Classification of Landsat 8 OLI image using support vector machine with Tasseled Cap Transformation

Abstract

Landsat acquires the longest space-based moderate-resolution land remote sensing images continuously. Compared with the other earlier Landsat satellites, Landsat 8 has several new characteristics in spectral bands, spectral range and radiometric resolution. Therefore, there is a strong requirement to analyze the characteristics of the Landsat 8 for land cover classification, global change research. In this paper, Landsat 8 OLI image was used with Support Vector Machine (SVM) and Tasseled Cap Transformation (TCT) for land cover classification. Firstly, the Top of Atmospheric (TOA) reflectance based TCT was developed based on Landsat 8 OLI images. Then comparison of ISODATA, K-Means and SVM of all original eight Landsat 8 OLI bands and both of TCT Greenness and Wetness in land use and land cover classification was done. The present results indicated that compared with using the original 8 Landsat 8 OLI bands, the classification results from ISODATA and K-Means based on both of TCT Greenness and Wetness had better robustness and accuracy, and the classification using SVM with TCT had better efficiency and accuracy.