Fast stereo matching based on probability assignment and ORB operator
Hui Gary Zhang, Zhaoxing Li · Journal of Interdisciplinary Mathematics · 2016
A stereo matching algorithm based on probability assignment is proposed to meet the real-time demand of stereo matching using SURF(Speeded Up Robust Features). In order to accelerate search of matching points, probability assignment of feature points is acquired based on the correlation between fuzzy clustering and probability assignment, as well as the necessary conditions of information entropy; then, when extracting ORB feature points, a data set is used to record the sequence of feature points after decreasing the edge effect; epipolar line, uniqueness and sequence constraints narrow down the search space; the Hamming distance between descriptors are calculated and the nearest neighbor matching becomes the stereo matching result. Experimental results show that the proposed algorithm accelerates matching and maintains high accuracy, even when applied to non-standard pictures.