The Research Of Rare Earth Mineral Area Objects Information Extraction From High-resolution Image Acquired By ZY-1Satellite

With the development of Remote Sensing (RS) technology, the high resolution images areused more and more in the actual production of life. The high resolution RS image containsabundant information, and its amount of data is large. It becomes a research hotspot of extractinginformation on a good precision for a high resolution RS image. Rare earth is a valuable strategicresource of our country. Rare earth application widely, it is called “industrial vitamin”. However,disordered mining, mining technology such like stripping out of the surface and in the rare earthore, which have caused destruction of vegetation, soil erosion, environmental pollution what hasbeen a threat to the lives of local people. The one problem needs to be solved right now is thathow to acquire the ground information in rare earth mine quickly by using RS technology, andimplement of mining dynamic supervision. The work carried out in this paper is beingcombination with the practical needs of rare earth mining regulation and requirement of highresolution image information extractionIn this paper, we take a rare earth mining area as the study area, which is located in XunwuCounty of Ganzhou city Jiangxi province.We take the ZY-102C satellite data as the primary dataand use object oriented classification method for feature information extraction. This paper filledthe blank of some related research. The extraction got vegetation, water, roads, buildings, and bareland5categories which divided into11sub-categoriesThe feature information extraction process can be divided into four parts.Firstly, do pretreatment for image data. Pretreatment includes ortho-rectification, imagecutting, color enhancement. The research area is chosen by pretreatment. The true color image,which composed by weighting method for visual interpretation in later process is been chosen,too.Secondly, image sharpening. Considering the resolution and spectral characteristics ofresource satellite space, choose10meters multispectral image and2.36meters panchromaticimage do image sharpening. The image sharpening methods are HSV transform, Brovey transform,Gram-Schmidt transform, and Principal components transform, Pan Sharpening. Compared the6sharpening images which produced by6transforms through subjective and objective, Pansharpening image is chosen as the source date for feature information extract.Thirdly, extract of feature information. The classification method be taken in this paper isobject oriented. It takes four stages, masks images layer by layer. And the research area issegmented, ground features are extracted, it is using rule classification and support vector machinemethod. According to the research area image, it is established features of interpretation labelingand feature list, set up segmentation parameter, then make feature information extraction done. Itis using the confusion matrix of evaluating the extraction results. The overall classificationaccuracy is78.71%, the total Kappa coefficient is0.7632, it means the classification accuracy is good.Finally, conclusion: According to the result of extraction, clearly this technology suitable forhigh precision information extraction of target in the study area.