Remote Sensing Image Automatic Classification Based on Texture Feature

Texture feature of ground object not easily change influenced by environment, so it is stable. In order to reflect object feature better, we extracted texture feature with wavelet transform method, Including contrast, correlation, energy and homogeneity. And we do the image classification based on texture feature. In order to test the method, we adopted QuickBird satellite image to experiment, and then compared with image classification based on spectral characteristics. Result suggests, in a way, that the image classification based on texture feature is able to improve the remote sensing image automatic classification precision and obtain the better classification effect.