Abstract: Aimed at improving roads in high resolution remote sensing image information extraction accuracy and the amount of information, this paper focuses on improving the image from the segmentation, and introduces texture and spectral characteristics of the image. An image clustering segmentation method based on combined features is proposed. We divide integrated features into three kinds of texture features and spectral characteristics.By selecting the size of the window from the gray levels and the relative orientation of the pixels on the four parameters, gray level dependence matrix extracts image texture features, after contrast five kinds of suitable texture characteristics for remote sensing image, this paper chooses contrast, angle second order moment and entropy as the texture characteristics, and chooses average gray image pixel and standard deviation as spectral features. Finally, based on segmentation image, which segmented by the comprehensive features of road, using image geometric characteristics and mathematical morphology such as open operation, corrosion, closed operation, refined, and the image processing algorithm to get the final road extraction results. The experimental results show that the method can be used to complex road information extraction.

Key words: high-resolution remote sensing images, the texture characteristics, spectral features, mathematical morphology, road extraction