Vehicle Detection Algorithm Based On Background Modeling

As the world’s urbanization has developed to an accelerated phase,it is certain that vehicle inventory will increase year by year, safety problems and economic losses caused by traffic congestion has become a major preoccupation of urban transport planning and management department.however,it has become more difficult to manage the current large trasportation network just rely on manpower. in the mean time,there is a large amount of development work being undertaken worldwide on Intelligent Transport Systems,which could make the full use of existing road infrastructure resources,achiving an intelligent programme between vehicle, pedestrian and road.As such,The vehicle detection technology is the important contributor to obtain road traffic information in Intelligent Transportation SystemThis paper studies the moving vehicle dectation through the road surveillance vedio, setting up the background model, whereby the target’s yanmo could be obtained among shipinzhen. In addition,the establishment of discriminant model toward the target vehicle could help identify and capture the targets as much as possible. The main contents including the following aspects:(1)A background modeling algorithm based on Splitting Gaussian Mixture Model is presented.By creating a multiple layers of extracting movint target information。establishing Splitting Gaussian Mixture Model seperatly on multiband, for which can provide the correct update feedbacks of the model group,enhacing the sharpness of the moving targets. Calculation result indicates that the algorithm can eliminate significant noise disturbances and solve the empty target.(2)Improved Relative Discriminative Histogram of Oriented Gradient(IRDHOG) is put forward.the approach taken by the IRDHOG is to analysis connected information between central block and surrounding block, integrating those information(RD-bins)into features. As with counting distribution of bins toward interest objectives(vehicle) and non-interest objectives the significant and non-significant sections can be calculated. On the premise of that the target vehicles characteristics can be highlighted, so as to improve the detection efficiency of the targets.