Hyperspectral Remote Sensing Image Classification Based On Multiple Kernel Learning

Hyper Spectral Image(HSI)classification is an important for information retrieval from remote sensing data.Feature selection aims at selecting a feature subset that has the most discriminative information and preserve most of characteristics from original features in hyperspectral image classification.Feature selection is the key for dimension reduction and it directly influences image classification accuracy.Many studies show that traditional methods based on the empirical risk minimization principle have shown obvious deficiencies.Support Vector Machine based on Statistical Learning Theory(SLT)has a limited number of sample study,high-dimensional space,nonlinear,etc.,and becomes the hot topic because of its superiority of hyperspectral remote sensing image classification.This paper studied ensemble learning,information theory and computational intelligent in remote sensing image classification to improve the accuracy of classification.In this paper,feature selection and classification methods were studied in detail based on the statistical learning,and some improvements of classification algorithms(such as kernel function construction,parameters optimal of SVM,and multiple kernel ensemble)were successfully applied in hyperspectral remote sensing data.The main work we have done in this dissertation is summarized as follows:To address the problem of HSI classification we present a novel criterion by standard deviation,Kullback-Leibler distance,and correlation coefficient for feature selection.We propose an ensemble learning framework,which applies the boosting technique to learn multiple kernel classifiers for classification problems.Experiments are conducted on benchmark HSI classification data sets.The evaluation results show that the proposed approach can achieve superior accuracy and efficiency over state-of-the-art methods.In order to improve the efficiency of SVM and Multiple Kernel Learning(MKL),in this paper,the Kullback-Leibler kernel function is derived to develop SVM.The proposed method employs an improved ensemble learning framework,which applies Adaboost to learning multiple kernel-based classifier.In the experiment for hyperspectral remote sensing image classification,we employ feature selected through optimum index factor to classify the satellite image.We conduct experiments on hyperspectral image data set for validating the performance and evaluate various parameters of the proposed method in order to achieve a tradeoff between accuracy and efficiency.This paper proposes a new two-stage feature selection approach based on mutual information and Jeffries-Matusita measure.In first stage,a feature subset with minimal redundancy maximal relevance criteria is selected.In second stage,we select further a featuresubset from which obtained in first stage by maximizing Jeffries-Matusita distance.We examine empirical performance of proposed approach on benchmark hyperspectral classification data set.Experimental results demonstrate that the proposed method obtains better feature subsets and is more effective and efficient than classical methods.To improve the efficiency of multiple kernel boosting framework for classification,we optimize the SVM classifier design by searching for the most appropriate value of the parameters using Particle Swarm Optimization(PSO)with mutation mechanism.Many MKL methods often formulate the problem as an optimization task.To avoid solving the complicated optimization problem,this paper presents an ensemble learning framework which applies AdaBoost and stochastic approach to learning multiple kernel-based classifier for multi-class classification problem.We extensively examine the performance of our approach in comparison to some relevant algorithms on hyperspectral remote sensing image data set.Experimental results show that our method has a stable behavior and a noticeable accuracy for different data set.