Research On Hyperspectral Remote Sensing Image Classification Algorithm Based On Generation Learning

Remote sensing image classification is a very important technology in the application of remote sensing technology.In recent years,the application of hyperspectral remote sensing image data is becoming more and more widely.It is inseparable from the remote sensing classification technology,whether it is surface information acquisition,vegetation coverage monitoring,or map making.However,the high dimensionality and redundancy of hyperspectral remote sensing data seriously affect the development of remote sensing classification technology.Aimed at the challenges encountered in the classification of hyperspectral remote sensing images,this paper studies the data reduction,fusion of spatial feature and spectral feature and classification algorithms.The generation learning algorithm is to determine the category of a sample by summarizing the features distribution and the prior probabilities of the various categories.The generation learning algorithm solves some problems in the discriminant learning model,but there are many shortcomings in its classification.Therefore,the algorithm needs to be improved and perfected.In view of its shortcomings,this paper proposes an improved classification algorithm,which firstly reduces the dimension of the hyperspectral data,then uses the linear discriminant analysis to reduce the dimension of spectral and spatial features extracted,and conducts classification in this space finally.This improved classification algorithm not only improves the recognition ability of ground objects,but also accelerates the classification process.The algorithm in this paper is divided into two steps.The first step is fusion of spatial feature and spectral feature.The principal component analysis is used to extract the spectral characteristics of hyperspectral images.On the basis of that,morphological filter operators are used to extract spatial features of hyperspectral image.Then,the two features are superimposed and fused.Finally,the linear discriminant analysis is used to further reduce the fusion feature,so as to obtain spatial and spectral feature space.The principal component analysis can extract the main features of hyperspectral images,which can effectively reduce the dimension of hyperspectral data and improve the classification efficiency.The extraction of spatial features is beneficial to the differentiation of different classes.The secondary dimensionality reduction and fusion for spatial and spectral information by linear discriminant analysis further increases the gap between classes,and reduces the intra class gap.In the second step,Gauss Discriminant Analysis is used to classify fused feature space formed.The experimental results show that,compared with other classification algorithms,the proposed algorithm has some advantages in the classification.In this paper,an improved algorithm is proposed in view of the shortcomings of the algorithm itself through the discussion and study of the generation learning algorithms.In hyperspectral remote sensing image classification,the proposed algorithm not only makes full use of spatial and spectral information of the hyperspectral image,but also effectively improves the classification efficiency and classification accuracy of hyperspectral remote sensing image.Finally,the experimental results are given to evaluate the classification performance of the algorithm,which lays a foundation for future research.