Segmentation For SAR Image Based On Spectral Clustering Algorithm

Synthetic aperture radar (Synthetic Aperture Radar, SAR) is the microwave imaging system, can't be limited by the weather, geographical and time factors, can get high-resolution imaging of the Earth's surface, and also can find the underground targets hidden by vegetation, providing plenty of land and marine geographic information and military intelligence, so it has been widely used in military and civilian fields, what's more, SAR image segmentation is the basis for SAR image understanding. This article combines the Nystr?m sampling technique with the spectral graph theory, it takes the SAR image segmentation as the goal, and it carries on the fast and efficient SAR image segmentation method, and through simulation experiment and actual SAR image experimental to verify the effectiveness of the method. The research contents are as follows:1. Carrying on the spectral clustering based on the automatic clustering method to achieve the SAR image segmentation. First of all, we bring in the matrix perturbation analysis theory, and construct the automatic clustering algorithm structure, which is suitable for SAR the image segmentation, further, we combines the automatic clustering method and the spectral clustering algorithm, and through the simulation experiment to verify the new algorithm performance, finally, we apply this new algorithm in the actual SAR image segmentation, and carry on the comparison and analysis with the traditional spectral clustering method.2. Researching on the automatic spectral clustering based on SAR image through the improved scaling parameter method. First we compare the Local N neighborhoods model, the global N neighborhoods model and the auto-adapted neighborhood model, next according to the global construction characteristic of the SAR image, we construct the auto-adapted neighborhood method to estimate the value of the scaling parameter, then use the auto-adapted neighborhood method to improve spectral clustering algorithm, and finally we compare these automatic spectral clustering algorithm in the simulation experiment and in the actual SAR image segmentation.3. Researching on the automatic spectral clustering based on SAR image through the improved similarity function method. First and foremost, according to the pixel value and the spatial location of each pixel in the SAR image to construct the similarity function, which can describe the essence structure of the SAR image, what's more, we through the new similarity function to improve the spectral clustering algorithm, last but not least, we apply this new spectral clustering algorithm in the simulation experiment and in the actual SAR image segmentation, and carry on the comparison with the traditional spectral clustering method.