Haoliang Yuan

Hyperspectral Image Classification Based on Three-Dimensional Scattering Wavelet Transform

Recent research has shown that utilizing the spectral-spatial information can improve the performance of hyperspectral image (HSI) classification. Since HSI is a 3-D cube datum, 3-D spatial filtering becomes a simple and effective method for extracting the spectral-spatial information. In this paper, we propose a 3-D scattering wavelet transform, which filters the HSI cube data with a cascade of wavelet decompositions, complex modulus, and local weighted averaging. The scattering feature can adequately capture the spectral-spatial information for classification. In the classification step, a support vector machine based on Gaussian kernel is used as a classifier due to its capability to deal with high-dimensional data. Our method is fully evaluated on four classic HSIs, i.e., Indian Pines, Pavia University, Botswana, and Kennedy Space Center. The classification results show that our method achieves as high as 94.46%, 99.30%, 97.57%, and 95.20% accuracies, respectively, when only 5% of the total samples per class is labeled.

Hyperspectral Image Classification Based on Regularized Sparse Representation

Sparsity-based models have been widely applied to hyperspectral image (HSI) classification. The class label of the test sample is determined by the minimum residual error based on the sparse vector, which is viewed as a pattern of original sample in the sparsity-based model. From the aspect of pattern classification, similar samples in the same class should have similar patterns. However, due to the independent sparse reconstruction process, the similarity among the sparse vectors of these similar samples is lost. To enforce such similarity information, a regularized sparse representation (RSR) model is proposed. First, a centralized quadratic constraint as the regularization term is incorporated into the objective function of ℓ1-norm sparse representation model. Second, RSR can be effectively solved by the feature-sign search algorithm. Experimental results demonstrate that RSR can achieve excellent classification performance.

Sparse Representation Based on Set-to-Set Distance for Hyperspectral Image Classification

Sparse representation-based classification model has been widely applied into hyperspectral image (HSI) classification. Its mechanism is based on the assumption that the nonzero coefficients in the sparse representation mainly lie in the correct class-dependent low-dimensional subspace. However, the high similarity of pixels between some different classes exists in the HSI, which makes the classification process very unstable. In this paper, we propose a sparse representation based on the set-to-set distance (SRSTSD) for HSI classification. Through utilizing the set-to-set distance, the spatial information is incorporated into the sparse representation-based model. Moreover, to further exploit the spatial structure of the pixel, we also propose a patch-based SRSTSD (PSRSTSD) model. Experimental results demonstrate that our proposed methods can achieve excellent classification performance.

Learning With Hypergraph for Hyperspectral Image Feature Extraction

It is known that hyperspectral image (HSI) classification is a high-dimension low-sample-size problem. To ease this problem, one natural idea is to take the feature extraction as a preprocessing. A graph embedding model is a classic family of feature extraction methods, which preserves certain statistical or geometric properties of the data set. However, the graph embedding model considers only the pairwise relationship between two vertices, which cannot represent the complex relationships of the data. Utilizing the spatial structure of HSI, in this letter, we propose a spatial hypergraph embedding model for feature extraction. Experimental results demonstrate that our method outperforms many existing feature extract methods for HSI classification.

Spectral-Spatial Classification of Hyperspectral Image Based on Discriminant Analysis

This paper proposes a spectral-spatial linear discriminant analysis (LDA) method for the hyperspectral image classification. A natural assumption is that similar samples have similar structure in the dimensionality reduced feature space. The proposed method uses a local scatter matrix from a small neighborhood as a regularizer incorporated into the objective function of LDA. Different from traditional LDA and its variants, our proposed method yields a self-adaptive projection matrix for dimension reduction, which improves the classification accuracy and avoids running out of memory. In order to consider the nonlinear case, this paper generalizes our linear version to its kernel version. Experimental results demonstrate that our proposed methods outperform several dimension reduction algorithms.

A Novel Sparsity-Based Framework Using Max Pooling Operation for Hyperspectral Image Classification

Various sparsity-based methods have been widely used in hyperspectral image (HSI) classification. To determine the class label of a test sample, traditional sparsity-based frameworks mainly use the sparse vectors to compute the residual error for classification. In this paper, a novel sparsity-based framework is proposed, which adopts the max pooling operation for HSI classification. Compared with the traditional sparsity-based frameworks using residual error, sparse vectors in our proposed framework are utilized to generate the feature vectors using max pooling operation. Experimental results demonstrate that our proposed framework can achieve the state-of-the-art classification performance.

