Study On The Method Of Automatic Incident Detection For Freeway Based On FA-SVM

As the main carrier of the transportation, the rapidly developed freeway has brought various economic and social benefits. But it is also confronted with a lot of traffic problems, such as the frequent occurrence of traffic congestion and traffic accidents, which affects not only the efficiency of freeway, but also people's lives and their property.With the variety of the traffic problems, how to improve the operation state of the freeway has become the study core of the freeway traffic monitoring system. One of the ways is to conduct the deep study of the freeway traffic incident detection technology.As a new information processing technology, artificial intelligence technology has shown its significant advantages in the study of traffic incident detection, which guides the direction of the AID algorithm development. However, the selection of the input variables in the existing AID algorithms is not standard. So the combination with the emerging information processing technologies and the design of a comprehensive, effective traffic incident detection input variables could improve the detection results and the existing freeway running. The research mainly reflects in the following aspects.1) Analysis the freeway traffic flow characteristics under the changing traffic incidentAnalysis the different traffic parameters trends under the events, describes traffic flow parameters and their combinations with different change characteristics, preliminary designs the traffic incidents characteristic variables.2) Study the traffic incident characteristics variable filter based on FAAnalysis the advantage and disadvantage in the data processing with the FA. Designs screening methods of input variables in freeway traffic incident detection based FA.3) Study on the method of automatic incident detection for freeway based on SVMThis paper has analyzed the advantages and disadvantages of data classification with the support vector machine as well as the selection of different kernel functions of the SVM; Combination of vehicle detector data, designs the methods of automatic incident detection for freeway based on SVM. The verification result with the combination of the simulation vehicle detector data shows that detection effect of the AID algorithm designed by this paper takes obvious advantages compared with the contrast algorithm testing.Based on the study of freeway vehicle detector data and the in-depth analysis of expressway traffic flow parameters changing characteristics, this paper has made a deep research on the freeway traffic incident detection algorithm combined with the good ability of "dimension reduction" FA method and excellent classification theory of SVM. Compared with the contrast algorithm, the result indicates that the AID algorithm has the better detection performance.