Study On Dimensionality Reduction For Hyperspectral Images

Hyperspectral images with over a hundred of spectral bands,together with increasing spatial resolution,can be simultaneously acquired.In a hyperspectral image,each band includes rich spatial structure information,while each pixel contains many spectral features across a continuous range of narrow channels,from which arouses many real-world applications of hyperspectral images.Rich spectral information from hyperspectral images can aid in the classification and recognition of the ground objects.Currently,hyperspectral images classification has already been applied successfully in various fields.However,the high dimensions of hyperspectral images cause redundancy in information and bring some troubles while classifying precisely ground truth.In order to maintain the information needed for target detection and classification,and reduce the dimension of hyperspectral imagery,this paper carried out the following two research works through different dimensionality reduction methods:A classification method for hyperspectral images is proposed by fusing spectral-spatial-texture features.The principal component analysis is first used to reduce the dimension of hyperspectral image,and then the texture is extracted by gray-level co-occurrence matrix(GLCM)from the obtained principal components.At last,the classification results are obtained by using of PNN classifier.This paper proposes a hybrid feature selection strategy based on the Genetic Algorithm and the Novel Binary Particle Swarm Optimization(GA-NBPSO)to reduce the dimensionality of hyperspectral data while preserving the desired information for target detection and classification analysis.The proposed feature selection approach automatically chooses the most informative features combination.The parameters used in support vector machine(SVM)simultaneously are optimized,aiming at improving the performance of SVM.To show the validity of the proposal,Indian Pines(AVIRIS 92AV3C)and Pavia U(ROSIS-3)data set which are widely used to test the performance of feature selection techniques is chosen to feed the proposed method.To demonstrate the validity of the proposed method,two real hyperspectral images with the different spatial and spectral resolutions are fed to the proposal.Experimental results show that the proposed method can achieve higher classification accuracy than traditional methods.