An improved algorithm for feature point matching

Abstract

According to the two video cameras in system of binocular stereo vision are at the same horizontal line and the collected images have the public areas, an improved feature point matching algorithm is put forward. This algorithm extracts the public areas from the two images, makes the feature point locate in a rectangular area and reduces the searching range in the matching process; then, it completes the coarse matching by the cross template matching and matches the feature point accurately using the epipolar line constraints and removes the edge points mismatched. Experimental results show that this method is simple and efficient, and it has the certain practical.