基于嵌入式Linux的运动目标跟踪视频监控系统  

Moving Target Tracking Video Monitoring System Based on Embedded Linux

作　　者：杨夏[1] 

机构地区：[1]湖南科技职业学院软件学院,长沙410118

出　　处：《西南师范大学学报（自然科学版）》2016年第11期148-154,共7页Journal of Southwest China Normal University(Natural Science Edition)

基　　金：湖南省教育厅科学研究项目(12C1094)

摘　　要：针对运动目标检测和出现严重遮挡时目标丢失的问题,提出了一种基于五帧差分和Surendra背景更新相结合的检测算法,同时设计了结合Kalman滤波和线性预测的MeanShift算法,引入了动态自适应窗口的CamShift算法,并提出了CamShift算法的全自动跟踪策略,减少手工选取目标引起的误差.在实现系统功能方面,搭建嵌入式交叉开发环境,通过相关应用程序的交叉编译实现了在嵌入式平台上目标跟踪的监控系统.实验结果表明,该文提出的算法跟踪效果相比传统的算法明显提高,所设计的视频监控系统运行可靠,改进后算法检测效果好,可以快速完整地提取出运功目标,并且进行稳定的跟踪.For moving target detection, two algorithms have been improved respectively in this paper. There is a algorithm based on the five combined detection of frame difference and surendra background up-date algorithm. For the traditional MeanShift algorithm is easy to lost goals q uickly. By combining detec-ted target centroid to improve, designs the combined w ith Kalman f i l te r and MeanShift algorithm to solve linear prediction, i t has obviously improved the tracking performance compared w ith the traditional algo-rithm . In addition , the dynamic adaptive window CamShift algorithm has been introduced, the CamShift algorithm of automatic tracking strategy proposed, and the error due to manual select target reduced. In the realization of system funct ion , building of embedded cross development environment, and compile the relevant application to realize the monitoring of target tracking system on embedded p la t fo rm. Experiments show that the design of video monitoring system is reliable operation, testing effect is good, the improved algorithm can quickly complete the extraction of the shipment work target and stable tracking.

关 键 词：嵌入式 OPENCV 视频监控系统 运动目标检测 目标跟踪 

分 类 号：TP311.5[自动化与计算机技术—计算机软件与理论]