Abstract: Due to the complex diversity of features, accurate identification of their classification accuracy is of great significance to remote sensing data processing. In order to improve the classification accuracy of multi-spectral remote sensing data based on Landsat 8 data, this paper proposes a method of fusing NDVI and NDBI with different scales to classify multi-spectral remote sensing images. Firstly, the first principal component of the multi-spectral data is extracted as the guide image, the original image is the input image, and the guide filter feature set with filter radii of 2, 4, 6 and 8 is extracted in turn. Then,the guided filtering feature set with different filtering radii is fused with the NDVI and NDBI features of the image, and the method of support vector machine is used to supervise the classification, so as to explore the influence of guided filtering of different scales on the classification accuracy of multi-spectral remote sensing images. The experimental results show that:①Guided filtering can better retain the edge features of the image while removing noise.②Guided filtering can improve the classification accuracy of multi-spectral remote sensing images, and the classification accuracy of different sizes of guided filtering radius images and original images has been improved to different degrees compared with the original image,the highest overall accuracy reaches 99.776 3%, and the Kappa coefficient is 0.997 1.③Guided filtering of different scales will obtain different classification results,and when the filter radii R=2, the classification accuracy of the image is the highest.

Key words: guided filtering, remote sensing imagery, classification accuracy, supervise classification