Research On Traffic Abnormal Incident Detection Algorithm Of Highway

In recent years,with the continuous improvement of computer storage and computing capabilities,the rapid development of artificial intelligence and pattern recognition technology,video traffic incident detection technology has become a hot issue in the field of intelligent transportation.Traffic incident automatic detection system is the key of intelligent and automation of traffic video surveillance system,it provides a conditions to deal with the traffic accident and reduce the traffic delay and avoid the secondary traffic accident,and provides a new breakthrough for the highway operation and management.But how to efficiently,accurately and quickly realize the automatic detection of traffic incidents is still a difficult problem in the field of intelligent transportation.This article embarks from the practical application,taking the expressway video sequence as the research object,from several key technologies of moving target detection,tracking and abnormal behavior description set about to study,designing the automatic detection algorithm for abnormal traffic incident of retrograde,parking and lane changing of highway.In this paper,study on the above three kinds of abnormal events include the following aspects:In the aspect of moving object detection,the background model is set up by means of mean value and the background difference method is used to extract the foreground of moving object;aiming at the foreground of moving object with shadow,bringing a method based on edge and HSV color is used to remove the shadow,and combined with morphological processing method to extract the whole foreground of moving object,which provides a basis for effective tracking of moving targets.In the aspect of moving object tracking,taking centroid and area of the moving vehicle as the basic characteristics of vehicle tracking,combined with Kalman filtering algorithm to seek optimal estimation of moving target features,Euclidean distance is used to calculate the distance and size difference of moving target to find the best match to complete the moving target tracking;aiming at the phenomenon which vehicle occlusion can make tracking target loss,a method of area selection is proposed to track vehicles in different ways,get the vehicle trajectory,which provides a gist for the judgment of abnormal events.In the aspect of abnormal event detection,with analysising the trajectory of vehicles can intuitively understand the motion direction of the vehicle,taking it compared with the positive direction of the road can judge whether the vehicle retrograde;with analysising the trajectory of vehicles can indirectly obtain vehicle instantaneous speed,acceleration,centroid position change traffic parameters,analyzing the change of these traffic parameters can judge whether the vehicle parking;with analysising the dispersion degree of the distance between the vehicletrajectory and the reference lane line can judge whether the vehicle lane changing.In this paper,the actual traffic video sequences of different sections of highway were tested,the experimental results show that the anomaly detection algorithm is effective and can accurately detect abnormal events of retrograde,parking and lane changing,which has a good practicality.