Abstract. Methods of automatic segmentation and simultaneous tracking on several objects at inhomogeneous background are reviewed. Adaptive correlation algorithms of object tracking are developed for the case when initial target destination is secured. Several methods are proposed for automatic capture and tracking of numerical objects. The correlation tracking algorithm employs adaptation of the slide window size to the object shape and dynamic signal accumulation technique. The problems of automatic segmentation and simultaneous tracking of several objects are solved on the base of motion vectors (optical flow) and Gaussian shape model. Presented results of experimental testing of developed algorithms makes it possible to evaluate their efficiency and reliability.

Keywords: segmentation, multi-object tracking, automatic capture, layering representation, correlation tracking, motion vector.