| Water is precious resources of human production and living, It is great significance for environmental protection, disaster monitoring and geological exploration.This article adopts the method of object oriented to high resolution remote sensing image to extract water body information. First by remote sensing image segmentation to obtain the smallest unit of image analysis-Image objects. Then the characteristics of image objects are extracted, In order to build a feature vector needed by classification recognition. Finally, water information was extracted from remote sensing image by the method of using machine learning. In this paper, the extraction methods of concrete is optimized, The main results were as follows:In the section of remote sensing image segmentation,This article adopts the method of mean_shift and the Merging rules of heterogeneity minimization to Image segmentation.The method is based on a variety of characteristics, the center of the segmented regions is independent, having good segmentation precision and adaptability. On the basis of segmentation by mean_shift method,region merging is carried out based on the Merging rules of heterogeneity minimization. Compared with the traditional image segmentation method, the method of this paper achieved better segmentation results and is More suitable for remote sensing image segmentationIn the section of feature extraction, spectrum, geometry and texture features is extracted. MBR is used to extracted Geometry information, GLCM is used to extracted Texture feature. The method of Water body information extraction Based on A variety of features is better than single spectrum characteristic method.In this paper, we use the SVM to the identification of water body information and use PSO to optimize parameters of SVM. The parameter Setting is based on Multi-Strategy.Compared with the method of pixel and Maximum likelihood, the method of this paper achieved better results on the accuracy of water body information extraction. |
PDF Full Text Request