Research Of Video Traffic Incident Detection Unit Based On DM642

As the video monitoring system becoming bigger and bigger, the video monitor content transforms from the visual explanation to the automatic explanation. Therefore, intelligent video monitor becomes necessary for the development of the video monitoring technology. Using advanced compute methods and technique such as computer vision, nerve network, and fuzzy logic and so on, the video traffic event detection system based on image processing carries on the event detection, the vehicles recognition and the road monitoring. It can forecast and discover the accident and obtain the important traffic parameter such as traffic flow, the vehicle speed, and the path's spatial occupancy proportion. Because of the limited computing power and the communication band width, the existing central processed video traffic event detection system is unable to apply to the large-scale video detection system.Begin with analysis to the current traffic event detection system, this article proposes distributional intelligent event detection system suitable in the large-scale detection system,and has conducted the deep research to the important component to the system——Video Traffic Incident Detection Unit (VTIDU). VTIDU uses high speed DSP processor DM642 as the core of the hardware system, and has the function of high speed video image gathering, intelligent vision analysis, video compression and the network transmission. It has the excellence of non-contact examination, convenient setting, powerful function, and credible performance.This article works over the software and hardware designing of the VTIDU and the intelligent vision algorithm in the embedded environment. DSP and the hardware system of its periphery modules have been given in the hardware designing. In the software designing, the function of the VTIDU and the realization of each module are given, including equipment's driving, transplant and optimization of H.264 and the video traffic event detection system, setting and controlling of the video server and the system based on the TCP/IP protocol. In the algorithmic analysis, aiming at the specific compute environment of embedded machine, the paper researches on the movement target detecting and tracking on VTIDU, and proposed the intelligent vision algorithm in the embedded environment. After the test, VTIDU can establish the TCP connection with many clients and transmit the compressed video flow by TCP or UDP. The compression speed of H.263 can reach 30 frames per second, while the speed of H.264 reaches 8 frames per second. The accuracy of the traffic parameter examination reaches above 90%.The traditional monitoring system can be promoted and upgraded, and the distributed intelligent monitoring system can be constructed which has the quality of "entire path monitoring, active warning, artificial confirming, automatically video switching and the immediately system response". The system is stable, safe and reliable and has high application value.