Abstract: According to the principle that an incident on a freeway produces a great impact on the traffic flow,an automatic incident detection algorithm based on wavelet transform and least squares support vector machines(LS-SVM)is presented. The wavelet transform is fist used for extracting feature data,LS-SVM is then applied to incident classification. Simulation experiments show that the proposed algorithm has such advantages as short detection time, high detection rate and low false alarm rate.This method is a novel and practical way to realize freeway incident detection.