Detection And Tracking Of Moving Object In Video Image Sequence

Detection and tracking of moving object in video image sequences belong to the intelligent video surveillance. It is very active in the computer vision field. It has broad prospect for application in the fields such as intelligent transportation, virtual reality, scene monitoring, medical diagnosis, meteorological analysis and entertainment, etc.On the basis of collecting and analyzing large amount of data from the field of moving object detection and tracking, this paper expatiates on the major problems in current research, and then concludes a method that is suitable for the improvement in solving these problems.In the aspect of moving object detection, this paper studies sophisticated methods of moving target detection, and discusses strong points and weak points of background subtraction algorithm and temporal difference algorithm, which is now widely used in moving object detection.While in the aspect of moving object tracking, a novel fragments-based self-adaptive feature selection of object tracking in video sequence is proposed to overcome the common and complicated problem of background cluttering, object occlusion and object pose variations. We skillfully divide object and background into several patches with the same size respectively, and choose the most discriminative 'object-background patch couple' to accomplish object tracking and get a satisfactory performance against the common and complicated problem of background cluttering, object occlusion and object pose variations. Meanwhile, integral histogram, as the key of our algorithm, greatly increases the efficiency of calculating color histogram in different rectangular regions.Extensive experimental results demonstrate that our moving object detection and tracking algorithm is more robust and shows better performance, compared with traditional tracking methods.