Research On Statistical Method Of Vehicle Traffic Based On Video

Image recognition can be applied in many applications of intelligent transportation systems.Through automatic traffic flow counting,the traffic information of a given area can be effectively displayed.After the surveillance video is processed by the existing image recognition model,the coordinates of the object in each frame can be easily extracted.Then,the extracted object coordinates are filtered to obtain the required vehicle coordinates.In order to implement the function of vehicle counting,it is necessary to identify the relationship of vehicles in different frames,that is,whether they represent the same vehicle.Although vehicle counting can be achieved through the use of tracking algorithms,the short recognition time can lead to incorrect tracking,which can lead to incorrect traffic counts.In this paper,we present a study of video traffic statistics based on video.This method is mainly divided into two major modules:the vehicle traffic statistics area division module and the vehicle traffic statistics module.The traffic flow statistical area division module includes based on video image data,first determine our ROI area through target detection,and then identify and divide each lane on the road into independent statistical areas through Hough transform in advance,and use this to detect vehicles;The traffic flow statistics module includes based on video image data,and uses deep learning-based target detection and recognition models to detect vehicles for each statistical area.Batch statistics are performed at time points to obtain vehicle statistics for the corresponding statistical area.Vehicle statistics are superimposed to achieve traffic statistics.The proposed method only needs to use the moving distance calculation of tracking a point to achieve the purpose of vehicle counting.In addition,by sifting edge vehicles to obtain tracking points,the system can mitigate the consequences of erroneous detection.The results show that this method can achieve high counting accuracy when the ambient light is sufficient.