Research On Key Algorithm Of Embedded Front-end Pre-processing Equipment In Visual Surveillance System

In today's awareness of prevention and for increasing public safety situation, the visual surveillance system has infiltrated our daily lives, hotels, school, station, shopping malls, residential, we can see the monitoring cameras everywhere; In recent years, as the rapid development of computer, image processing and network transmission technology, visual surveillance system is no longer a simple devices of transfer and store the images. More and more function of intelligent monitoring been moved to the Embedded front-end equipment from the background equipment. But, the contradiction between the complex algorithms of intelligent monitoring and embedded device with limited resource already become a main problem which needs to solve. These conditions make the application of intelligent monitoring algorithm in the embedded devices to become a research focus in visual surveillance systems.First, the thesis analysis the development of visual surveillance systems, discuss the type and structures of video systems. Focus on the visual surveillance systems, the thesis analysis the definition, function and application status of embedded front-end equipment. Throng a detailed analysis about embedded front-end equipment software and hardware resource, we find the problem of achieved the intelligent monitoring in embedded front-end equipment. Through analysis about algorithms of image pre-processing and theses implementation method, we get the experiment and guide of the image preprocessing algorithm and its implementation. Based on those analyses, we get the objectives of algorithm proposed: assurance the detection rate, as much as possible to reduce the complexity of the algorithm to reduce computation times and storage area occupied.Then, the thesis also focused on moving object detection algorithm. After conclusion of other algorithms, we selected a background updating algorithm based on adaptive median filter algorithm, and proposed improvement. Based on this background updating algorithm for moving object detection, the thesis is given in detail each step of the methods and data, with a large of experimental results and picture, to supplement shows that the algorithm results.Secondly, the thesis also focuses on moving object tracking algorithm, based on the overlap area of a single object tracking algorithm, combined with histogram-based color matching algorithm to solve multi-objective tracking problem in scenario of multi-object tracking and occlusion handling. Articles in this section are presented in each step detail of the method and experimental data, the simulation results.Finally, the thesis discuss the whole algorithm time complexity and space complexity, and give an instructive program with guidance which estimate performance on embedded device, and two examples.