A New Method for Stereo Matching Based on Fiducial Marker

This paper proposes a new method to automatically recognize and stereo match fiducial markers at the stage of data coarse alignment in 3D profile measurement. Markers were detected by OpenCV contour extraction function, a set of criterions were proposed for eliminating pseudo markers. Then fiducial markers were divided into different encoded regions by projecting encoded images on target object. Encoded regional identity and epipolar constraint were combined to achieve stereo matching. The algorithm has been approved by experimental test that markers were recognized and matched correctly.