Study On The Road Traffic Flow Reconstruction Based On Surveillance Video

Traffic flow estimation,simulation and visualization are the important contents in the field of traffic engineering research,it is the research hotspot in this field to study the traffic flow estimation method which is in line with the actual road traffic law and realize the realistic traffic flow simulation with virtual reality technology.Its research results has a wide application prospect in traffic management,traffic planning and urban intelligent transportation construction.At present,most of the traffic flow simulation is using simulated data combined with complex mathematical model to generate road traffic flow,but this method has high complexity and low efficiency,and the authenticity of the road traffic flow generated by simulation data is not strong.In addition,there is less work on the traffic flow simulation research combined with the actual road traffic information at present.Aiming at the above problems,this paper studies and proposes a road traffic flow reconstruction method based on road surveillance video,the method utilizes the discrete data extracted from the surveillance video and simulates the vehicle trajectory with the A~*algorithm to realize the road traffic flow reconstruction which uncovered by surveillance video.The proposed method combines the road surveillance video processing technology with the traffic flow simulation technology,not only can make better use of the road surveillance video resources,obtain the vehicle information from the video to provide the data basis for the intelligent traffic analysis,but also can take the road vehicle information as the input data of road traffic flow simulation algorithm,thereby improving the authenticity of virtual simulation of traffic scene.Specific research work includes:(1)In order to obtain the real information of road vehicle data,this paper realizes a vehicle trajectory acquisition method based on surveillance video.The method is used to track the vehicle in the video,and obtain the pixel coordinate sequence of the moving vehicle in the video.Because of the different shooting angle of different surveillance video,only the practical value of the pixel coordinates is not high,so the mapping relationship between the video image coordinate system and the real road coordinate system is established,and the vehicle position information of different video is correlated.On the basis of this,the calculation of the vehicle running speed is realized by measuring the actual distance and time difference corresponding to the pixel coordinate trajectory acquired when the vehicle moves.(2)In order to obtain the traffic flow of the road area which is not covered by the surveillance video,this paper proposes a vehicle routing algorithm based on the A~*algorithm to reconstruct the road traffic flow.First,the road environment information is simplified and building a static vehicle roadmap to simplify the vehicle search range to improve the calculation speed.Secondly,taking the number of the lane changed,the acceleration changed and the safety distance as the decision basis of the decision function of A~*algorithm,the heuristic search is made in the search space of the vehicle,so as to obtain a path that best meets the constraints.Finally,the order of the vehicle path generated is determined by the method of the priority value.The priority value is defined as the position of the vehicle on the static road map,that is,the car which its relative position in front can generation path with priority,and the subsequent vehicle is only required to consider the vehicle in front of the same lane and keep a safe distance to avoid collisions when it generates the path.(3)A set of traffic flow reconstruction system based on the surveillance video extraction vehicle trajectory acquisition discrete data is designed and implemented.The system integrates the vehicle detection and tracking with the traffic flow generation and display,it can track the vehicle in the video,obtain the vehicle trajectory and speed as the input data of the simulated traffic flow,and combine the traffic flow reconstruction algorithm to generate the vehicle path to simulation the road traffic flow,and ultimately show the simulation results in the three-dimensional rendering engine.Experimental results on the actual road video data show that the road surveillance video processing proposed in this paper and the vehicle speed calculation method based on the vehicle travel trajectory provide the data foundation for the analysis of vehicle driving behavior in the video.At the same time,the reconstruction of traffic flow based on real road video vehicle data increases the authenticity of traffic simulation,and provides a new application direction for the combination of surveillance video and traffic virtual reality.