Moving Object Segmentation And Object-based Mpeg-4 Encoding

Recently, as the technology of communication and computer science developing, the need for global safety defending growing fast, video surveillance and monitoring has attracted more and more attention by researchers. Video surveillance and monitoring analyzes video sequence by segmenting people, animals, vehicles and describing their action and behavior, which can be used for judging the occurrence of unusual events and reporting their warnings. This technology includes object segmenting, object extraction, object tracking, target partition and behavior understanding. It also has connection with many key problems of computer vision, pattern recognition and AI. Therefore, segmenting moving objects from video sequence and tracking them is a basic tache of video surveillance and monitoring.Besides, as multimedia application turning from broadcasting to object-based interacting, MPEG-4 comes into being, which makes it possible that video surveillance and monitoring is implemented from far away by internet as multimedia communication.This paper includes six parts, and targets at the application of vehicle tracking on super-way. The whole paper analyzed the algorithms of object segmenting, object extraction, object tracking and object-based MPEG-4 coding. First, object segmenting based on Gaussian mixture model was optimized, and object extraction method was improved. On the basis of traditional algorithms, a new object tracking algorithm according to this certain application was presented. Then, this paper analyzed the support of objects by MPEG-4 from both system stream and visual stream, and realized the object-based coding of vehicle tracking results on visual stream.