Research On Target Tracking System Based On The Movement Of The Selective Attention

Visual attention mechanism is based on cognitive science and neuropsychology relatedfields research results, using the method of mathematics establish perception model to realize thesimulation of biological vision system. The main work of this paper is for the problem that invideo sequence, the tracking algorithm limited by the background conditions will influence thetracking results, introducing visual attention model, making Itti model and Mean-Shift algorithmcombined, completing the development of target tracking system based on the movementselective attention. The main work as follows:First, on the basis of the analysis of the relevant tracking algorithm and visual attentionmodel, focusing on the Mean-Shift algorithm, although this algorithm has small amount ofcalculation, in the complex background tracking stability is not high, therefore introducing Ittimodel, compensating for the instability of the Mean-Shift algorithm by means of finding theregion of interest automaticly to solve the difficulty in tracking.Second, combining movement feature and weigh coefficient to improve Itti model. ClassicItti model only take into account the color, intensity, toward feature, but the attention of humanvisual perception system input information is not limited to the above three, in the field of targettracking, movement feature is a class of visual stimulation related to time, also importantinformation of biological perception target. So, movement feature is introduced to the Itti model.In different scence, the contribution of each feature causing the attention of the perceptualsystem is different. Itti model introduce feature weigh coefficient to measure the size ofcontribution that each feature attention to target, adapting to the needs of the different scence. Last, using sift feature matching algorithm to make improved Itti model and Mean-Shiftalgorithm combined. In the tracking process, using mean-shift algorithm search the target area inthe current frame and comparing sift feature points with the region of interest that Itti model findin the current frame, if meet matching success condition, target final position is the Mean-Shiftalgorithm to determine the position. Instead, comparing the region of interest that Itti model findwith the target area in the first frame, matching success terminate matching, otherwise, usingattention shift mechanism to find the next region of interest continue to compare until match thecontidion, target final position is the Itti model to determine the position. The relatedimplementation results show that the effectiveness of proposed method.