Human Body Detection And Tracking On Intelligent Surveillance System Based On Embedded System

Intelligent video surveillance technology is an important research direction in the field of computer vision, with a wide range of applications. Moving target detection and tracking is the key technology in intelligent video surveillance, especially human target detection and tracking is a very useful research directions. In the video processing, it need to deal with the large amount of data, and to meet the real-time. Digital signal processor (DSP) has a low power, high performance, programmable features, and TI’s DaVinci series DSP processor with extensive development resources and powerful parallel processing capabilities, can meet the large volume of data and real-time requirements. In this article, the DSP platform is TMS320DM6437EVM.This study is the indoor environment, the camera stationary case, in the intelligent video surveillance system we designed and achieved human target detection and tracking algorithm. Topics include:the design of human target detection and tracking algorithm, the use of matlab simulation tools for algorithm verification, hardware platform configuration about DM6437, DM6437-based algorithm transplantation and the use of Image Processing Library (IMGLIB) and Video Processing Library (VLIB) for program optimization.This human target detection uses adaptive background updating background subtraction method to extract moving targets, through the analysis of the connected domain,namely:the location of the centroidand and the size of area and the aspect ratio and so on, and the non-rigid characteristics, extracts human targets. This article uses a human body tracking target centroid position to represent the position of the body, using the Kalman filter for human centroid position prediction, the prediction location near the target, the target histogram by calculating the similarity histogram matching to achieve human target tracking. This method of Kalman filtering and histogram matching tracking can improve the body tracking speed and accuracy.Using matlab simulation tools to simulate the algorithms of the detection and tracking which have been designed, while on DM6437EVM platform transplanting the algorithms, using CCS3.3development tools, based on DSP/BIOS, using C language to program, complete algorithm programming, and the use of Image Processing Library (IMGLIB) and Video Processing Library (VLIB) for program optimization. Using FVID model achieves video capture, processing and display of results.