Study on data mining technology in hyperspectral remote sensing

In this paper, the problems rise in hyperspectral data mining and some key issues should pay attention to were proposed based on the analysis on the state-of-art of hyperspectral data mining. The problems are as follows: data mining precision, mining algorithm efficiency, new hyperspectral data mining algorithms, the uncertainty in hyperspectral data mining, visualization in hyperspectral data mining process, and knowledge presentation, interpretation, estimation and management etc.. Some key issues should emphasize in the future are: systematic hyperspectral data mining theory, dimensionality reduction, mining spatial and temporal knowledge from images, and mining distributed data and mining multi-agent data. Also the framework and architecture of hyperspectral data mining were put forward in this paper. Hyperspectral data mining framework includes some subparts as follows: data selection, data preprocessing, data transfer, data mining and pattern estimation. And the architecture is composed of database, data warehouse, database management system, repository, mining process, user interface etc.. At last, an algorithm which named Relational perspective map (R-PM) was introduced into the field of hyperspectral data mining. By the experiment on the spectra data from USGS spectral library, it proves that this algorithm is suitable to discover those spectral features and to identify and discriminate object classes based on their spectra.