Research On Fast Targets Detection Algrithms Based On Visual Attention In SAR Images

Synthetic aperture radar(SAR) is widely used in the military field with its all-weather, all-time, strong penetration and other advantages. It provides the information of reconnaissance and target detection for military operations. In the meantime, there are also a lot of applications in the civilian field, such as the weather, geomorphology, road detection and the disaster monitoring. SAR target detection is an image processing technology, which makes use of the information of targets’ gray, texture, shape, edge and direction to determine their position in the SAR image, and separates the targets from background. With the constantly improvement of SAR image resolution, target information shows explosive growth, how to detect the SAR target quickly and efficiently is becoming the hot issue in the SAR image processing field.The human eyes can filter and process surrounding massive amounts of information quickly, and only to respond to useful information, this mechanism is called the visual attention mechanism. This initiative and selective behavior of human eyes is called attention. The computational models of visual attention use mathematical modeling methods to simulate human visual perception system for information processing in a computer simulation environment. Visual attention mechanism is a hot research topic in the field of image analysis and processing.The visual attention mechanism is introduced to the field of SAR target detection in this thesis. The main contents and innovations are as follows:(1) The mature visual attention models at home and abroad are researched. The classical model of Itti and Hou spectrum residual model, which are widely used, are analyzed in detail combined with the work process of human visual perception system. The deficiencies of two models in SAR image processing are described.(2) A visual attention model is researched, which is suitable for SAR images. The model is based on the framework of Itti’s model. The features that can characterize the SAR target are extracted, the saliency map of SAR image is computed by feature integration mechanism. The simulation results of comparative experiments show that the new model can generate better saliency map, the contours of targets are clear, the speed of calculation is fast, and the results of detection is accurate.(3) Based on visual saliency, a new SAR target fast detection method is proposed. Firstly, the proposed visual attention model is used to calculate the saliency map of the input SAR image, and then the salient regions are smoothed and filtered combined with the size information of SAR targets. Lastly, through the rectangle completion and registration process of the regions of interest, the detection of SAR targets is finished. The idea of this method is novel, which can detect the SAR targets combined with visual attention. The results of experiments show that compared with traditional methods, our method has fast calculation speed and accurate positioning.