Spectral–Spatial Shared Linear Regression for Hyperspectral Image Classification

Classification of the pixels in hyperspectral image (HSI) is an important task and has been popularly applied in many practical applications. Its major challenge is the high-dimensional small-sized problem. To deal with this problem, lots of subspace learning (SL) methods are developed to reduce the dimension of the pixels while preserving the important discriminant information. Motivated by ridge linear regression (RLR) framework for SL, we propose a spectral–spatial shared linear regression method (SSSLR) for extracting the feature representation. Comparing with RLR, our proposed SSSLR has the following two advantages. First, we utilize a convex set to explore the spatial structure for computing the linear projection matrix. Second, we utilize a shared structure learning model, which is formed by original data space and a hidden feature space, to learn a more discriminant linear projection matrix for classification. To optimize our proposed method, an efficient iterative algorithm is proposed. Experimental results on two popular HSI data sets, i.e., Indian Pines and Salinas demonstrate that our proposed methods outperform many SL methods.Classification of the pixels in hyperspectral image (HSI) is an important task and has been popularly applied in many practical applications. Its major challenge is the high-dimensional small-sized problem. To deal with this problem, lots of subspace learning (SL) methods are developed to reduce the dimension of the pixels while preserving the important discriminant information. Motivated by ridge linear regression (RLR) framework for SL, we propose a spectral-spatial shared linear regression method (SSSLR) for extracting the feature representation. Comparing with RLR, our proposed SSSLR has the following two advantages. First, we utilize a convex set to explore the spatial structure for computing the linear projection matrix. Second, we utilize a shared structure learning model, which is formed by original data space and a hidden feature space, to learn a more discriminant linear projection matrix for classification. To optimize our proposed method, an efficient iterative algorithm is proposed. Experimental results on two popular HSI data sets, i.e., Indian Pines and Salinas demonstrate that our proposed methods outperform many SL methods.

Spectral-spatial linear discriminant analysis for hyperspectral image classification

We propose a spectral-spatial linear discriminant analysis method (LDA) for dimensionality reduction in hyperspectral image. The proposed method uses a local scatter of the small neighborhood as a regularizer to incorporate into the objective function of the LDA. The intrinsic idea is to design an optimal linear transformation that makes these samples among the neighborhood approximate the local mean in the low-dimensional feature space while simultaneously preserving the original property of LDA. Experimental results based on both adequate training samples and inadequate training samples demonstrate that the proposed method outperforms several traditional dimensionality reduction methods.

Spectral-spatial hyperspectral image destriping using low-rank representation and Huber-Markov random fields

This paper presents a novel spectral-spatial destriping method for hyperspectral images. The ubiquitous striping noise in hyperspectral images might degrade the quality of the imagery and bring difficulties in hyperspectral data processing. Although numerous methods have been proposed for striping noise reduction recently, most of them fail to consider the spectral correlation and spatial information of the hyperspectral images simultaneously. In order to remedy this drawback, the proposed method integrates the spectral and spatial information to remove the striping noise in the hyperspectral images. To this end, firstly, the low-rank representation (LRR) is used to take advantage of the spectral information. Then, the spatial information is included using a Huber-Markov random field (MRF) prior model, which is convex and can well preserve the edge and texture information while removing the noise. The experimental results on simulated and real hyperspectral data sets demonstrate the effectiveness of the proposed method.

Semi-supervised graph-based retargeted least squares regression

A semi-supervised graph-based retargeted least squares regression model is proposed for multicategory classification.Our aim is to utilize a graph regularization to restrict the regression labels of ReLSR, such that similar samples should have similar regression labels.Linear squares regression and graph construction are unified into a same framework to achieve an overall optimum. In this paper, we propose a semi-supervised graph-based retargeted least squares regression model (SSGReLSR) for multicategory classification. The main motivation behind SSGReLSR is to utilize a graph regularization to restrict the regression labels of ReLSR, such that similar samples should have similar regression labels. However, in SSGReLSR, constructing the graph structure and learning the regression matrix are two independent processes, which cant guarantee an overall optimum. To overcome this shortage of SSGReLSR, we also propose a semi-supervised graph learning retargeted least squares regression model (SSGLReLSR), where linear squares regression and graph construction are unified into a same framework to achieve an overall optimum. To optimize our proposed SSGLReLSR, an efficient iteration algorithm is proposed. Extensive experiments results confirm the effectiveness of our proposed methods.