Geo-information Science Foundation Series: hyperspectral remote sensing image classification with support vector machine applied research(Chinese Edition)

Hyperspectral remote-sensing is one of the hotspots in the current remote sensing science and technology development, while the classification of hyperspectral remote-sensing images is the most important task of hyperspectral remote sensing image processing. Based on introducing the basic knowledge and commonly used methods, the book focuses on exploring the classification of hyperspectral remote sensing image based on support vector machine (SVM) deeply. Firstly, it introduces the common hyperspectral remote sensing image classifier, and discusses the dimension reduction processing and feature extraction methods in the classification. By introducing the basic principle of SVM and its strategy in hyperspectral remote sensing image classification, it focuses on studying the kernel function design methods of the SVM, and based on analyzing all kinds of scheme or realizing SVM, proposes the binary tree multiclass SVM based on JM distance. Finally it discusses the comprehensive application of characteristics of SVM classifiers, and studies the SVM classification based on the multinuclear SVM combination spectra and special texture feature.