Spectral-spatial Classification Of Hyperspectral Images Research Articles

A Hybrid Multi-scale Spatial Filtering and Minimum Spanning Forest for Spectral–Spatial Hyperspectral Image Classification

Integration of spatial and spectral information is an effective way in improving classification accuracy. In this article a new framework, based on multi-scale spatial weighted mean filtering (MSWMF) and minimum spanning forest, is proposed for the spectral–spatial classification of hyperspectral images. In the proposed framework, at first the image is smoothed by MSWMF and then the first eight principal components are extracted. Using support vector machine, at each scale of MSWMF, a classification map is produced in order to generate a marker map in the next step. Then, the minimum spanning forest is built on the marker map. Finally, in order to create a final classification map, all the classification maps of each scale are merged with a majority vote rule. The experimental results of the hyper-spectral images indicate that the suggested framework enhances the classification accuracy, in comparison with previously classification techniques. So, it is interesting for hyperspectral images classification.

Ideal Kernel-Based Multiple Kernel Learning for Spectral-Spatial Classification of Hyperspectral Image

Using multiple types of features can effectively improve the classification accuracy of hyperspectral image (HSI). Multiple kernel learning (MKL) provides a flexible framework to fuse different features in a very natural way. In this letter, a novel MKL algorithm is proposed to integrate multiple types of features [i.e., principal components of original spectrum, multistructure morphological profiles (MPs), and multiattribute MPs] for HSI classification. The basic kernels are constructed with each feature subset separately, and the weights of basic kernels are determined by solving an optimization problem with the objective of the ideal kernel. Then, linear programming (LP) and signal sparse representation are adopted to solve the optimization problem, thus leading to two variants of the proposed algorithm, ideal kernel MKL (IKMKL)-LP and IKMKL-sparse, respectively. Experiments carried out on real hyperspectral data show that the proposed algorithms outperform several state-of-the-art MKL algorithms and reveal the capability of ranking the relevant features.

A Hybrid DBN and CRF Model for Spectral-Spatial Classification of Hyperspectral Images

Hyperspectral image classification plays an important role in remote sensing image analysis. Recent techniques have attempted to investigate the capabilities of deep learning approaches to tackle the hyperspectral image classification. This work shows how to further improve the hyperspectral image classification through using both a deep representation and contextual information. To implement this objective, this work proposes a new Conditional Random Field (CRF) model (named DBN-CRF) with the potentials defined over the deep features produced by a Deep Belief Network (DBN). The newly formulated DBN-CRF model takes advantage of the strength of DBNs in learning a good representation and the ability of CRFs to model contextual (spatial) information in both the observations and labels. Within a piecewise training framework, an efficient training method is proposed to train the whole DBN-CRF model end-to-end. This means that the parameters in DBN and CRF can be jointly trained and thus the proposed method can fully use the strength of both DBN and CRF. Moreover, in the proposed training method, the end-to-end training can be implemented with a standard back-propagation algorithm, avoiding the repeated inference usually involved in CRF training and thus is computationally efficient. Experiments on real-world hyperspectral data show that our method outperforms the most recent approaches in hyperspectral image classification.

An adaptive framework for spectral-spatial classification based on a combination of pixel-based and object-based scenarios

Remotely sensed image analysis using spectral-spatial information plays a key role in modern remote sensing applications. This article presents a new semi-automatic framework for spectral-spatial classification of hyperspectral images. The proposed framework benefits from a combination of pixel-based and object-based classification scenarios in which the main parameters are adaptively tuned. In order to reduce the complexity of the method, an unsupervised band selection technique is used as well. Meanwhile, the wavelet thresholding is applied in order to smooth the selected bands. The classification results after applying the proposed method to well-known standard hyperspectral datasets are better than those of the most of the other state-of-the-art approaches. As an example, the overall classification accuracy achieved by applying the proposed semi-automatic spectral-spatial classification framework to the Salinas dataset is more than 99% for 10% training samples per class. Moreover, the vital parameters are adaptively set in our approach.

