Expressway Throwing Objects Detection Based On Video Stream

The expressway has the characteristics of many lanes,wide roads and large traffic capacity.The transportation volume of the expressway is much higher than that of ordinary highways,which is of great significance to the development of the national economy.In the event of a traffic accident on the expressway,it will cause heavy casualties and property damage.Objects are droped by vehicles on the expressway during driving,and the rear vehicles are prone to traffic accidents with serious consequences.The existing expressway safety detection system installs video cameras to collect video images of the road in key parts along the expressway,and the monitoring coverage is not comprehensive.In order to solve this problem,an expressway throwing objects detection technology based on driving vehicles is proposed in this paper.The detection of expressway throwing objects is a mobile monitoring network composed of vehicles on the road.It records all kinds of traffic incidents and detects them in real time.Once there are abnormal events such as throwing objects,an alarm message is sent to the control center.After the analysis and confirmation of the control center,it notifies the driving vehicles of the relevant sections and handles the accidents in a timely manner.A video acquisition module,a video detection and processing module,an image transmission module,and a control center are included in the expressway throwing object detection system.The moving target detection and tracking algorithm in the throwing object detection system is focused in this paper.The phase correlation method is used to estimate and compensate the pseudo-motion.An acting background model is established by using the common information of the historical frames of the current frame.The actual moving objects are extracted from the foreground effectively by using the dissimilar historical frames of the current frame.The two-step morphological operation is used to refine the target area and obtain the accurate contour and position of the moving target.Kalman filter is used to track moving targets,and the centroid,size and intensity distribution of moving targets are used to effectively solve the problem of data association in the tracking process.Based on HOG features,SVM classifier is used to classify and identify the detected moving targets,and the non-vehicle targets(throwing objects)are detected and tracked.The experimental results show that the proposed method can detect and track moving targets accurately,and can distinguish the attributes of moving objects effectively,so as to realize the detection of throwing objects.