Model And Analysis Of Automatic Incident Detection Based On Support Vector Machine

In recent years, china's roadway in particular of the highways had a rapid development. This brought tremendous changes to the people, but at the same time it brought a lot of traffic jams and accident. Through research the traffic incident detection, we can find the incident as soon as possible, and reduce the losses to the minimum.First, we give an introduction of the research status of the traffic incident detection. Then give the advantages and disadvantages of the research. After this, we give a description of the traffic detector. We give the traffic incident an definition. Then give an analysis of the principle of the traffic incident detection, single-cross and double-cross section detection. Then give the analysis of the traffic incident detection methods, and of the support vector machines, and the evaluation indicators so on. We give an improvement of the evaluation indicators.Give an analysis of the factors on traffic flow. Then get the incident-related factors. This factor will be the model input factor. This paper provides an orthogonal experiment method to select the parameter of SVM. Then we do experiment based on the I-880 database, and get a satisfactory test result. Finally we analyze the application on setting of the traffic detector and the rebuild of the road and so on.Through the experiments, the traffic flow and the traffic incident have a very close relationship. The paper verifies the SVM can give a satisfied result of the diction. And we give an improve of the evaluation index, and this index is more convenient and effective.