Study On Intelligent Road Surveillance System Based On DSP

With the rapid development of computer vision technology and embedded technology, road surveillance system is more digitalization, intelligent and convient. Intelligent road svirveillance system is designed to acquire and analyze the information of the moving target within the scope of surveillance accurately and quickly, and to provide decision aids to the traffic management. The general roadmap of this design includes image preprocessing, background establishment, background updating, moving target detection and moving target tracking. The main research topics of the paper are to detect and track the motion targets, which are the key technologies of intelligent road surveillance system.In the research on moving target detection, this paper summarizes some common methods of motion target detection:optical flow method, frame differential method and background subtraction method, and analyzes their advantages and disadvantages respectively. Then we achieve the motion objective detection based on the background subtraction algorithm, which includes the initial background establishment, background updating, background subtraction, image binarization as well as the morphological processing.In the area of motion target tracking, this paper studies the tracking method based on Kalman filters and particle filter, and introduces the basic principles and steps of each method. We also establish the motion model by utilizing Kalman filter to match the targets after predicting the general area, and realize the multiple targets tracking ultimately. As to the particle degeneracy problem with standard particle filter, an improved particle filter based on particle swarm optimization is proposed, which has bean proved the robustness and the validity of the algorithm by the experiments of single target tracking.This paper designs and develops a hardware development platform based on the TMS320DM643. The whole system consists of several functional modules, such as video collection module, DSP video process module, video display module, storage module, etc. The software design and development based on DSP side by adopting TI’s DSP/BIOS and RF5reference framework. At last, we transfer the related image processing algorithm from PC to DSP and achieve the optimization of the whole system proposed instantaneity.