Research On The Key Algorithms Of Intelligent Video Surveillance System In The Vehicles Crossing

As a new research direction in computer vision technology field, intelligent video surveillance system is the combination of computer science, machine vision, image processing and other subjects. The application of intelligent video surveillance system in the streets can be realized by monitoring the real situations through camera, detecting and tracing the moving cars in the traffic, and handling traffic violation behaviors, like jaywalking and speeding.There are already some researches on intelligent video surveillance system in the traffic both in China and abroad, but most of them based on ground induction coil video surveillance system. Since ground induction coil has many disadvantages like difficulty of laying, damaging the roads and inflexibility of monitoring, this research will focus on key algorithms in intelligent video surveillance system in the Vehicles Crossing based on these achievements. The main contents are as follows.Firstly, overall plans of key algorithms of intelligent video surveillance system in the traffic are the main focus of this research. These include the setting of camera and virtual coil and the specific process of shooting the pictures in jaywalking and speeding.Secondly, the reconstruction and updating the algorithms of background in the real situations are being studied. After analyzing and comparing several algorithms of reconstruction and updating of background, one algorithm on background updating algorithm based on running mean is summarized, which can reflect the real situations in the traffic exactly and authentically.Thirdly, the detecting of moving cars is studied in the real traffic situations. The advantages and disadvantages of several algorithms of moving objective detections are compared. Considering the authenticity of this system, background updating algorithm is adopted in this research to judge if there will be the moving cars. Finally, main focus is put on introduction of tracing algorithm with feature based suitable to this intelligent video surveillance system. By taking advantage of background updating algorithm based on running mean, the reconstruction and updating algorithm of background for continuous pictures can be realized in the designated area. After detecting moving cars, the mer-minimum enclosing rectangle of every moving car will be counted and then can successfully trace the cars. Coordinating transformation can be also realized between actual coordinate system and image coordinate system and matrixing the real speed in the designated areas. If the speed exceeds the top of limited speed in the real traffic, the system can shoot the pictures of the cars and then seed the related information like pictures to backstage center.