Design And Implementation Of Moving Target Detection System Based On Binocular Stereo Vision

In recent years,road traffic detection and advanced driving assistant system have become increasingly important to reduce traffic accidents,and with the rapid development of technology for target detection and visual ranging and the increasing complex traffic environment,only visual ranging for stationary targets or non-depth target detection can no longer meet traffic safety requirements.Therefore,an efficient moving target detection system with depth information is required.As a popular research field of computer vision,binocular stereo vision technology can be widely used in road traffic and car driving to achieve target detection and target ranging,thereby improving driver safety.At present,the depth perception system based on binocular stereo vision is mostly implemented by pure hardware;although the real-time performance of the system is guaranteed,the system lacks human-computer interaction.What's worse;the moving target detection method based on embedded system is obviously inferior in measurement accuracy and real-time performance.Based on the above problems,this paper proposes a moving target detection system based on binocular stereo vision,which is mainly composed of four modules: data acquisition and storage,depth information perception,target detection and target ranging.Firstly,considering the low detection accuracy of moving targets in the embedded system,this paper designs the moving target detection algorithm based on the depth information of binocular stereo vision,and the problem of "ghost" phenomenon and incomplete target detection in traditional algorithms is improved.Then,the target-ranging algorithm is designed based on the target detection region of the improved algorithm.The algorithm classifies and discards the depth information in the region,and obtains the distance of the moving target by the mean method.Finally,this paper proposes a hardware and software co-design approach for moving object detection system based on binocular stereo vision.The approach combines the advantages of hardware and software to perform hardware and software partitioning of the system,and designs efficient and accurate data control and transmission between hardware and software,and combines the specific algorithm and logic control in the system to design the system.The approach has good human-computer interaction while ensuring high real-time performance of the system.The experimental results show that the frame rate of image data of the moving target detection system designed in this paper can reach 121.43 frames/second with a resolution of 640*480 when the main frequency of the FPGA is 100 M.And the system can achieve real-time display and human-computer interaction with the control of the graphical user interface.At the same time,the improved target detection algorithm in the system can accurately detect the moving target,the target ranging algorithm has an error value of less than 2.0% within seven meters,and they all have better robustness and precision.All the design and test results of the system have achieved the desired results.