SVM-based detection of traffic incident.

Traffic incident are non-recurrent events that disrupt the normal flow of traffic and create a bottleneck in the road network, reliable automatic detection of traffic incidents is required for efficient traffic management. With the development of statistical learning theory, support vector machine (SVM) has been recently proposed as a new learning network for pattern recognition with good generalization performance. In this paper we proposed a SVM-based classifier which is used to detect incidents, and the discrete wavelet transform (DWT) feature extraction model is used to get features for classification. Using data from simulations, the experiments show that the automatic incident detection algorithm can work effectively and robustly.