now I face the same error as you.

[Joe Joe]

i don't know how to resolve this until now.

[Victor Medina]

I just managed to solve this same error. In my case, it seems I was using too many images, so the resulting equation system was overdetermined and, therefore, the matrix could not be computed. I just tested with a number of images similar to the example (6, in my case) and enough camera movement from frame to frame, and it works like it should.

Hope this helps.

[UBAID ULLAH]

@Victor Medina ... how'd you solve this problem? i cannot solve it even after number of images reduced to 6. still it gives me the same error. can you please help me out sir?

[Mostafa Hossam]

same error please help

[Abdalmalek Abu-Raddaha]

Have you guys solved this?

I modified the helperEstimateRelativePose function decreased the confidence and increased the max distance but still getting the same error

Answer 1

[jhz]

Consider increasing the maximum distance or decreasing the desired confidence. You can do this by modifying the Matlab built-in function helperEstimateRelativePose.m: line 43. Then you can add as many images as you want. After modification, it should look like this:

 [E, inlierIdx] = estimateEssentialMatrix(matchedPoints1, matchedPoints2,...
    cameraParams, 'Confidence', 50, 'MaxDistance', 5);

But be careful editing the built-in functions. In my case, I modified the function and saved into another folder by another name and add that folder into path. I hope this will help someone.