Visual Object Detection And Tracking Based On Dual-core Embedded Platform

With the development of digital signal processing, and the decline in hardware cost, surveillance technology will lean more and more toward frontal intelligent analysis. In this paper, according to the developing direction of intelligent surveillance, we design a frontal intelligent moving object detecting and tracking system based on a dual-core embedded processing platform. With this system, we can automatically detect and trace the moving target in the monitoring setting.We use background subtraction based on adaptive Gaussian background model to detect moving object, and we trace the moving object accurately by using its color histogram and motion state information based on SIR particle filtering solution. In this paper, we adopt DaVinci DM6446 EVM, which is a platform with ARM+DSP open dual-core structure. Firstly, we package the detecting and tracking algorithm in the DSP section, and then generate the bottom layer library. After that, in ARM section, we adopt modularize design method to build a system which has multiple functions including video capturing, intelligent processing, output displaying, and performance testing. Also, we can transplant much more video intelligent analysis algorithm in this system.After testing the performance of the system, we show that the moving object tracking algorithm based on DM6446 dual-core embedded system performs well. It realizes real-time processing of videos with CIF resolution.