Invariant moments represent a very important shape feature of the image. With their invariant function of geometric transformation，they have been widely used in the field of image analysis. In this paper，shape features extracted from images by using three types of commonly used invariant moments，namely Hu moments，Zernike moments and Wavelet moments，were applied to high resolution remote sensing image classification and compared with the image classification only utilizing spectral information. The results show that，when the shape features defined by invariant moments are included in high resolution image classification，accuracies significantly increase. Higher accuracies can be especially achieved for those classes which have similar spectral features but different structural and shape features.