Spectral-spatial classification of hyperspectral image based on discriminant sparsity preserving embedding

The last few years have witnessed the success of sparse representation in hyperspectral image classification. However, the high computational complexity brings some worries to its applications. In this paper, a novel sparse representation based feature extraction algorithm, called discriminant sparsity preserving embedding (DSPE), is proposed by constructing a sparse graph and applying it to the graph-embedding framework. The proposed algorithm encodes supervised information mainly in stage of sparse graph construction, in which only the training samples in the same class are used to calculated the reconstructive coefficients during sparse reconstruction. An approach combining l1-norm and l2-norm is applied to solve the reconstruction weights, where l1-norm ensures the sparsity of the graph weights, l2-norm shrinks the weight coefficients to make the construction more stable and alleviate the reconstruction errors possibly caused by small-size training samples. On the premise of satisfied classification results, here a spectral-spatial classification strategy which takes spatial information into consideration is used to evaluate the efficiency of the proposed algorithm. Experiments on the Indian Pines and Pavia University hyperspectral image datasets demonstrate the superiority of the proposed algorithm.

Discriminative Low-Rank Gabor Filtering for Spectral–Spatial Hyperspectral Image Classification

Spectral–spatial classification of remotely sensed hyperspectral images has attracted a lot of attention in recent years. Although Gabor filtering has been used for feature extraction from hyperspectral images, its capacity to extract relevant information from both the spectral and the spatial domains of the image has not been fully explored yet. In this paper, we present a new discriminative low-rank Gabor filtering (DLRGF) method for spectral–spatial hyperspectral image classification. A main innovation of the proposed approach is that our implementation is accomplished by decomposing the standard 3-D spectral–spatial Gabor filter into eight subfilters, which correspond to different combinations of low-pass and bandpass single-rank filters. Then, we show that only one of the subfilters (i.e., the one that performs low-pass spatial filtering and bandpass spectral filtering) is actually appropriate to extract suitable features based on the characteristics of hyperspectral images. This allows us to perform spectral–spatial classification in a highly discriminative and computationally efficient way, by significantly decreasing the computational complexity (from cubic to linear order) compared with the 3-D spectral–spatial Gabor filter. In order to theoretically prove the discriminative ability of the selected subfilter, we derive an overall classification risk bound to evaluate the discriminating abilities of the features provided by the different subfilters. Our experimental results, conducted using different hyperspectral images, indicate that the proposed DLRGF method exhibits significant improvements in terms of classification accuracy and computational performance when compared with the 3-D spectral–spatial Gabor filter and other state-of-the-art spectral–spatial classification methods.

Class-Specific Random Forest With Cross-Correlation Constraints for Spectral–Spatial Hyperspectral Image Classification

A class-specific random forest (RF) model with cross-correlation constraints is developed for the spectral-spatial hyperspectral image (HSI) classification. The novelties of this letter are as follows: 1) normalization of the spectral feature vector by using cross correlation in the stochastic process and proposal of a spectral-spatial hybrid feature extraction based on the cross-correlation analysis; 2) establishment of an RF classifier model by using class-specific trees (CSTs); and 3) exploration of the performance of the proposed method by comparing with its several traditional classification methods on two real HSI data sets. Further research on the effects of parameter setup, such as the number of CSTs, spectral constraint scale, and size of the spatial neighbor, is discussed in terms of classification accuracy. Experimental results show that the performance of the proposed method is better than that of the traditional methods.

Spectral-spatial classification of hyperspectral imagery using a dual-channel convolutional neural network

ABSTRACTIn this article, a novel dual-channel convolutional neural network (DC-CNN) framework is proposed for accurate spectral-spatial classification of hyperspectral image (HSI). In this framework, one-dimensional CNN is utilized to automatically extract the hierarchical spectral features and two-dimensional CNN is applied to extract the hierarchical space-related features, and then a softmax regression classifier is used to combine the spectral and spatial features together and predict classification results eventually. To overcome the problem of the limited available training samples in HSIs, we propose a simple data augmentation method which is efficient and effective for improving HSI classification accuracy. For comparison and validation, we test the proposed method along with three other deep-learning-based HSI classification methods on two real-world HSI data sets. Experimental results demonstrate that our DC-CNN-based method outperforms the state-of-the-art methods by a considerable margin.

