Research On Rough Sets Based Remote Sensing Images Fusion And Classification Method

Remote sensing images fusion methods based on rough sets and remote sensing image classification methods based on spectrum and texture features are studied in this paper.Firstly, a images fusion method based on rough sets and contourlet transformation is proposed in this paper. Equivalence class concept in rough sets theory is used in the division of high-frequency coefficients by image contourlet transformation. The high-frequency coefficients are divided into different equivalence classes by the gradient property. Different classes of the high-frequency coefficients are combined using different rules. Use the new fusion high-frequency coefficients and new fusion low-frequency coefficients to get the fusion image through inverse contourlet transformation. Experiment result results show that this images fusion method compared with the method"Using the Second Generation Curvelet to Improve IHS Transform Merge Remote Sensing Images"can not only preserve spectrum information but also make better in enhance detail information.And then a images fusion method based on rough sets and genetic algorithm is proposed. The weights of low- frequency coefficients and every class of high-frequency coefficients are optimized by genetic algorithm. Use the optimizing weights to combine the coefficients and get the fusion image through inverse contourlet transformation. Comparing with the method"Image Fusion Algorithm Based on Wavelet and Rough Set in Intelligent Transport System", This images fusion method is better.The study of remote sensing image classification methods in this paper is based on the paper"Research on the Classification Method of Remote Sensing Images Based on Spectral and Texture Features Fusion". Two improving classification methods based on D-S evidence theory and a improving classification method based on fuzzy reasoning and similarity are proposed. In the two classification methods based on D-S evidence theory, spectrum probability distribution and texture probability distribution are combined through direct combination method and selection combination method. In the classification method based on fuzzy reasoning and similarity, spectrum and texture memberships are combined through selection combination method. Experiment results show that accuracy of the classification method using the fusion image can effectively improve. And the multi-feature classification method applying combination of spectrum and texture features is better than the single spectrum feature classification method. Comparing with the former methods, The classification methods are better.