The Extraction Of Depth Information About The Object Based On Binocular Vision

Binocular vision technology is a hot spot in the field of computer vision, which is widelyused in machine vision, sculpture and architecture, engineering, testing, and virtual reality, etc.Binocular vision technique transform two or more2-d image to the depth information of the3-d scene with the steps of camera calibration, feature extraction, feature matching and3-dreconstruction. Based on the analysis of the related technology and a large number of relatedpapers, this paper mainly do research in the field of feature matching and3-d reconstruction.The main research is as follows:(1) This paper get more matching points by improving the gray matching method, itsprinciple is that sharpen the image gray gradient with the technology of the pixel differenceproduct instead of the traditional mode gray magnitude, which enhance the high frequencycomponent of the image, and make the image details more clear, so as to get more matchingpoint.(2) Because of the less number of point cloud based on binocular vision, it is difficult toform the outer contour with the sparse point cloud in the step of3-d scene reconstruction.Unified the coordinate system with the second calibration between the Kinect depth map andthe binocular camera sparse point cloud, this paper uses Kinect to assist the3-d reconstructionof the sparse point cloud, and extract the object contour with the most depth value constrainttechnology, which gain better sensory effect.The paper takes image by the binocular camera, and complete the whole3-dreconstruction process with camera calibration, the SIFT feature matching and software programming, which verify the feasibility of improving methods, and achieve the intendedpurpose.