基于多CPU的目标识别与跟踪系统设计  

Design of Target Identification and Tracking System Based on Multi-CPU

作　　者：谢勤岚[1] 陈红[1] 

机构地区：[1]中南民族大学电子信息工程学院,湖北武汉430074

出　　处：《计算机测量与控制》2009年第8期1525-1526,共2页Computer Measurement &Control

基　　金：国家自然科学基金项目资助(40774089)

摘　　要：设计了一个实时图像采集、目标识别与跟踪系统;系统功能包括图像采集、运动估计、目标识别、摄像机云台控制等;系统功能主要由3片CPU完成,其中由2片DSP构成的图像采集处理单元实现两路图像采集、目标识别和运动估计,识别和估计结果通过系统总线送给跟踪控制系统;由1片ARM构成的目标跟踪控制系统对2路摄像机的云台和镜头等实现实时控制,使目标一直位于视场中;该系统采样率为10fps,当运动目标的距离在100 m以内,速度小于15 m/s时,可以实现实时识别与跟踪;该系统可以应用于视频监控等应用领域。A system used for real--time image acquisition, target identification and tracking was designed. The system can implement mainly by three CPUs functions such as image acquisition, movement estimation, target identification, camera control. Two image acquisi- tion and processing units, composed mainly of the two DSP CPUs respectively, accomplish image sampling, target identification and move ment estimation. The estimation results is then sent to camera control unit, which uses a ARM CPU, to control two camera consoles and lens, so that the target can be kept within the camera＇s scene. When the sampling rate is up to 10pfs, and the target is within 100 m at velocity within 15 m/s, the real--time tracking can be implemented. The system is fit for video surveillance.

关 键 词：多CPU 图像处理 目标识别 跟踪系统 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]