Ensemble learning based on multi-features fusion and selection for polarimetric SAR image classification

Aim at the problems of low classification accuracy rate of the traditional single feature and the multi-features dimension disaster, a ensemble learning algorithm based on multi-features fusion and selection is proposed, and is used for polarimetric SAR image classification. Firstly, various features of SAR image is extracted and fused by normalized; then, different feature selection methods are used to select features, and different feature subsets are generated; thirdly, different feature sets are used to train the SVM classifier, and the individual classifiers will be got; finally, each individual classifier is ensembled to a ensemble classifier. The experiments indicate that higher classification accuracy can be obtained by the algorithm.