Dimension Reduction of Self-organized Neural Network Classification for Multi-band Satellite Data

To meet the productive application of satellite data in land cove classification the higher resolution, the more bands are used, the more accurate results can be produced. A clustering in 3D self-organized neural nodes is used. ASTER data is a new kinds of sensors, 3 bands with 15m resolution and 6 bands with 30m resolutions. Dagang region in Tianjing is selected as a case study area. The Wavelet fusion is applied to fused different bands with different resolutions, then the self-organized neural network classification for land use is performed. Finally classification results is compared with that of the maximum likelihood classification with the same training samples. The accuracy of validation result is over 94%.