Abstract: At present, the method of extraction forestland is mainly based on selecting samples by supervised or semi-supervised, and the efficiency is low. For this reason, this paper proposes a multi-feature HSV transform hyperspectral image forestland extraction method. We first performs relation correction processing of original remote sensing images.Then use the parameters of normalized vegetation index (NDVI) and principal component analysis (PCA) to obtain the composite images. Finally, HSV transform is used to extract forest land information through color segmentation of image by setting color value range. The results show that the extraction accuracy of the high-resolution forestland can reach 96.29％, and indicated the effectiveness of the method.