Design And Implementation Of Motion Target Detection System Based On FPGA +

Moving target detection is the key technology of the target search and track system, and with the development of mobile devices, moving target detection problem under dynamic background has become a key issue to be solved. However, due to the complexity of moving target detection algorithm in dynamic background, it is difficult to guarantee the real-time of the algorithm. Therefore, to solve the above problem, this paper presents a scheme of moving target detection algorithm based on FPGA+DSP.Study of this subject is divided into two parts:hardware platform design and the implementation of moving target detection algorithm base on FPGA and DSP. First, according to the actual needs of the system, we designed the schematics and PCB of FPGA circuit board and completed the soldering and debugging of this circuit board. Then introduced the basic principles of image matching and moving target detection, and emphatically analyzed four corner detection algorithm:KLT, SIFT, Harris and FAST, after the simulation-based comparison, we selected the KLT feature point extraction and LK optical flow as the FPGA implementation of image matching algorithm. Moving target detection section introduced the solution process of affine transformation matrix based on RANSAC and the MHI algorithm. In this paper, we improved the MHI algorithm to solve the hollow phenomenon. We optimized the structure of KLT feature point extraction method and LK optical flow algorithm in FPGA, and the use of adaptive threshold and watchdog method improves the automaticity and stability of this algorithm. We realized the affine transformation matrix based on RANSAC and the improved MHI algorithm in DSP by the use of multi-core parallel computing method. And the communication of FPGA and DSP is based on the RapidIO protocol by the SMA interface.Through a lot of debugging and experiments on the platform of FPGA and DSP, we can come to a decision that this method can not only ensure the accuracy of the moving target detection algorithm, but also satisfy the real-time of the algorithm.