Binocular vision position algorithm using hue-saturation histogram back-project combined with feature point extraction

Abstract

In low-cost binocular vision system, an inexpensive non-measure camera usually results in a low visual positioning accuracy. To obtain higher visual positioning accuracy, there exist some algorithms to reduce or compensate positioning errors. A binocular vision positioning algorithm based on H-S (hue-saturation) histogram back-projection combined with feature point extraction is proposed in this paper. In this algorithm, the H-S histogram back-projection method is employed to extract the feature points of images from the left and the right cameras; and then, the feature points of the left image and the epipolar constraints are employed to correct the feature points of the right image, to achieve binocular vision positioning. In order to further improve the positioning accuracy and reduce the cost of computation, the SURF (speed up robust feature) algorithm is used to extract the interest points of both left and right images, and an indirect stereo matching approach is employed to improve the feature point extraction of the right image. Experiment results show that for the low-cost visual positioning system this algorithm can effectively improve the positioning accuracy.