Research On DM642 Based ITS Supervision System

Intelligent traffic supervision system is an important branch of Intelligent Transportation System. With the development of computer vision algorithms and hardware processing speed, computer vision technology is widely used in ITS system.Detecting moving vehicles and related traffic informations is an import task in ITS's research. Vehicle detecting result is the foundation of ITS. A good detecting method would greatly improve overall system's performance. This paper studies on how to detect vehicles in realtime road and develops a DM642 based vehicle detecting system to finish this task. Moreover, we also improved the original vehicle detecting algorithm to make it more efficiently and more accuracy. Experiment results indicate that our system is stable, efficient and has broad prospects. The major work and achievement of this paper are presented as follows:1. This paper used a vasual based detecting methold to detect vehicles. This method could provide rich vehicle informations with less cost and easy to be implemented. After studied exist motion detecting algorithm and other intelligent supervision system, we used a running-average background model to segement vehicles. Further, focused on the existing problem of traditional algorithm, we adopte many new methods to optimize segmentation results. We use two-threshold method to avoid choosing threshold's problem, use HSV space shadow detecting method to eliminate the shadows' affection, use a motion feedback method to eliminate blur introduced by continued vehicles. This algorithm significantly improves the accuracy and reliablility.2. This paper studies on the existing ITS supervision system's architecture. The original system uses PC in monitoring center to process multiple vehicle video streams. However, this has cause a problem of high computation and network bandwidth cost. In our system, we use DSP based system to finish vehicle detecting task. And our monitoring center only needs to combine multiple detecting results to make traffic controls. This architecture boosts overall system's performance.3. We finish the designment of DSP based vehicle detecting system with TI's DM642 processor. Then, with the help of DSP/BIOS development tools and RF5 reference framework, we finished the transplant of our vehicle detecting algorithm.