Research On Intelligent Vehicle Traffic Detection System Based On Video

The Intelligent transportation system is future development direction of the transportation industry.Traffic flow is an important part of ITS.Through the application of digital image processing technology to the video can directly achieve the vehicle counting.The algorithm can avoid the inefficiency of manual counting and damage to pavement by the ground coil.Based on the existing algorithms,this article has done research on detecting,tracking and counting vehicle.In detecting target's section,a target detection algorithm is proposed based on the combination of ViBe method and frame difference method.Using the ViBe algorithm to obtain foreground images,then using the frame difference method for adjacent foreground images to obtain vehicle target detection results.At the same time,this system records the number of background model updates per frame to prevent changes in light causing accuracy reduced.In terms of the adhesion target,if using morphological processing,the shape of the vehicle will be seriously deformed.Concerning the issue,we first use the image thinning algorithm to get the skeleton of the adhesion target,then use Shi-Tomasi algorithm to get all the corner points,at last use the DBSCAN clustering algorithm to classify the corner points.The clustering algorithm doesn't need to input the number of categories,the adhesion target can be divided regardless of the number of vehicle adhesion.Through the experimental testing,the extracted vehicle has a clear appearance,“ghosting” phenomenon can be quickly eliminated and the adhesion can be divided into single targets.To reducing the search range of the matching target,Kalman filtering algorithm is used to predict the target position and then the true position is obtained by using the similarity of the histogram in vehicle tracking's section.At the same time,a list is created for each target and the latest feature information of the target is added to the end of the table.The system can get vehicles track.In terms of vehicle statistics,two virtual detection lines are set in the video screen.One of lines is the starting counting line,other line is the stopping counting line.By comparing the relative positions of the target position and the virtual detection line,the statistics of the traffic flow are completed and the current road traffic is judged.