A Video Surveillance System Based On Embedded Linux And H.264

With the improvement of living life and enhancement of security awareness, people pay more and more attention to their own life and property. Video surveillance, one of the most important aspect of security field, can supply the real reflect of the spotted region, and has become a hot research point.With the rapid growth of computer science and technology, network technology, video encoding and decoding technology, image processing technology and embedded technology, the video surveillance system develops toward network, more compact, more intelligent, more embedded, longer-distance fashion.An embedded network video surveillance system has been designed and implemented in the paper. The stream media server has been built on the development board, using the application interface supplied by linux kernel to capture video. The captured video image is transmitted by the network, after encoding, encapsulation and pack according to specific protocols. The client gets data packet from network, and decodes the encoded data in a reverse way to acquire the raw data for display, thus accomplishing the aim of real-time video surveillance. The system here has many merits, like lower power consumption, easier to implement, smaller volume, and lower price and so on, full of theoretical and practical significance.The dissertation emphasizes on UMHexagonS motion estimation algorithm, describes the search steps of UMHexagonS algorithm in depth:â‘ the starting search point prediction;â‘¡the unsymmetrical-cross search;â‘¢the multi-hexagon-grid search;â‘£the extended-hexagon search. A hierarchical search strategy has been adopted in UMHexagonS algorithm, the current optimum point of last step becoming the center of next search step. The UMHexagonS makes use of many search templates, greatly improves the search speed, and the final match precise is close to full search. In order to modify its shortcomings, three new strategies, including self-adaptive selection of search layer, dynamic search windows size, and search points grouping, are proposed to improve its performance. Comparing to the original algorithm, the experimental result indicates that the improved algorithm is able to save motion estimation time, decrease the algorithm time complexity and increase the real-time property of encoding process.Moreover, the paper has described each functional module of the entire system, including the video capturing module, the video encoding and decoding module, the video transmission and receipt module, the video display and storage module. After analyzing the system modules, the paper tells the basic principles of each module, and summarizes and concludes used technologies. Finally, the paper has completed each module and accomplished the entire video surveillance system.The final result indicates that the surveillance system is able to acquire stable and clear image, acknowledging the aim of real-time monitoring. It can be extended to perform post-processing on captured image for intelligent field.