The Design And Implemention Of Image Target Detection Based On Visual Saliency

The visual saliency model, which is based on studies and hypotheses on psychology, neurosciences and epistemology, uses mathematical model to simulate human visual system. Visual Saliency is a visual cognition mechanism to guide visual attention distribution in the Human Visual System. It is important for the visual cognition system, and detects regions of interest by detecting prominent points in an image, for it takes regions of interest, which contain potential target as the muster of points that have prominent features, so that the machines can obtain useful information from these data quickly and accurately. In image processing tasks such as image retrieval, target detection and surveillance, it would be more convenient to focus on detection of the regions of interest by visual attention model than focus on the whole image.We studies on visual attention Itti and SR model based on the different space of image processing in this thesis. Analysis of the formation mechanism of the two models in detail, whose strategies of the detection of primary visual features and combination of conspicuity maps are investigated and improved. In this paper, we propose a saliency detection algorithm based on color features. This thesis first introduces visual attention mechanism and progresses that have been made in the field. The basic concept and theory, the measurement method and the characteristics of general model framework related to visual attention mechanism are also introduced. Summarize disadvantage of the existing models, obtain the ideas of research on feature extraction method and saliency figure fusion strategy.Since there are some cells in the visual system which react to the color information, the color feature is added into the detection of primary visual features as a kind of the primary visual feature in the new model, which makes use of the RGB information and broadband tuning information to form conspicuity maps. The conspicuity maps are combined as saliency map by a global enhancement method to guide attention and detect regions of interest.This thesis also designs the random rectangle, sliding window, symmetry around the window, the multi-scale sliding window to improve the initial figure, can effectively eliminate the background information in the graph, highlights the targets of interest in the images, bring convenience for the subsequent application. Experiment results have demonstrated that the improved model can detect the target in the image, maximize theelimination of background and noise information, will be interested in the human eye is highlighted areas.Saliency algorithm is developed in MATLAB GUI platform.Spectrum analysis and the binarization and morphological image processing method are introduced into the method to detect and locate the target in remote sensing and some other region such as the vehicle license plate, traffic signs and traffic signs. It makes up the defects of the existing methods, for example,leak detection and fault detection and is proved practical...