Optical Flow Method Accelerate Research Based On Hybrid Reconfigurable Hardware Platform

Motion target detection and tracking in image sequence, is one key research direction in the field of image processing and pattern recognition. Optical flow method is a kind of important moving target tracking algorithm. This method can achieve high calculation precision, and obtains movement direction and speed of movement target directly. But the calculation amount of optical flow method is big; lead to poor real-time performance in the existing serial computing platforms. So the designing of real time flow algorithm and corresponding hardware processing platform has become the research focus at present.Lucas-Kanade algorithm is an important method for optical flow calculation; it has high calculation precision and simple structure. Also, it is very suitable for hardware acceleration implementation, and achieve very good balance in terms of real-time and accuracy.Combine heterogeneous computing technology with optical flow method, this paper designs hardware acceleration of optical flow algorithm based on a heterogeneous multi-core platform system. By making sufficient use of computing platform computing resources, the efficiency of the optical flow method can be improved. Also, the application level of heterogeneous computing platforms and optical flow method are improved. This paper uses the Zynq-7000 chip's FPGA part to implement prophase and metaphase stages of image processing algorithm with large data amount but less algorithm complexity. And chip's CPU part was used to implement algorithm with high complexity but small calculation amount in later stage. In this paper, by improving optical flow method of calculation in the process of moving target detection, and using Zynq heterogeneous computing platform to speed up the calculation, expected processing result and effect was achieved.