Measuring object size with stereo vision

Hi, I am working on a stereovision project trying to measure diameter of a object using the matlab triangulate function. The way I do this Is using blobanalysis on both Images to detect the objects(two red circles on a piece of paper pinned to a wall). Then I Insert a square box around both circles, and uses triangulate on the corner coordinates for the left top corner In both Images, and the same for the right top corners. Then I substract the resulting X koordinate on the left side corner from the resulting right side corner X coordinate. Is this a possible way to get the diameter? because I am getting numbers that are way off from the reality.

I have done this with a single calibrated camera, and got good results. But now I want to do the same with two cameras, hopefully some of you could spare some time to explain what I am doing wrong.

[Stalin Sabu Thomas]

If you get a bbox of the detected object, here a circle, isn't it safe to say that the width of the bbox is the diameter itself? the third column of the bbox is the width. I don't think you need to have a stereo image or triangulation to get the diameter of the object as it can be done with one single camera image. But one possibility is to triangulate the left & right top corners (world coordinate points) of the bbox and get the distance between them using Euclidean distance ie. sqrt( (x1-x2)^2 + (y1-y2)^2 + (z1-z2)^2 )

Answer 1

[Tohru Kikawada]

Have you tried this example?

https://www.mathworks.com/help/vision/ref/triangulate.html

Generally, the accuracy of stereo vision depends on the estimated stereoParameters. You might want to review the results of stereo camera calibration.

[Shaik Ahmad]

Thank you for the information. I have seen the implementation about triangulate function, it explained how to calculate the distance from the camera to the object. But it didn't mention about the dimension of the object.

Could you please provide some information related to measuring the object dimensions.

Thank you