Remote Sensing Image Fusion And Classification Based On FastICA

With the rapid development of remote sensing science and technology, a largenumber of different sensors (multi-temporal, multi-spectral, multi-sensors,multi-platform and multi-resolution) for Earth Observation applications increaseswidely. Compared to single-source remote sensing image data, multi-source remotesensing image data is redundancy, complementarity and cooperative.Contrapose the current situation of multi-source remote sensing data fusionmethods，applied research and existing problems, with the application background ofland use classification, combined with the characteristics of SAR data and SPOT data,different fusion methods were used in data fusion, an ideal result of fusion andclassification was obtained. The main contents and results are as follows:1. Look up the SAR image denoising references, take experiments on SAR imagedenoising, better denoising SAR image was gained to improve registration accuracyof SAR image and SPOT image and the classification accuracy of fused image;2. Comparative study of several common fusion method: IHS method, Broveymethod, HPF method, WT method, PCA method, and a new proposed method-FastICA method, the results showed that: a) ICA and PCA algorithm have theadvantage over retained linear features and details, while the ICA algorithm has theadvantage over noise filtering than PCA algorithm; b) Combined with all theevaluation indicators，ICA algorithm has better fusion effect. The quality of fusionresult is not entirely related to the merits of the algorithm, the image data types areclosely related, the results do not apply in all cases. In actual work, we should selectthe fusion processing method according to the image data and work requirements;3．SAR image and SPOT image fusion not only to retain the spectral informationof SPOT images, but also retains the details of the SAR image, it can be used forbetter visual interpretation.4. Compared to the IHS method, Brovey methods, HPF method, WT method andPCA method, the FastICA method of SAR image and SPOT image fusion has betterresult that has higher classification accuracy image.