Abstract: A new multi-features selection method on traffic videos which is based on genetic algorithm is proposed to improve the accuracy and rate of the traffic video event detection.This method firstly extracted multiple features of traffic videos to gain information of a variety of traffic events as much as possible,and then fused to those features to get a high-dimensional redundancy feature vector that can characterize the video,then used genetic algorithm to optimize and select multiple feathers and finally obtain the optimal feature set by SVM classifier to detect and analysis traffic events.Experimental results show that this method can effectively reduce the dimension of features and improve the accuracy and rate of the traffic events detection.