摘要
描述了基于TI公司TMS320DM642型DSP平台的视频车辆实时检测系统的实现,针对DSP系统在运算和内存方面的局限性,提出了一种线性时间复杂度下的矩形区域多目标搜索算法,解决了DSP栈深度的困难。应用高性能RF5参考框架进行算法构建,通过改进的阴影检测和运动目标检测算法,从而实现视频车辆的精确实时跟踪。实验证实了算法的有效性和处理的实时性,实现了车辆实时检测的微型化。
This paper proposed an approach of implementation of real-time video vehicle detection based on TMS320DM642 DSP platform.Considering the limitations of performance and the stack depth on DSP system,presented a new algorithm to search the rectangular region of multiple targets in linear time complexity.Then,based on RF5 architecture,improved shadow detection and adopted motion detection algorithms.Finally,the implementation on DSP testified the efficiency of these algorithms.Experiment results show that the ...
出处
《计算机应用研究》
CSCD
北大核心
2009年第2期757-759,765,共4页
Application Research of Computers
基金
国家自然科学基金资助项目(60672099)
