The Reseach Of Super-resolution Algorithm Of Hyperspectral Imagery Based On Geologic Feature

With the rapid development of remote sensing technology,the application of hyperspectral images in contemporary social production and life has become more and more extensive.However,due to the imaging mechanism and hardware conditions,the spatial resolution is usually low,to the subsequent data processing and other processes caused a lot of inconvenience.Therefore,through the need for hardware participation in the signal processing method that is super-resolution recovery method to improve the image resolution has become the most meaningful means of improvement.The existing super-resolution technology mainly uses the information correlation of multiple images in space to reconstruct the high-resolution information,and cannot give full play to its spectral dimension information.Based on the characteristics of hyperspectral image and imaging principle,this paper studies the hyperspectral algorithm of hyperspectral images based on geologic feature redundancy dictionary.Based on the characteristics of similar spectral curves,the hyperspectral image is sparse and decomposed along the spectral dimension to make full use of its spectral dimension information and improve the spatial resolution.The main research contents include:A hyperspectral image redundancy dictionary training and sparse decomposition algorithm based on the geologic feature is designed and implemented.The clustering algorithm based on K-mean is realized,which can effectively divide the image data according to the feature,and the clustering effect is significant which has good adaptability and high efficiency.Based on the clustering results,a new algorithm is proposed to eliminate the redundant dictionary of hyperspectral image.This method can effectively introduce the feature of the feature into the process of establishing the dictionary library by reducing the classification of the different geologic feature,thus reducing the negative influence of the other dictionaries in the process of sparse representation.In the case of less data volume and sparseness,the accuracy of using the sparse representation of the image is improved,which lays the foundation for the super-resolution of hyperspectral images.A super-resolution algorithm for hyperspectral images based on geologic feature redundancy dictionary is studied and implemented.Based on the hyperspectral image feature information,the high and low resolution redundancy dictionary pairs of each kind of objects are obtained from the training data according to the correlation of the sparse coefficients between the high resolution image pixels and their corresponding low resolution image pixels.The image pixel is reconstructed by the high and low resolution constraint relation,and the high resolution image is reconstructed by the combination of the pixel space dimension.This method improves the pertinence of the algorithm to each kind of objects,so that it is superior to the traditional algorithm in the preservation of spectral features,which is beneficial to the use of the results in geological analysis and other fields.A hyperspectral image super-resolution algorithm based on least squares support vector machine(LS-SVM)is designed and implemented.Based on the super-resolution restoration theory based on high and low resolution redundant dictionary pairs,LS-SVM is used to improve the classification method.This method avoids the complexity of the high-dimensional space of the spectral image,and can achieve better classification effect in the case of small sample,so that the spectral classification redundancy dictionary elements are more suitable for the typical feature spectral characteristics.The experimental results show that the algorithm can effectively reduce the data loss of sparse representation of hyperspectral image signal in the process of super-resolution,and further improve the image resolution under the premise of guaranteeing the image restoration quality.