Research On Moving Object Detection And Tracking Algorithms On Dynamic Image Sequences

Detection and tracking of moving object in video image sequences belong to intelligentvideo surveillance. It is a very active subject in field of computer vision. It involves subjectabout computer graphics, image processing, pattern recognition, artificial intelligence, neuralnetwork, psychology and physiology, etc.. It has broad prospect for applications in the fieldssuch as intelligent transportation, scene monitoring, medical diagnosis, meteorological analysis,entertainment and so on.As to detecting and tracking of moving object in video image sequences, there are stillmany problems no matter in theoretical research or in applications. A mass of researchers aredevoted to this area and have already achieved many progresses. The dissertation studies the keytechnologies based on the current conclusion, and discusses the technologies about the movingobject detection, segmentation and tracking.In moving object detection and segmentation, the dissertation summarizes the objectdetection and segmentation technologies. Several common detection and segmentationalgorithms are introduced, and their advantages and disadvantages are discussed. A newmethod is proposed based on the image sequence group and the improved OSTU algorithm. Itcan effectively and fast detect low-speed moving objects in stable scene.In moving object tracking, the dissertation introduces the mean shift algorithm's basicprinciple and application, and the algorithm is extended to continuous image sequences intocamshift algorithm. It can effectively track the moving object based on dynamic changes of thecolor probability distribution. The experimental results show that the algorithm is rather robustand effective.