Study On Pre-processing Algorithm In The Front-end Of Video Compression And Design System For Smart Surveillance

With security awareness, video surveillance requires more and more features. Traditional surveillance systems on the passive monitoring mode can't automatically detect target, identify, track and alarm, etc. However, the intelligent control is to overcome the shortcomings of traditional monitoring methods relying solely on management control, and gradually become the trend of development in the monitoring field. High-quality video image as the source of target in surveillance system is the base of intelligent control. The result of collecting de-interlaced video by traditional cameras, not only makes image quality serious, but also does not benefit for the rate control of H.264 encoder. And once applied to multi-channel video monitoring system, in the premise of the fact that H.264 coding efficiency closes to the limit, it makes the transfer network bandwidth occupied by the increasing amounts of data scanty. Certainly, this will greatly limit the application of intelligent monitoring, thus, we should find other means to reduce the mounts of data. In the context of this technology factors, the subject uses de-interlacing and down-sampling algorithms to preprocess video sources in the front of the monitoring system, aiming at achieving the elimination of interlace effect, reducing the amounts of data and improving coding efficiency .So that the system may have higher quality video signals under the same bandwidth for more efficient intelligent video analysis to benefit better monitor results.On the basis of analysis to the intelligent video surveillance system, the paper comes up with the overall design scheme of the front-end video pretreatment system. Through researching on the core preprocessing algorithm, combined with the characteristics of the hardware FPGA logic implementation, the paper uses an improved motion adaptive de-interlacing and down-sampling algorithms based on the haar wavelet transform for video image's processing. The processed video data will connect with video surveillance system seamlessly to interact and share data through the hardware platform's PCIE interface. In the program, motion-adaptive de-interlacing algorithm used in motion detection followed by a morphological opening operation, this step not only improves the accuracy of de-interlacing, but also maintains the clarity of the static region. The down-sampling algorithm takes the principle of wavelet transform, enabling the concentration of the original image information to the low-frequency sub-band image effectively. It only lost small amounts of edge information, retaining most energy of the image and achieving fluent effects in vision. The FPGA hardware design based on the core algorithms that described above, implements through the addition, subtraction and shift operations. The design adopts pipeline structure, not only greatly improves the processing speed of the system, but saves hardware logic and system cost.The front-end video pretreatment system designed by this subject proves to have normal operation, attaining the real-time processing effect of video de-interlacing and down-sampling in a smart surveillance system, and finding a better balance point between working efficiency and hardware cost, possessing higher practical value.