Comparison of Land Cover Types Classification Methods Using Tiangong-2 Multispectral Image

Abstract

In this paper, Qinghai Lake and Taihu Lake are used as experimental areas, and the visible and near infrared spectrum range of Tiangong-2 Wide-band Imaging Spectrometer are selected for classification research. On the basis of preprocessing, the images are classified by several common classification methods such as Minimum Distance Classification (MDC), Maximum Likelihood Classification (MLC), Spectral Angle Mapping (SAM) and Support Vector Machine (SVM). The classification results are verified using confusion matrices. In the land cover types classification of Qinghai Lake area, the overall classification accuracy of SVM is the highest, which is 99.04%, followed by SAM of 98.78%, MDC of 97.84%, and MLC of 86.89%. In the land cover types classification of Taihu Lake area, the overall classification accuracy of SVM is the highest, which is 92.44%, followed by MDC of 88.90%, SAM of 84.01%, and MLC of 71.01%. After comparative analysis, the practicality and superiority of the SVM method in the image classification of visible and near infrared spectrum range of Wide-band Imaging Spectrometer are proved, which provides a technical reference and theoretical basis for the classification research of Tiangong-2 data.