Research On Feature Classification Based On Support Vector Machine Model For Hyperspectral Remote Sensing Image

The spectral information of the features in the hyperspectral remote sensing image is very rich.Make good use of hyperspectral remote sensing images to classify features is one of the most important applications of hyperspectral remote sensing images.This application has been widely used in national production or scientific research fields such as environmental monitoring,geology and mineral analysis,land use evaluation,and agricultural remote sensing.At present,the use of machine learning methods to classify features of hyperspectral remote sensing images is a hot topic.The traditional classification method originally used for multi-spectral imagery is no longer suitable for classifying hyperspectral remote sensing images with high dimensionality and large data volume.The theoretical assumptions of the traditional classification method are based on the fact that the sample approaches infinity.However,in the actual classification problem,the acquisition of samples is not simple,and it is difficult to obtain a sufficient number of samples to satisfy the traditional statistical theory.Support vector machine is a learning method with excellent two-class classification performance.It is based on statistical learning theory.It is feasible to extend the support vector machine to the multi-class classification problem of hyperspectral remote sensing image classification,and there are already many well-researched results.In order to obtain higher accuracy of classification results in small sample cases,this paper proposes a semi-supervised support vector machine model for hyperspectral remote sensing image classification based on statistical learning theory.In this paper,the K-means++ clustering algorithm is used to cluster the unlabeled sample points in the case of a small number of training samples per class,and the structural information in the unlabeled sample points is used to assist in the construction of the SVM.Complete the classification of hyperspectral remote sensing images.The main work and innovations of this paper are as follows:(1)This paper introduces the background of remote sensing research and the development and prospects of hyperspectral remote sensing applications,analyzes the status of current methods for classification of hyperspectral remote sensing images,introduces the development of support vector machines and points out its good performance in classification applications.(2)This paper analyzes the acquisition of hyperspectral remote sensing image data,the expression of data and its data characteristics are analyzed.The principle and main classification strategies of hyperspectral remote sensing image data classification are introduced.(3)This paper introduces the concepts of optimal classification plane and kernel function involved in support vector machine through the form of formula calculus,and summarizes several characteristics of support vector machine.(4)This paper reviews the principles of two types of classification of support vector machines,and introduces two guiding ideas for extending support vector machines to multi-class classifications and their corresponding implementation methods.(5)In order to obtain higher image classification accuracy under smallsample conditions,an optimized semi-supervised support vector machine model for hyperspectral remote sensing image classification is proposed.This method uses the structural information of unlabeled sample points to modify the support vector machine.The design experiment proves that the optimized support vector machine model proposed in this paper can obtain higher classification accuracy when performing hyperspectral remote sensing images in the case of small samples.