Classification of visible and infrared hyperspectral images based on image segmentation and edge-preserving filtering

The classification of hyperspectral images with a few labeled samples is a major challenge which is difficult to meet unless some spatial characteristics can be exploited. In this study, we proposed a novel spectral-spatial hyperspectral image classification method that exploited spatial autocorrelation of hyperspectral images. First, image segmentation is performed on the hyperspectral image to assign each pixel to a homogeneous region. Second, the visible and infrared bands of hyperspectral image are partitioned into multiple subsets of adjacent bands, and each subset is merged into one band. Recursive edge-preserving filtering is performed on each merged band which utilizes the spectral information of neighborhood pixels. Third, the resulting spectral and spatial feature band set is classified using the SVM classifier. Finally, bilateral filtering is performed to remove “salt-and-pepper” noise in the classification result. To preserve the spatial structure of hyperspectral image, edge-preserving filtering is applied independently before and after the classification process. Experimental results on different hyperspectral images prove that the proposed spectral-spatial classification approach is robust and offers more classification accuracy than state-of-the-art methods when the number of labeled samples is small.

Integration of 3-dimensional discrete wavelet transform and Markov random field for hyperspectral image classification

Hyperspectral image (HSI) classification is one of the fundamental tasks in HSI analysis. Recently, many approaches have been extensively studied to improve the classification performance, among which integrating the spatial information underlying HSIs is a simple yet effective way. However, most of the current approaches haven't fully exploited the spatial information prior. They usually consider this prior either in the step of extracting spatial feature before classification or in the step of post-processing label map after classification, while don't integratively employ the prior in both steps, which thus leaves a room for further enhancing their performance. In this paper, we propose a novel spectral-spatial HSI classification method, which fully utilizes the spatial information in both steps. Firstly, the spatial feature is extracted by applying the 3-dimensional discrete wavelet transform (3D-DWT). Secondly, the local spatial correlation of neighboring pixels is modeled using Markov random field (MRF) based on the probabilistic classification map obtained by applying probabilistic support vector machine (SVM) to the extracted 3D-DWT feature in the first step, and then a maximum a posterior (MAP) classification problem can be formulated in a Bayesian perspective. Finally, α-Expansion min-cut-based optimization algorithm is adopted to solve this MAP problem efficiently. Experimental results on two benchmark HSIs show that the proposed method achieves a significant performance gain beyond state-of-the-art methods.

Attribute profile based feature space discriminant analysis for spectral-spatial classification of hyperspectral images

An initial feature reduction is necessary to reduce the data dimensionality before applying attribute filters to hyperspectral images. Unsupervised methods such as principal component analysis are not good choices for classification purposes. On the other hand, supervised methods such as linear discriminant analysis have no good efficiency in small sample size situations. In this article, we propose to extract features using feature space discriminant analysis (FSDA), which has been recently proposed in 2015. FSDA only uses the spectral information and ignores the spatial information. In this paper, we overcome this indigenous disadvantage of FSDA and develop FSDA for spectral-spatial classification of hyperspectral images. Our proposed method, called attribute profile based feature space discriminant analysis (APFSDA), extracts spatial features with high class discrimination ability and as little as redundant information. The experimental results on several real hyperspectral images show the superiority of APFSDA compared to some state-of-the-art spectral-spatial classification methods.

Local histogram and discriminative learning-based hyperspectral data classification

ABSTRACTA spectral–spatial hyperspectral image classification method is proposed in this article. The proposed method is local histogram and discriminative based classification (LHD). It is implemented in two steps. In the first step, a projection matrix is obtained by maximizing the class discrimination and preserving the local structure of data using both the spectral and spatial neighbours. In the second step, the contextual features are extracted from the local regions using the histogram as a nonparametric statistical estimate. Finally, with incorporating spectral and spatial features, the classification map is obtained. The experimental results demonstrate that the proposed method is superior to some state-of-the-art alternatives.

