邱晓金　陈名松　黄高见

(桂林电子科技大学信息与通信学院　桂林　541004)



基于Cortex-A8的视频监控跟踪系统设计与实现*

邱晓金陈名松黄高见

(桂林电子科技大学信息与通信学院桂林541004)

摘要为降低数字化视频监控设备成本、缩短开发周期和解决PC机上对视频运动目标跟踪时延较大的问题,设计了一种基于Cortex-A8的视频监控跟踪系统。该系统分为嵌入式视频采集传输部分,网络视频流接收部分和视频监控跟踪部分。嵌入式视频采集传输部分,用具有视频压缩功能的摄像头进行视频采集,通过构建嵌入式视频流服务器实现远程监控功能。网络视频流接收部分,创新地将视频流服务器的网页接收代码与OpenCV的函数库结合完成网络视频流数据的接收,极大地降低开发成本和精力。运动目标跟踪部分,提出了一种基于Camshift算法和Kalman滤波器结合的算法,使得系统满足实时性的要求并且跟踪目标在短时间内被遮挡的情况下,仍然可以实现对目标的有效跟踪。最终一套形成集视频的采集、传输、监控、跟踪为一体视频监控跟踪系统。

关键词Cortex-A8; MJPG-Streamer; OpenCV; Camshift; Kalman

Design and Implementation of Video Surveillance and Tracking System Based on Cortex-A8

QIU XiaojinCHEN MingsongHUANG Gaojian

(School of Information and Communication, Guilin University of Electronic Technology, Guilin541004)

AbstractTo lower the cost and shorten the development period of digital video surveillance device and reduce the latency of moving object detection on PC, an innovative video object detection system based on Cortex-A8 is devised. The system can be divided into three parts, including the embedded video acquisition and transmission subsystem, the video stream receiving subsystem, video monitoring and tracking subsystem. To implement the embedded video acquisition and transmission subsystem, video is captured by camera with the function of video compression and embedded video stream server is constructed for remote monitoring. As for the video stream receiving subsystem, the receiving code of video stream server and the OpenCV function library are combined innovatively to receive video stream via network,which can greatly reduce the development cost. A new algorithm based on Camshift and kalman filter is proposed to implement the video monitoring and tracking subsystem. The algorithm fulfills the requirement of real-time and is even even able to track objects that had been blocked for a short time. After all these work, a video surveillance and tracking system with the function of video acquisition, transmission, monitering and tracking is completed.

Key WordsCortex-A8, MJPG-Streamer, OpenCV, Camshift, Kalman

Class NumberTP391

1引言

在人类的感官系统所接收的信息中,视觉的信息比例是最大的,占到了80%[1]。除了文字和声音,视觉是最客观、最形象、最逼真的描述,是人类最重要的信息承载体。由于视频的信息量大,人工处理速度慢等原因,迫切要求计算机对视频进行编辑、处理。因此,计算机对视频处理的一个重要的方向就是视频监控。目前国内的视频监控也已从传统的模拟,模数混合监控逐步向数字化、智能化视频监控方向发展[2],高清视频监控也正逐步进入人们的视野。……