特定对象的立体对应和3D位置（ROI）

Question

I am interested in getting the 3D location of a specific object (Using stereo vision) given its location and ROI in the left image, in addition to obtaining its location in the right image.

All I found online is how to get the disparity map using (StereoBM::operator()/ cvFindStereoCorrespondenceBM).

What is the approach? How can I get the disparity of that object? Is there a direct way of doing it after getting the disparity map or I should match the object first in the right image to get the disparity?

Can you suggest a good algorithm to match the object? please include codes in the answer.

Answer 1

Your task has the following steps:

1) Camera calibration
2) Extracting features (manually or via algorithms. For example - SURF, SIFT);
3) Matching features (it's up to you. OpenCV has a brute force matcher)
4) Fundamental and Homography matrices estimation
5) Rectification (OpenCV stereo matcher works only with rectified images)
6) Disparity map
7) Triangulation based on your camera model (might not be pin-hole)

Good explanations are available from these books with code:
1) OpenCV 2 Computer Vision Application Programming Cookbook by Robert Laganière
2) Learning OpenCV by Gary Bradski
3) An Introduction to 3D Computer Vision Techniques and Algorithms by Boguslaw Cyganek and J. Paul Siebert

Hope it helps! =)