Design Of 3D Reconstruction System Based On Embedded Stereo Vision

With the continuous development of science and technology,uav autonomous navigation and unmanned technology will be applied in more occasions.In order to achieve these key functions,3d reconstruction technology is indispensable.It is mainly based on laser reflection,deep camera and 3d reconstruction of binocular stereo vision.Compared with other two approaches,binocular camera's low cost,light weight,convenient installation makes it has a bright application prospect in the future.This article is based on binocular stereo vision,as the basis of 3 d reconstruction,mainly completed the binocular image collection,binocular camera calibration,image calibration,stereo matching and 3d reconstruction works.Build an embedded platform based on raspberry pi and successfully run on the platform.The main work of this paper is as follows:(1)In order to solve the mapping relationship between the spatial three-dimensional coordinates and the image coordinates,the calibration and correction of binocular camera are realized.In this paper,the camera reference model is researched,and the transformation of each coordinate system between the cameras is deduced,and the binocular camera is calibrated using the method of zhang zhengyou camera.Researching the constraint condition and matching strategy of image matching in stereo vision,and to make polar correction for binocular camera.(2)In order to obtain the parallax information of the left and right views,the stereo stereo matching algorithm is realized.The basic principle of stereo matching algorithm is researched,and focus on Semi global matching algorithm SGBM and NLCA algorithm.An improved method for NLCA is proposed and implemented.Finally,the matching algorithms are compared in the two aspects of mismatching rate and processing speed,and the optimal matching algorithm is selected as the stereo matching algorithm in this paper.(3)An embedded platform based on raspberry pi is built.Set up the hardware and software environment to open the raspberry pie camera.For unmanned aerial vehicle(uav)in the process of actual flight jitter frequency will be high,that may lead camera out of sync.Therefore,a software optimization based on the image acquisition of raspberry pi camera is proposed,and the time gap between the left and right cameras is shortened,so that it can be more suitable for actual flight.(4)The QT ground station based on PC platform is realized.Using Opencv and OpenGL mix programming,and completing the collection of images,the import of the calibration parameter,the function of stereo matching and 3D measurement and Open GL 3d point clouds reconstruction of images,the 3d reconstruction based on binocular stereo vision is realized.