Research On Vehicle Detection And Tracking In Highway Video Surveillance Based On Lane Line

In the western region,most of the highways have the characteristics of far-reaching views and a straight line.Due to the spatial perspective projection,the area characteristics of moving vehicles vary greatly within the effective monitoring range,and the overall phenomenon is "far small and nearly large".It is difficult to find a suitable threshold for the entire image,so that it can filter out the larger noise in the near field of view and preserve the smaller vehicle target in the far field of view.Therefore,the vehicle detection and tracking method in highway video surveillance based on lane line is proposed.In this way,we can use lane line to describe this spatial relationship and introduce influence factors to improve the feature representation of the moving target.The background image is established by using the mixed Gaussian model before the target detection,and the established background model can be dynamically updated,and then the foreground target is obtained by the background difference method,and the shadow elimination and morphological processing are performed.The lane line position and color feature extraction,edge detection,Hough line transformation and K-means clustering are successively calculated in the background image to obtain the line information of the three lane lines and the intersection coordinates of the lane line.It will lay the foundation to calculate the influence factor of the subsequent lane line.For the detected vehicle target,its position in the monitoring scene is determined by its angle with the reference lane line and the distance from the intersection point.The calculation of the influence factor of different positions is the area characteristic of the actual acquisition target of the lane division in the entire monitoring field of view,and then the area of the maximum position(near field of view)is selected as the standard area,and the influence factor at other positions is calculated by calculating the ratio of the standard area to the area at the position;so that the calculation of the influence factor of the entire effective monitoring area can be completed.Therefore,the target detection result in this paper includes not only the direct area feature value of the target in the video image,but also the product value calculated by the influence factor and the direct area feature value.So that the target area feature performance at different positions can be adjusted,and the rationality of the target is measured.In this paper,the target detection method can remove the pseudo target while retaining the vehicle target in the far field of view,then,to filter out the background noise of the abnormal position.Meanwhile the motion lane or direction of the vehicle target can be obtained,so that the target detection result will be more accurate.Based on the completion of target detection,the feature-based target tracking method is selected for vehicle tracking research.The vehicle's centroid position and area features and lane information are selected as the matching features of target tracking in two adjacent frames.Therefore,in the target matching,the similarity comparison is performed only for the vehicles in the same moving direction,so that the unnecessary target matching calculation is reduced,thereby improving the calculation efficiency.