A binocular stereo vision system and its parallax optimization method

Aiming at the problem that the output depth map of binocular stereo vision system has low accuracy and less effective depth range, a binocular stereo vision system is developed, and the hardware selection and construction and the design and development of the algorithm are completed. In order to improve the effective range of disparity map, a disparity filling method is proposed. The small connected regions are selected by using the connected region labeling algorithm of binary graph, and the small connected regions that meet the requirements are filled with the edge mean value. In order to reduce the error of converting disparity to depth at a long distance, a method of converting disparity to depth by polynomial regression is proposed. The experimental results show that the effective depth range of the depth map output by the binocular stereo vision system is significantly improved, and the accuracy of disparity conversion depth is significantly improved.