基于OpenCL的光流法对运动目标检测跟踪应用  

Application of Optical Flow in moving target detection and tracking based on OpenCL

作　　者：夏雨 秦工[1] 谢烨 熊绍薇 吴琦 Xia Yu;Qin Gong;Xie Ye;Xiong Shaowei;Wu Qi(School of Artificial Intelligence,Jianghan University,Wuhan Hubei,430056)

机构地区：[1]江汉大学人工智能学院,湖北武汉430056

出　　处：《电子测试》2020年第23期41-42,96,共3页Electronic Test

基　　金：湖北省高等学校2019年省级大学生创新训练项目。

摘　　要：该系统采用LK(Lucas-Kanade)光流法对摄像头捕获的运动目标实现检测和跟踪。但由于LK光流法算法冗杂,时效性差,在对实时性有苛刻要求的情况下并不适用。为解决以上问题,引入基于OpenCL(Open Computing Language,开放运算语言)的LK光流异构并行算法,而设备端采用FPGA(可编程门阵列),实现硬件加速。实验结果表明,该算法对比于普通CPU上调用OpenCV的API对图像处理,算法处理速度上实现了很高的加成,实时性也得到有效的改善。The system uses LK(Lucas-Kanade)optical flow method to detect and track the moving target captured by the camera.However,it is not applicable in the case of strict requirements on real-time performance,because of the complexity and poor timeliness of LK optical flow algorithm.To solve the above problems,a parallel algorithm for LK optical flow based on OpenCL(Open Computing Language)was introduced,and FPGA was adopted as the end device to realize the hardware acceleration as well.The experimental results show that the above algorithm achieves a high addition in processing speed and an effective improvement in real-time performance compared with the common multi-core CPU which calls OpenCV API for image processing.

关 键 词：OPENCL Lucas-Kanade光流算法 OPENCV 并行计算 FPGA硬件加速 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]