The Research And Design Of Embedded Network Video Surveillance System Based On ARM9

For the purpose of building a cheap, practical and low power consumption embedded network video surveillance system, according to the actual needs of video surveillance systems, the S3C2440 microprocessor with an ARM9 core is chosen as the core of hardware platform, the embedded Linux operating system is chosen as the core of software platform, the low-cost and commonly used USB camera is chosen as the video capture device, and the design of video surveillance system is completed. The whole system is divided into two parts: the server at monitoring place and the remote client. The server uses Video4Linux technology and multi threads technology to capture images, uses motion object detection technology to detect anomalies, uses Xvid Encoder to compress the video as MPEG-4 format, and uses RTP/RTCP protocol to transmit video flow through network. The client mainly realizes the functions such as video receiving, decoding, display and storage.According to the characteristic of limited processing speed in embedded systems, the time-consuming motion estimation part in video coding is lucubrated, and a hexagon fast motion estimation algorithm based on motion vector sets prediction is presented. The algorithm combines technologies such as multi vector sets prediction method, effective search strategies and SAD adaptive partial sampling method. By using this algorithm, the image quality could be assurance, and at the same time, the computation of motion estimation is significantly reduced.According to the situation that both camera and monitoring background are relatively still, a motion object fast detect algorithm based on YUV color space is presented. The algorithm uses the idea of background difference method, uses micro block as the basic unit, uses both brightness and color information in YUV color space to detect, updates the background according to the test result, and then by using threshold and morphological filter processing to get the final motion object zone. The algorithm performs quickly and accurately, which could meet the requirement of detection result and speed in embedded video surveillance systems.