Study On The Technologies Of The Object Tracking In The Surveillance Video

Motive object tracking in video is a very important and active research task in the field of computer vision, image processing and pattern recognition. A focal problem in the field of object tracking is multiple objects tracking. The method of human tracking with fixed camera is widely applicable.On condition that camera is fixed motive region is detected by background subtraction based on Gaussian mixture model. Color distribution feature in HSV space is extracted as an object template. There are always occlusion statuses in multiple objects surveillance video. The objects in occlusion and the ones not in occlusion are distinguished by object activity statuses prediction and classification. The object activity status in video can be classified into six classes: new object, single object, abnormal single object, occlusive object, static object and leaving object. And the occlusion or slipping statuses of objects in the next frame can be estimated by motion estimating with kalman filter. After objects are classified, the different locating and template update strategies are applied to the objects in different statuses.Some experiments are carried out on video sequences. Experiment results show the method of tracking is available.