中文摘要：

该文提出一种改进式背景差分算法,并应用于监控系统中.针对人流量较少的监控情况,提出一种基于计算机视觉的嵌入式监控系统解决方案,其以DSP DM642为核心处理芯片,可对3路视频视角同时处理.系统利用动态权值的改进式背景差分算法对视频流进行实时监控,若发现异常事件,则自动存储一段时间的视频数据,并利用H.264压缩后保存至外存中以供事后取证.由于监控算法巧妙,普通的SD卡即可替代传统的硬盘,系统精简,使用方便.试验表明：该系统灵敏度可调,非常适用于外景和内景的库房监控.

英文摘要：

In this work, a kind of improved background difference algorithm is put forward and applied in the monitoring system. A novel embedded monitoring system based on computer vision is presented to build simple and intelligent monitoring platform for less population flow. DSP DM642 is used as the core processing chip, which can process three video views at the same time. The improved background difference method based on dynamic weight is utilized to perform real-time monitor to the video. When the accident is discovered, the system would capture a section of video automatically, then H.264 compression algorithm is used to compress the video data and the video data is saved in the external memory as evidence. Thanks to ingenious surveillance algorithm, the traditional hard disk can be replaced by the SD card. Furthermore, the monitoring system is a simple and easy process. The experimental results show that system sensitivity is adiustable, and this system is suitable for out-door and in-door warehouse monitoring.