Research On Recognition Of Urban Traffic Spillover Based On Video Information

With the increasingly severe of traffic congestion problems every year, traffic spillover phenomenon, which was one of the important manifestations of traffic congestion, had become a hot issue. The Automatic Incident Detection (AID), which is one of the vital parts of Intelligent Transportation System (ITS), plays an important role in improving intellectualized management level of urban transport. In recent years, Video detection technique had developed rapidly and been widely used in various professional fields due to its high reliability and fast detection speed.In this paper, video detection technique was used to be the major technical means. Geometric model、occurrence mechanism and properties of traffic spillover were described and defined firstly as the foundation of following research. In order to achieve the automatic identification of traffic spillover phenomenon, algorithms were proposed innovatively which aimed to solve the problems of traffic lights detection and vehicles queuing detection:To solve auto-recognition problem of multi-type traffic lights in complex background, a novel method was proposed. Firstly, self-localization was realized by brightness value division、geometric features analysis and classification statistics. Along with signal color judgment by K-means clustering algorithm, the type and direction information of traffic lights would be acquired by analyzing the histograms of foreground objects; accordingly auto-detection was realized. The experimental results showed that the proposed method had high detection accuracy for multi-type traffic lights in different scenarios. To solve auto-recognition problem of traffic spillover by video detection technique, a novel algorithm based on fuzzy inference theory was proposed. Firstly, image features, which were input quantities of the fuzzy recognizer, were extracted by means of image processing. Along with the formulation of inference rules based on experts’ experience and field observations, the fuzzy recognizer was designed. Finally, the fuzzy query rules table was built by defuzzification; accordingly the problem of traffic spillover recognition was solved. At last, a real-time detection system for traffic spillover phenomenon was designed and realized, the core theories of which were above-mentioned algorithms.In summary, a novel method by video detection technique, which was used to recognize traffic spillover phenomenon, was proposed in this paper. Meanwhile, an integrated system based on these algorithms was build. Experiments show that the algorithms were practical and effective, and the system had good running performance. Due to above studies, innovative ideas were proposed in the research field of traffic spillover. And the studies also laid the foundation for the follow-up researches.