An Algorithm For Extraction Of ROF In Surveillance Scene

Video of surveillance scene contains a lot of information. But the focus regions arethe ones with high value. The extraction technology of the Region of Interest (ROI)commonly used in image processing extracts the important regions by imitates the wayof subjective visual attention when human watch the images. So the extractiontechnology of ROI is subjective. It cannot extract the whole important imformationfrom the video of surveillance scene.According to the application of intelligent surveillance of scenes, the concept ofthe Region of Focus (ROF) in surveillance scene has been researched in this paper. Andthe model of ROF extraction has been proposed. The ROF is different from the ROI.The foundation of the extraction of ROF is that whether the regions benefit the videosurveillance and the video analysis.According to the problem of the segmentation of the initializing ROF in the modelof ROF extraction, an algorithm for initializing ROF segmentation based on fusingmultiple features is proposed. This paper analyzes the charaters of ROF. The motiondetection, the image local entropy detection and the color features are used for thesegmentation. The initializing ROF can be segmented by fusing the multiple features.According to the problem of the final ROF extraction in the model of ROFextraction, an algorithm for ROF extraction based on the multi-tuple is proposed. Thesenior characters of the initializing ROF should be extracted. Then the bottomcharacters and the senior characters can be used for the recognition of types of theregions. Construct the multi-tuple by the types and the characters of the regions. Thedetermination and the extraction of the ROF are based on the multi-tuple.According to the problem of classifying the ROF the in the model of ROFextraction, a method for ROF classifying based on the scene type is proposed. In thispaper, the main types of surveillance scenes and ROF have been analyzed. A model ofclassifying importance has been proposed. The model classifies the importance of theROF by the type of scene and the ROF and rank the ROF by the importance.