Obstacle detection technology based on 2D image segmentation and stereo vision fusion

• 1、College of automation, Beijing university of posts and telecommunications

Abstract：Obstacle detection based on vision sensors and location has been had many research, but there are various deficiencies.First detection algorithm based on TOF camera sensor surface reflector materials due to light and obstacles will lead to the influence of the depth data detection error is bigger.Obstacle detection algorithm based on monocular vision sensor can\'t detect the obstacles of 3 d data.And based on 3 d point cloud segmentation algorithm of binocular stereo vision can only be obstacles to do a simple segmentation, target edges due to the limitations of stereo matching algorithm area appears a lot of false matching points, resulting in 3 d data object detection accuracy is not high.In order to solve this problem, the paper puts forward a kind of image segmentation based on two-dimensional and three-dimensional point cloud data with the combination of obstacle detection and localization algorithm.In conclusion by stereo matching algorithm 3 d point cloud, so as to obtain the three-dimensional information, on the basis of introducing the edge information of the object, get the more accurate 3 d information of object.The experimental results show that the proposed algorithm is better than single by stereo matching algorithm to obtain the three-dimensional information of accuracy to improve a lot

Keywords： Machine Vision Stereo Matching 2D Image Segmentation 3D Information