Research On Binocular Stereoscopic Vision System Based On Zynq SoC

Binocular stereoscopic vision is one of the important research topics in the field of computer vision.Binocular stereoscopic vision is to imitate the human eye imaging,as well as the human visual distance perception process,with the relative position of the binocular camera to determine the target scene to shoot,and finally based on the binocular image to calculate its three-dimensional information.Binocular stereoscopic vision system generally includes image acquisition,camera calibration,epipolar correction,stereo matching,depth map to restore the five parts.Compared with other 3D information acquisition methods,binocular stereoscopic vision has the advantages of simple structure and low cost,so it has important application value in robot autonomous navigation and depth measurement of workpieces.However,the stereo matching algorithm involved in the binocular stereo vision system is generally complicated,and the traditional general processor such as(CPU)cannot meet the real-time requirement of the algorithm well.The use of binocular stereoscopic vision to obtain the depth of the scene information,is the use of binocular camera to capture the spatial relationship between the single frame image,restore the depth of the scene information.In order to better meet the real-time requirements,this design selects the local matching algorithm in the local stereo matching algorithm for stereo matching.After the simulation of datasets,confirmed the algorithm can get the correct disparity map.In this paper,Zynq-7000 SoC(Field Programmable Gate Array and Dual-Core ARM)is used as the hardware implementation platform for the practical application of the real-time requirement in the stereo matching algorithm.Considering the software algorithm and hardware implementation,the high-level comprehensive tool is used to realize high performance Binocular stereo vision system,and verify the use of this algorithm and field programmable gate array real-time and reliability.Finally,the real-time acquisition,correction and stereo matching of the binocular image resolution of 640*480 and the frame rate of 30fps are achieved,and the result of the disparity map is taken into the real-time output of the HDMI interface output.The total power consumption of the system is about 4W,accounting for about 90%of the programmable logic hardware resources,compared to dual-core ARM processor to achieve efficiency of about 50 times,verify the programmable logic algorithm acceleration.