Mineral Identification By Hyperspectral Based On Data Mining

Hyperspectral Remote Sensingis in the electromagnetic spectrum visiblelight, near infrared, mid-infrared and thermal infrared range, the use of imagingspectrometer to get a lot of very narrow spectral image data of continuoustechnology. Hyperspectral remote sensing with high resolution, typically reach10~2λ of magnitude. Due to the high spectral resolution remote sensingtechnology in terms of the great advantages and related technology matures, itsapplications are increasingly widespread around the world for the developmentof remote sensing very seriously.Mineral identification hyperspectral remote sensing technology is the mostsuccessful, and most can play one of its strengths applications, it enables theidentification of lithology geological remote sensing to identify a single mineraldevelopment as well as the chemical composition and crystal structure ofminerals. However, due to certain minerals, especially the spectralcharacteristics and mineralization related hydrothermal alteration mineralsdifference smaller, more affected by the mixed spectrum of minerals and otherfactors, most of the spectral identification of some of the spectral characteristicssimilar mineral easily confusion and false phenomenon. In addition, the samekinds of minerals due to different developmental processes and the developmentof the state, its composition, structure and spectral characteristics will have somedifferences, so that the spectral characteristics of geographical areas; and becauseof the different spectral characteristics of minerals with the measurementconditions will be light, background color, particle size and other factors for thevarious changes.Data Mining refers to the automatic search from large amounts of datawhich has hidden in a special relationship of information process. Given the highnumber of spectral images with many bands, information and large massive dataredundancy features, data processing method based on traditional multi-spectralremote sensing image data processing can not meet the needs of high-resolutionquantitative hyperspectral imaging spectrometer measurement system datamining technology has become an important means of extracting usefulinformation from the mass model hyperspectral data and discovered knowledge.Related technical field of data mining methods will be applied to the field of hyperspectral remote sensing, marine extracts knowledge of the data is theinevitable trend of development of hyperspectral remote sensing.In this paper, the characteristics of hyperspectral data characteristicparameters related to Naive Bayes, K-Means clustering algorithm, such asclassification, based on spectral modeling development for hyperspectral dataprocessing and application of spectral matching and optimal band selectiontechnology, and using the software that comes with Envi conduct hyperspectraldatabase application potential of hyperspectral data mining technology research,exploration hyperspectral remote sensing data in mineral extraction, mineralidentification and other aspects.