Study On The Methods Of Automatic Incident Detection For Urban Road

Along with the fast development of the social economy and the unceasing expansion of urban road network, the traffic incident which happens on the urban road has aroused more and more people's interest. It not only threatens seriously the people's life and the property security, but also gives severe influence on the transportation. At the same time, the early artificial methods which are used to discover the traffic incident already could not maintain the need for traffic management. In order to enhance the security and the operating efficiency for the urban transportation, each country starts to study the AIDFU technology which can be carried on to monitor traffic condition in the urban road.The dissertation stems from subproject"the study on traffic congestion prediction and prevention theory and method"of the item"the study on key theories and simulation technologies of dynamic traffic management and control system in urban road network", funded by National Science Foundation Item under grant 50338030. Through analyzing and summarizing the characteristic of traffic flow and the classification of traffic incident in urban road, the dissertation proposes the automatic incident detection algorithms based on macroscopical traffic flow parameter and facility on vehicles. The concrete content is as follows:(1) The classification for urban road traffic flow. This part analyzes the traffic flow characteristic of the urban road, proposes the thought which divides urban road traffic flow into the heavy flow and the light flow, providing basic theory to Incident Detection of the different situation in the following parts.(2) The foundation of AID technology. This part introduces the classified method to the traffic incident Detection, summarizes separately the present situation and the history in China and foreign countries according to the classification of urban road traffic flow, which provides the spot for this article. Also this part introduces the index and method evaluating the performance of AID algorithm.(3) The research of AID algorithm based on macroscopical traffic flow parameter. Through analyzing and summarizing traditional AID algorithm, this part designs separately single section and the double section AID algorithm based on the Support Vector Machine according to the distributing characteristic of detectors in urban road control system. Through validating to new algorithm and contrasting with traditional algorithm based on simulated data, the result shows the excellent performance of the new algorithm. Finally, this part produces the detectors'distributing principle of traffic Detectors in urban road from AID angle.(4) The AID method based facility on the single vehicle. In view of the situation in the light flow, this part brings forward a new Detection thought which considers uniformly the driving difference and the non-driving difference. The driving trait refers mainly to the change of vehicles'acceleration. The non-driving trait refers mainly to the noise and vibration information of vehicles. In virtue of the principle of breakdown diagnosis and alarm on the vehicle, the part designs the automatic incident detection system based facility on the single vehicle. Through analysis to driving characteristic of the vehicles under different condition, this part proposes the automatic incident detection system's logical frame, and elaborates the concrete realization technology and the feasibility analysis.At last, concludes the whole dissertation, analyses the findings and achievements, and brings forward further research directions and projects.