A Study On Image Processing In Video Detection System Of Traffic Incident

With the rapid development of economy and the fast construction of roads,the number of vehicles increases largely.However,owing to the backwardness of road management,traffic incidents occur frequently.Traffic incidents seriously affect operation of capacity of roads;meanwhile,it would threat people's safety on the road. In order to reduce the effect of traffic accidents,a great deal of human and material resources have been invested on this research in many counties in the world.There are many methods for detecting traffic incident.However,as compared with other methods,Video detection is much better due to its own advantages.Video detection has become a research hot research topic in traffic incident detection system. Technologies of moving object detection and tracking,which have a direct impact on traffic incident judgment,are the base technology of Video detection.This thesis focuses on technologies of moving object detection and tracking based of microwave monitor platform in the lab.The main contents are as follows:(1) This thesis first gives the definition of traffic incident and its effect,and compares methods of traffic incident detection.The video detection has been paid attention due to its advantages.(2) Second,the components of traffic video monitor system have been introduced in detail.(3) The background subtraction method has been used to detect vehicles from image.This method consists of background building and background updating, threshold segmenting noise-removal processing,and can be extracted label processing of Connection region.As a result of this method,the whole object vehicles.(4) Fourth,based on moving object detection and the features of moving object, the thesis also proposes a matching tracking method based on Kalman Filtering,to predict the next movement of the object in the object zone and estimate its possible location at the next moment.Subsequently we could search for matching object in a small zone and create the link of the object between frames.(5) Finally,the incident detection algorithms are implemented based on Visual C++.The result shows that the algorithms proposed in this thesis are available.