Classification extraction of land coverage in the Ejina Oasis by airborne hyperspectral remote sensing

摘要

The hyperspectral data of the Ejina Oasis is performed dimensionality reduction by the minimum noise fraction transform, classified by the maximum likelihood method and clustering processing and finally, the classification of land coverage is obtained. The thesis analyzes the results of two dimensionality reductions, and finds the superiority of the minimum noise fraction. In the classification of land coverage, the overall accuracy and Kappa coefficient are respectively 87.75% and 0.8401. The classification is of high precision and can provide effective parameters for ecological research.