Research On Detection Of Moving Object Characters And Guarantee Of Real-time Data Transmission In Traffic Surveillance

In the recent year, traffic congestion in not only developing country but also developed country is getting more and more serious. Usually video camera, image gathering card and computer constitute a video surveillance system for traffic. This system can simulate the visual perception process of the human eye to monitor the traffic condition. It's a transformation of transport with the development of information technology in the 21st century. But because of the complexity and variety of traffic environment, the precision and speed of video surveillance system still can be much more advanced and optimized. Vehicle surveillance is the core part of the traffic surveillance system. According to this core part, the traffic surveillance system , can be separated into three parts, abstraction of vehicle features, vehicle information matching and prediction of vehicle location. The abstraction of vehicle features part can be divided into three parts again, moving object detection, shadow removal and character abstraction.This dissertation mostly researched on detection of moving object characters and guarantee of real-time data transmission in traffic surveillance. First, search on moving object detection, shadow removing and features abstracting, which ensure the robusticity, accuracy and realtime. Secondly, as regard to guarantee of real-time data transmission, test the realtime system and analyse the influencing factors.The detection of moving object: the separation of moving vehicles from video traffic sequence is the basic of distilling vehicle features. The dual background algorithm based on the advanced Surendra algorithm which is put forward by the paper, has increased the speed of background update, and it can also solve the recognition problem when vehicles moving again after station. We can achieve better results.The removal of shadow: after the abstraction of vehicle, commonly there are some concomitant shadows in the object image. It will affect the location of vehicles, the measuring of information and the tracking of objects. Hence, shadow removal has become a necessary part in traffic surveillance system. In the dissertation, we improved the illumination invariance image shadow removal algorithm based on the research of many shadow suppression method. Eventually we compared the improved algorithm and original one with the image sequence in traffic surveillance.The abstraction of vehicle features: the vehicle features abstraction after shadow removing is the premise of feature matching. Only single feature abstraction algorithm will induce real-time or robusticity insufficiency. So the paper mainly pays attention to the integration of different algorithms and to the option for different algorithms based on different environments.The guarantee of real-time data transmission: according to the data type and characteristic of transmission in the video surveillance system, we chose the most appropriate real-time system. After the confirmation of the open source protocol stack RTnet which is based on RTAI, we researched on the TDMA mechanism of RTnet deeply and finished the test of real-time communication on RTnet. At last, we analyzed the effect of memory interaction operation.Finally, we summarized all the research and looked forward to the work in the future.