Study On The Classification Method Of Rock Ore Features Based On Hyperspectral Remote Sensing

Because the hyperspectral remote sensing image has rich spectral and spatial information,so the spectral curves of materials can be obtained,furthermore,it provides valuable information for material and object recognition,which can be classified and recognized.In order to make use of remote sensing technology to find deposits and carry out prior geological mapping work,this paper takes the rock deposits as the research object and finds out in the analysis of the existing hyperspectral remote sensing rock and Mineral Feature Classification Algorithms,most of them only use spectral dimension information to classify rocks and minerals(such as SAM?SVM?etc.),the complexity of classification may be increased due to the change of crystal structure,the amount of doped elements,which leads to the unsatisfactory classification accuracy.In this paper,the spectral and spatial characteristics of ores are studied,which lays a theoretical foundation for the following classification work;the existing hyperspectral image classification method based on spectral features,hyperspectral image classification method based on SVM is studied,and combined with the experimental analysis of the reasons affecting its classification accuracy;a supervised classification algorithm based on depth learning is proposed,which combines the spectral and spatial information of pixels with convolutional neural network,four kinds of ores,Alunite,Kaolinite,Montmorillonite and Chalcedony,are classified in the AVIRIS hyperspectral data.The experimental results show that the overall classification accuracy of the ores is 90.58%,the Kappa coefficient is 0.8676,and the algorithm has good classification performance,it is shown that the algorithm proposed in this paper is of practical significance for the classification of hyperspectral rock and mineral features.