Study On Target Recognition And Tracking Based On Visual Selective Attention Mechanism

The visual selective attention model is based on the hypothesis or research results of Cognitive Science and Neuropsychology. it has important contributions for the development of Artificial Intelligence that introducing the visual selective attention model to the Information Processing and researching.Main research content of this paper is building a visual selective attention model which has a better quantitative capacity, to realize the recognition of the specific objects as well as the detection and tracking of the moving objects. First of all, it develops around building a visual selective attention model, whose main research content is analyzing a number of key technologies. It puts scale, salience and recognition in framework as well as analyzes the scale problem,representation method of scale space and saliency measurement in vision. This paper chooses features, which include luminance, color, orientation, and scale to lead attention. It uses the local iteration strategy to combine feature map and form the saliency measurement of feature space. Second, it estimates main scale of images, builds scale space of feature map and realize the salience measurement combining scale with feature, chooses the best scale for fixation point. Last, according to the scale of fixation point, it approximately estimates the size of fixation target to realize the modeling of ROI (region of interest).Finally, here conduct simulation experiment. The visual selective attention model of this paper together with feature recognition which based on local invariance realizes the recognition of specific objects. And according to information abstracting model based on ROI (region of interest), detect the regions containing potential targets. Meanwhile, use invariant moments to find the best match based on that the thought of"Detection-orientation- detection"is used to realize target tracking of moving objects.Under the above analysis, the model in this paper improves the disadvantage of classic attention model, which includes false focus of attention and lack of integrity that describing target, and then the improved attention model has the simulation experiment. The result shows that features and scale leading attention together as well as the improvement to consolidation strategy can retain important information of images more effectively and get focuses of attention and areas which fit visual system of human beings better.