Research On Urban Traffic Flow Statistics Algorithm Based On Video Image

The rapid economic development has accelerated the progress of urbanization in China.Traffic congestion caused by the increasement of motor vehicles in cities has become a common problem in major cities.The seriousness of urban traffic congestion has led people to shift the focus of solving traffic problems to the field of traffic management,thus the concept of ITS has been proposed.Traffic flow statistics is an important basis for ITS as well as traffic management.By monitoring the traffic flow,the real-time status of traffic operation is obtained,traffic guidance is made in time,and vehicle congestion is reduced,thereby achieving the purpose of intelligent transportation.This thesis aims to detect and track vehicle traffic based on the surveillance video information on the road,and calculate the traffic flow at the same time.The thesis compared the common traffic detection techniques,then proposes an improved method to reduce the"ghosting" phenomenon in Vibe used for the detection process.As the background model is initialized,14 neighborhoods of the pixels are selected to run Vibe in combination with the inter-frame difference method,which erases the "ghosting" phenomenon of Vibe in some extent.After detecting the moving vehicle target,the HSV-based shadow removing algorithm is used to eliminate shadows in the image,filter out noise and extract the contour of the moving vehicle target.The moving vehicle target is tracked based on the moving target feature and the update of motion history image.The vehicle contour tracking is combined with the virtual line detection to divide the detection area for each lane,so that the traffic flow in each lane can be counted.Traff-ic videos photographed in sunny,rainy and night environments were tested respectively to verify the effect of algorithms proposed by this thesis.The result show that the final average traffic flow accuracy rate can reach 90%,which can meet the daily needs.