基于FPGA的多运动目标检测算法实现

Implementation of Multi-moving Target Detection Algorithm Based on FPGA

针对目标检测对实时性要求越来越高的情况,提出了一种基于现场可编程门阵列(Field Programmable Gate Array,FPGA)实现的多运动目标检测算法。该方法通过帧差法提取运动目标,然后基于距离阈值对形态学处理后的图像进行目标分割,最后对运动目标标记显示。系统通过CMOS摄像头采集视频,采用FPGA平台外接DDR3高速大容量缓存,实现了视频图像的采集、存储、目标检测和显示。实验结果表明,提出的多目标检测系统能够有效的实时检测出多个运动目标,并且在分辨率为1 024×600的情况下帧率达到38 fps。

In view of the increasingly high real-time requirement of target detection, a multi-target detection algorithm based on field programmable gate array was proposed. Firstly, the moving targets were extracted through the frame difference method, and then the target segmentation of images after morphological processing was performed based on the distance threshold. Finally, the moving targets were marked and displayed. The system was based on CMOS camera, FPGA, external DDR3 high-speed large-capacity cache, and realized the functions of video image acquisition, storage, target detection and display. The experimental results showed that the target detection system could effectively detect multiple targets in real time, and could reach 38 fps at a resolution of 1 024×600.