基于支撑向量机概率输出的高光谱影像混合像元分解

Unmixing of Hyperspectral Imagery Based on Probabilistic Outputs of Support Vector Machines

提出利用支撑向量机(SVM)后验概率来分解高光谱影像的混合像元,通过支撑向量机的输出值转化为两两配对的后验概率,再由两两配对的概率值求得多类后验概率,并以像元所属类别的后验概率作为地物的组分信息。实验结果表明,该方法能较好地估计出混合像元的组分比。

This paper proposes to estimate abundances from hyperspectral image using probability outputs of support vector machines （SVMs）, trains a SVM with a gauss kernel function, and trains the parameters of an additional sigmoid function to map the SVM outputs into probabilities. An experiment of real hyperspectral image is conducted to validate the procedure. The abundances estimated by SVM comparison with those of the linear spectral un- mixing are also given. The experiment results show that the method can provide effective result of abundance estimation for hyperspectral image.