Based On Dictionary Learning And Feature Fusion For Hyperspectral Image Classification

As the development of hyperspectral remote sensors,hundreds of images at different wavelength channels can be collected for the same area on the surface of the Earth,which can provide spectral information that is several ten times larger than the multispectral image.However,it also increases the difficulty of the classification of remote sensing images by providing a large amount of spectral information,in which the major challenge is the large number of spectral features and limited training samples(due to the difficulty and expense of manual labeling).So how to exploit the wealth of information available to effectively tackle the small sample size(3S)problem is an important issue and has received much attention in hyperspectral image analysis.Texture feature is one of the important factors to distinguish different objects,helping to reduce the spectral variability and adverse effects of objects with the same spectrum.Gabor filter is one of the powerful tools that can capture the image texture direction,size and other internal structure information.Two dimensional Gabor filter has been widely used in the field of texture analysis,texture segmentation and successfully applied to hyperspectral image classification.Discriminative dictionary learning aims to learn a dictionary from training samples in order to improve the discriminative ability of their coding vectors.Gabor wavelets have recently been successfully applied for hyperspectral image classification due to their ability to extract joint spatial and spectrum information.Due to the high discriminative power of Gabor features,an efficient method,called Gabor feature based Support Vector Guided Dictionary Learning(GSVGDL),has been proposed in this paper for hyperspectral image classification.After Gabor features have been extracted from the hyperspectral image,the augmented Gabor feature matrix is used to construct the initial dictionary.The dictionary learning model formulates the discrimination term as the weighted summation of the squared distances between all pairs of coding vectors,which can greatly improve the discriminative ability of the dictionary.The structure of the dictionary and the corresponding linear classifier are obtained simultaneously by dictionary learning.Since the spatial distribution of Hyperspectral imagery generally exhibits high regularity and local continuity,spatial texture information should be introduced to improve the classification accuracy of hyperspectral image.The extended morphological profiles(EMP)have been created from the raw hyperspectral image,which has proven to be effective and robust of reflecting the spatial structural features of hyperspectral data.Meanwhile,due to the three-dimensional(3D)Gabor wavelets have been introduced to exploit the joint spectral-spatial features of hyperspectral image.In this paper,in order to combine the advantages of the EMP operator and Gabor wavelet transform together,an extended morphological profile-based Gabor wavelets has been proposed for hyperspectral image classification.Compared with the other algorithms,such as the Sparse representation-based classification(SRC),support vector machine(SVM)and so on.The experimental results have shown that the proposed method can achieve better classification performance than other method.In the end of the paper,we have summarized the disadvantages and the advantage of the proposed method and pointed out the direction of the further word in this field.