Research On Abnormal Event Detection Algorithm Of Urban Traffic Video In Sparse Learning

With the rapid development of monitoring instruments in public areas,there is an urgent need for automatic and non-monitoring anomaly detection system,especially traffic monitoring.Efficient video anomaly detection system has become one of the most important video anomaly detection systems.However,the variability of scenes and the complex state of events in traffic surveillance video make it difficult to detect abnormal events.Influenced by the rapid development of intelligent transportation system,the research of traffic anomaly detection using computer vision technology has attracted the attention of relevant personnel and made some breakthroughs.However,due to the variability of the traffic environment,there are still some bottlenecks in the practical application of some methods proposed,and there is still a lot of room for development for the research of traffic anomaly detection system.Based on the research results at home and abroad,this paper mainly studies the urban traffic abnormal event detection algorithm based on video image.This paper studies a real-time and robust traffic anomaly detection algorithm in urban traffic video streams.In order to improve the performance of edge detection,an image edge detection algorithm based on fusion of multi-scale morphology and wavelet transform is proposed.In addition,a target detection method based on HSV and edge gradient information is proposed to solve the problem of poor robustness in complex traffic environment,which improves the traditional detection method and the problems of duplication and void.A new self-adaptive sparse reconstruction method for vehicle behavior learning based on video surveillance system is proposed.Half algorithm is used to solve the sparse solution by introducing₁l_/2 regularization.Finally,based on sparse solution and trajectory similarity,a trajectory classifier based on sparse reconstruction and similarity is designed to learn vehicle behavior.Based on the existing study and Research on automatic learning of vehicle behavior from video,the research contents of this paper are mainly divided into three parts:moving vehicle detection,vehicle behavior trajectory learning and vehicle behavior trajectory recognition.?1?Motion vehicle detection.Through video image preprocessing,the foreground image of moving vehicle is obtained by using improved VIBE method and three frame difference method.At the same time,the shadows generated by moving vehicle are removed according to the color information and edge gradient information of vehicle image.?2?Vehicle behavior trajectory learning:A regularizedl_p?0<p<1?sparse reconstruction model for vehicle behavior learning is designed,and the lower bound theory of l_p?0<p<1?norm is applied to obtain more sparse reconstruction coefficient vectors,which guarantees the learning performance of vehicle behavior trajectory.?3?Vehicle Trajectory Recognition:A trajectory classifier based on sparse reconstruction and similarity is proposed for vehicle trajectory recognition.The proposed method has high accuracy for classification and recognition of vehicle behavior.