Gaussian kernel-based Fuzzy Rough Set for information fusion of imperfect images

Imperfection of remote sensing data greatly affects the performance of information fusion algorithm. To solve this problem, a Gaussian kernel-based Fuzzy Rough Set fusion algorithm is proposed, since Fuzzy Rough Set theory is an effect tool to model uncertainties of data. For feature reduction a novel index is proposed to evaluate the significance of features, considering both the relevance between features and decisions and the redundancy of features. Thus the most informative features are selected for classification. Experiments with standard test data and real remote sensing data show that the classifier can achieve a high accuracy using feature subset selected by the proposed method than using the full feature set.