Research On Automatic Detection Algorithm For Urban Road Traffic Congestion

Real-time information on the status of urban road traffic congestion helps the traffic management department to take corresponding measures in a timely manner,which plays an important role in alleviating road pressure and improving the capacity of urban roads.It is of great significance to make full use of the network resources of traffic surveillance cameras to develop road congestion detection algorithms based on visual image information.However,the traffic conditions of urban roads are affected by many uncertain factors,such as temporary traffic control,traffic light failure,bad weather,and traffic accidents,the occurrence of road congestion is very sudden and random.The method of traditional computer vision is difficult to process effectively.In response to this problem,this paper uses a reliable vehicle detection algorithm based on deep convolutional neural networks as a breakthrough,and conducts research from four aspects:road detection,road maximum carrying capacity estimation,two-way multi-lane vehicle counting,and congestion detection.The specific content is as follows:(1)Aiming at the problem that the detection of urban road congestion will be interfered by vehicles in other non-detection areas,an effective road surface area estimation algorithm based on offline video data and deep learning object detection algorithm is proposed.Use object detection algorithm and tracking algorithm to obtain the position and direction information of road vehicles,and count the number of vehicles passing through each point of the road,and then reserve the road area of interest by setting appropriate threshold,and finally realize the extraction of effective road information.(2)Aiming at using the instantaneous number of vehicles to estimate the maximum load on the road will be affected by the precision of the object detection algorithm itself,a road maximum load estimation algorithm based on offline video of the road combined with object detection and non-maximum suppression algorithm is proposed.On the basis of the road detection algorithm,by saving the position and size information of the road vehicles detected in a period of time,and using the non-maximum suppression algorithm to filter the redundant vehicle information,the maximum load capacity of the road is effectively estimated.(3)In order to solve the problem of two-way multi-lane counting on urban roads,a vehicle counting algorithm based on the combination of deep learning object detection algorithm and virtual coil is proposed.The virtual coil on the road is divided into several independent virtual sub-coils according to the actual number of lanes,and the two-way multi-lane is realized by judging the position relationship between the center of the vehicle detection box and the virtual sub-coil and the overlapping area of the vehicle detection box and the virtual sub-coil count.(4)Aiming at the fact that the existing traffic congestion detection algorithm cannot accurately obtain the situation of urban road congestion,a method of detecting congestion on different directions of the road is proposed.On the basis of considering the influence of traffic lights,the maximum carrying capacity and the maximum traffic flow of the road are used to detect traffic congestion on the roads in different directions,and the degree of congestion of the road is concretely quantified and graded.In addition,a video-based road traffic light state detection algorithm is also implemented.Experiments show that the algorithm designed in this paper can effectively extract the road area,estimate the maximum number of vehicles on the road,and can count the number of vehicles on the road and judge the state of traffic lights,and realize the effective estimation of urban road congestion.It has good applications reference value.