Research Of Regions Of Intensity Based On Visual Perception

When flying in the upper air, hawk can keenly catch and track the game in the complex background , which is running on the ground .When apperceiving and catching targets, vision system of animals is very fast and accurate. The formidable vision function of animals has been attracting many workers to study their optic nerve information processing system. The origin project of this tissue is called" automatic tracking theory and technology of movement goals based on visual perception ", and its aim is to automatically detect and track the goals by simulating visual function of animals. The origin project achieves the aim of detecting and tracking the targets by extracting their spatial and dynamic characteristics, and fusing them to get space-time remarkable map. The article attempts to improve the veracity of interest maps through three ways, .i.e. the visual model, standardization and combination, and to extract space characteristic and initially realize the function of detecting targets.At first, this article introduces the research situation of the visual attention models, discusses the structure and characteristics of visual perception system, and focuses on the description of Itti model, which is driven from bottom to up and used widely.Considering Itti model's lack which can't notice all goals, and even any goal when the goals is disturbed by strong background and noise ,this article adds the extraction of outline characteristic into the Itti model so that promotes veracity of interest maps describing tracks , and according to experiment ,it confirms that the improvement model is the a better successful model.After introducing the frame of the improved model, this article compares the standardization of the overall strengthening with the partial iteration, through the experiment. The partially iterative standardization causes the interest maps to satisfy sparse distribution, suppresses the background and the noise effectively, and conforms to the human visual characteristic.Because the average value combination which has not considered the priority of the different feature maps can not suppress the strong background and noise effectively, this article proposes the dynamic combination based on the forecast mechanism. It dynamically assigns the weight to each map according to its contribution to the interest map to get the interest map. According to two experiments, the dynamic combination based on forecast mechanism can effectively describe the interest map.