The Research On Traffic Congestion Identification Method Based On Road Surveillance Video

With the rapid development of economic, road traffic conditions become more and more complex as a result that the number of cars increases. And accurately determining traffic situations is the basis for resolving the problem of road congestion. With the development of universal road monitoring system, image processing and pattern recognition technology, making extraction of video-based traffic characteristic parameters to determine traffic conditions becomes a research hotspot. In the actual scene, loss of road traffic parameters occurs due to the malfunction of road information system, restoration of this parameter is very important.In order to solve the problem of road conditions distinguish, this paper gets the road traffic characteristic parameters by rroad traffic video processing, then proposes a traffic congestion identification method based on kernel fuzzy c-means clustering(KFCM), while the time-space compression sensing is used to restorate the lost traffic flow data. The main work is as follows:For obtainning the road traffic characteristic parameters from real-time traffic video, we should achieve the goal of detecting the motion of the vehicle firstly. This paper proposes an adaptive Vibe target detection algorithm based on threshold detection, which improves the traditional pixel-level Vibe target detection algorithm. In order to reduce the presence of ghost detection, the ghost suppression methods based on the Otsu threshold are introduced, which could combine the background single pixel discrimination with the characteristics of the entire image. In order to adapt to the situation of large changes of foreground target motion status, we can adjust the rate of background update adaptively according to the motion speed of the foreground objects centroid. Experimental results show that the improved algorithm of this paper, can quickly and effectively suppress ghosting, while improving the target detection accuracy and robustness.Secondly, this paper presents a framework for traffic congestion identification based on KFCM. Traffic jams are checked by three parameters such road space occupation ratio, traffic flow numbers and the whole optical flow speed of road. We get the road by traffic video detection of multi-frame fusion. The road space occupation ratio is the ratio of the number of pixels of foreground objects and the number of pixels of road; get the traffic flow numbers by the virtual coil method and Vibe algorithm; then calaulate the whole optical flow speed of road by Harris corner detection algorithm and H-S optical flow algorithm. On this basis, according to the state having the ambiguity between traffic using KFCM clustering algorithm to find the center of the states of traffic, build traffic congestion identification model, and finally get the current state of traffic congestion by calculating the Euclidean distance. Experimental results show that the proposed method can be carried out quickly and accurately determine the status of road congestion.Finally,the video traffic characteristic parameters obtained during the traffic flow parameters may be lost, the structural characteristics of road traffic volume makes it have a certain redundancy and compressibility, therefore it can be compressed sensing theory in space-time traffic flow parameters repair. This paper constructs a road network traffic matrix, combined with the low rank of the road traffic and time- spatial correlation, we proposed construction method time traffic parameter correlation matrix and spatial correlation matrix and use matrix approximation missing interpolating elements fix on reconstructed traffic data. Result shows that the method can accurately and efficiently recover the missing traffic flow parameters.