Tracking Of Moving Objects In Videos

Video processing is a technique of processing individual frames or images. The processes involve acquisition, manipulation, transmission, analysis and compression. This paper focuses on video analysis; it includes motion segmentation and motion tracking. Tracking objects in a video containing extremely crowded scenes is a challenging due to the motion and appearance variability produced by the large number of people within the scene. The individual pedestrians collectively form a crowd that exhibits a spatially and temporally structured pattern within the scene. The video is divided into sub volumes. The local spatio-temporal motion of the sub volume is extracted. Hidden Markovian model is used to train on the spatio-temporal motion pattern. From the model the spatio-temporal motion pattern that describes how the object moves in a video is obtained. The extracted information is used as the priori for tracking. Key words: Bayseian, HMM, spatio-temporal pattern, Tracking,