Research And Implementation Of Embedded Wireless Network Video Surveillance System Based On JPEG2000

A new solution of video surveillance is designed on basis of analysis on principle of Adaboost human face detection algorithm and JPEG2000 compression algorithm. In the scheme, Adaboost and ROI algorithm technology are combined to avoid disadvantages of large data size and low bandwidth. The S3C2440 microprocessor with an ARM9 core is chosen as hardware carrier of video capture and transmission. Face detection algorithm is migrated on the platform to realize the corresponding function. ROI image compression algorithm is realized on TMS320DM642,a kind of DSP chips made by TI. In this way, human face area has high definition and the video image has smaller size.Implementation of human face detection is based on Adaboost. By training of face database, many cascade classifiers are constructed to detect faces. The method has higher probability of detection and faster detection speed. New forms of rectangle feature are added for capturing linear gray changes of the old forms. By experimental comparison and analysis, it has higher detecting probability and lower false alarm rate.In order to get a clearer face area and reduce the size of image,JPEG2000 is used as video image compression algorithm. Maxshift method is used to realize the function of ROI encoding. Multi-ROI regions with different shapes can be encoded and the shape of ROI need not be encoded. Compare with scaling-based shift method, maxshift has smaller computation complexity.