Research On Automatic Detection Technology Of Urban Road Traffic Events Based On Floating Car Data

After traffic incidents such as traffic accidents,vehicle breakdowns,and cargo spills occur in urban road networks,the bottleneck of road capacity will be formed,resulting in the phenomenon of vehicle congestion and queuing,forming occasional traffic congestion.If the traffic incidents can not be dealt with in timely and effectively,the traffic congestion may spread and lead to a large area of congestion within the network,which may lead to traffic paralysis in severe cases.Rapid and effective detection and facilitation of urban road traffic incidents is an effective way to alleviate traffic congestion and reduce losses,and is also an important part of urban road traffic management and control.In this paper,an algorithm of urban road traffic incident detection based on floating car data is proposed,and the corresponding traffic incident detection system is designed and developed.To begin with,this paper summarizes and analyzes different types of event detection algorithms,and compares their advantages and disadvantages.Secondly,it analyzes the characteristics of floating car data,cleans and repairs the data,and studies the matching algorithm of floating car GPS data on the road network,and realizes the accurate positioning and trace drawing of floating car data on the road network from the time,direction,position,etc.Then,through analyzing the changing characteristics of traffic flow parameters when urban road traffic events occur,an automatic traffic event detection algorithm based on convolution neural network is proposed,and multi-parameters are used to distinguish event categories,and the algorithm is tested and verified using Wuhan floating vehicle data and traffic accident data.At the same time,an experimental environment was built by using VISSIM simulation software to verify the algorithm;Finally,based on the above research results,the functional design and interface design of the traffic incident detection system based on convolutional neural network are carried out..Through comparing the detection rate,false alarm rate and other detection indicators with the traditional neural network detection algorithm,the algorithm proposed in this paper has higher accuracy,and can realize the identification of event types.