Research on Fog Ranging Based on Binocular Vision

Abstract

For the robot to travel in foggy weather, the target feature information is not obvious, which affects the robot’s ranging speed and accuracy. In order to realize the rapid and stable advancement of the robot, a binocular vision ranging detection method for the target under foggy conditions is designed and implemented. First, the dehazing algorithm MSBDN (Multi-Scale Boosted Dehazing Network) is used to dehaze the target image to restore the characteristic information of the target; then, the Zhang Zhengyou calibration method is used to obtain the internal and external parameters of the camera; then, the semi-global matching algorithm SGBM (semi-global-blockmatching) to match pixels to obtain the target disparity map; Use WLS (weighted least squares) filtering to smooth and denoise the initial disparity map to obtain the optimal disparity map. Convert disparity map to depth map. The experimental simulation shows that the method can restore the actual features of the image better in the poor image quality of foggy days, improve the distance measurement accuracy, and meet the requirements of robot travel in foggy days.