A New Method to Improve Classification Accuracy of Fused RADAR and Optical Data

In past few decades, feature selection and learning have been considered by many researchers in terms of reducing the dimensionality of feature space and optimal feature selection. In traditional methods, feature selection and learning, are separately done. In this paper, a new method of supervised feature selection and learning, based on sparse regularization, was used to improve the classification accuracy of two pairs of fused radar and optical data for the first time. NMF features extracted from the images and the extracted features were used in two learned and unlearned forms as input to the SVM classifier, which choose as a base classifier. The results showed significant improvement in classification accuracy, resulting from the implementation of the sparse regularization algorithm based on L-2,L-p norm.