Sereda A. Method of object tracking in video and its implementation on parallel computer systems

The object of research is the process of automatic computer detection and tracking of objects in a video stream. The goal of this work is to increase efficiency of computer object tracking systems by the means of development a new method of objects detection and tracking in video able to track partly occluded objects and its implementation on parallel computer systems. The methods of research are the methods of computer vision theory, computing geometry, set theory, fuzzy logic, probability theory, theory of computation, theory of parallel computation; computer modeling on multiprocessor computers, cluster and multithreaded graphic processors. Main results: a new method and software subsystem for automatic detection and tracking of objects in video on parallel computer systems is developed; a new method of increasing background hovelling accuracy. Scientific novelty: the new method of automatic detection and tracking of a set of objects in video is developed; methods of search of image fragments in a frame are improved; the new method of lowering of amount of errors during initialization and update of background model is offered; the mapping of proposed tracking method to the MIMD parallel computer systems is developed; the new hardware-oriented algorithms for search of image fragment in a frame for CUDA architecture are developed. The adoption: in the NPP TOV "AMI" in the development of a digital security system; in the educational process of SHEE "Donetsk national technical university". The spheres of use: video surveillance systems, automation of sporting events analysis.