高速动态图像目标识别算法的设计与实现

Design and realization of high-speed dynamic image target recognition algorithm

摘要:

为了实现对高速动态图像进行稳定的快速识别，并将识别出的目标图像传给上位机，应用优化的背景差分原理，设计并验证了一种可并行计算的高速动态目标识别算法，利用FPGA作为主控芯片实现了有效图像的高效采集以及实时传输功能。首先该系统采用流水线处理方式实现对数据的实时采集，然后利用乒乓操作来实现目标识别算法，通过对DDR3进行分页操作，将识别后的动态目标图像进行缓存，最后采用USB3.0芯片实现上位机与FPGA进行实时传输数据。实验结果表明，所设计的动态目标高速识别算法可以有效识别出6mm的BB弹，捕获率高达99%，同时该系统可以实现动态目标数据的实时传输。

Abstract:

In order to realize the stable and rapid recognition of high-speed dynamic images, and transmit the recognized target images to the host computer, using the optimized background difference principle, a high-speed dynamic target recognition algorithm that can be calculated in parallel is designed and verified. FPGA as the main control chip realizes the efficient collection and real-time transmission of effective images. First, the system adopts pipeline processing to realize real-time data collection, and then uses ping-pong operation to realize the target recognition algorithm. Through paging operation on DDR3, the recognized dynamic target image is cached, and finally the USB3.0 chip is used to realize the upper computer. Real-time data transmission with FPGA. Experimental results show that the designed high-speed recognition algorithm for dynamic targets can effectively identify 6mm BB bullets, Capture rate up to 99% and the system can realize real-time transmission of dynamic target data.