摘要
针对目前视频目标检测匹配跟踪算法不能满足视频监控的高实时性要求,不能满足当前硬件平台主流技术的问题,研究了差分目标检测和匹配跟踪算法的优化实现问题。为优化算法减少计算量,选用了连续帧训练背景的方法,利用背景差分检测出场景中的运动物体,采用模板匹配跟踪算法,将目标检测和跟踪算法在DM642上优化并实现。仿真结果表明,经过算法和程序级的优化,程序运行时间大大减少,可在CIF格式下较好地进行多物体的实时检测与跟踪。
Intelligent video monitoring system must fill demand about real time.This requires that detection and tracking algorithm are simple,and platform runs quickly.Consecutive frames training background method is used in this article.The subtraction between background and frame detects moving object.The template matching algorithm is used to track object.These algorithms are realized on DM642.After optimization in algorithm level and program level,the program can be run in real time with CIF（30fps,352＊288）.
出处
《计算机仿真》
CSCD
北大核心
2010年第10期234-237,245,共5页
Computer Simulation
关键词
运动物体检测与跟踪
算法和代码优化
差分方法
Moving objects detection and tracking
Optimization on algorithm and program
Subtraction