Spectral-Spatial Classification of Hyperspectral Images Using Joint Bilateral Filter and Graph Cut Based Model

Hyperspectral image classification can be achieved by modeling an energy minimization problem on a graph of image pixels. In this paper, an effective spectral-spatial classification method for hyperspectral images based on joint bilateral filtering (JBF) and graph cut segmentation is proposed. In this method, a novel technique for labeling regions obtained by the spectral-spatial segmentation process is presented. Our method includes the following steps. First, the probabilistic support vector machines (SVM) classifier is used to estimate probabilities belonging to each information class. Second, an extended JBF is employed to perform image smoothing on the probability maps. By using our JBF process, salt-and-pepper classification noise in homogeneous regions can be effectively smoothed out while object boundaries in the original image are better preserved as well. Third, a sequence of modified bi-labeling graph cut models is constructed for each information class to extract the desirable object belonging to the corresponding class from the smoothed probability maps. Finally, a classification map is achieved by merging the segmentation maps obtained in the last step using a simple and effective rule. Experimental results based on three benchmark airborne hyperspectral datasets with different resolutions and contexts demonstrate that our method can achieve 8.56%–13.68% higher overall accuracies than the pixel-wise SVM classifier. The performance of our method was further compared to several classical hyperspectral image classification methods using objective quantitative measures and a visual qualitative evaluation.

Bilayer Elastic Net Regression Model for Supervised Spectral-Spatial Hyperspectral Image Classification

In this paper, we propose a novel bilayer elastic net (ELN $^2$ ) regression model for hyperspectral image (HSI) classification, exploiting the spectral-spatial information. The proposed model is designed to address the special problematic characteristics of HSI, namely, high dimensionality of hyperspectral pixels, limited labeled samples, and spatial variability of spectral signatures. To alleviate these problems, by exploiting the spectral and spatial information, the proposed model features in the following two components: 1) spectral-only elastic net regression in the first layer and 2) spatial contextual driven elastic net regularization in the second layer. In the first layer, to encourage a grouping effect and feature selection, we use multinomial logistic regression (MLR) model penalized by the ELN to optimize the spectral-only classifier parameters for the initial HSI classification. In the second layer, spatial Markov-random-field-based gradient profiles are incorporated into the ELN penalty over the hidden marginal probability of the posterior distribution to encourage the spatial smoothness. Furthermore, the true labels of training samples are fixed as an additional constraint in the second-layer optimizing model to further improve the classification accuracy. Moreover, an effective algorithm named as ELN $^2$ _RegMLR is developed by coupling the path-wise coordinate descent algorithm, and variable splitting and augmented Lagrangian approach to solve the proposed model. Experimental results on several popular datasets show that the proposed method outperforms many state-of the-art classifiers in terms of the overall accuracy, average accuracy, statistic coefficient kappa, and visual classification map quality.

Tensor Block-Sparsity Based Representation for Spectral-Spatial Hyperspectral Image Classification

Recently, sparse representation has yielded successful results in hyperspectral image (HSI) classification. In the sparse representation-based classifiers (SRCs), a more discriminative representation that preserves the spectral-spatial information can be exploited by treating the HSI as a whole entity. Based on this observation, a tensor block-sparsity based representation method is proposed for spectral-spatial classification of HSI in this paper. Unlike traditional vector/matrix-based SRCs, the proposed method consists of tensor block-sparsity based dictionary learning and class-dependent block sparse representation. By naturally regarding the HSI cube as a third-order tensor, small local patches centered at the training samples are extracted from the HSI to maintain the structural information. All the patches are then partitioned into a number of groups, on which a dictionary learning model is constructed with a tensor block-sparsity constraint. A test sample is also expressed as a small local patch and the block sparse representation is then performed in a class-wise manner to take advantage of the class label information. Finally, the category of the test sample is determined by using the minimal residual. Experimental results of two real-world HSIs show that our proposed method greatly improves the classification performance of SRC.

Parallel probabilistic relaxation labelling based on Markov random fields for spectral-spatial hyperspectral image classification

ABSTRACTThe large volume of data and computational complexity of algorithms limit the application of hyperspectral image classification to real-time operations. This work addresses the use of different parallel processing techniques to speed up the Markov random field (MRF)-based method to perform spectral-spatial classification of hyperspectral imagery. The Metropolis relaxation labelling approach is modified to take advantage of multi-core central processing units (CPUs) and to adapt it to massively parallel processing systems like graphics processing units (GPUs). The experiments on different hyperspectral data sets revealed that the implementation approach has a huge impact on the execution time of the algorithm. The results demonstrated that the modified MRF algorithm produced classification accuracy similar to conventional methods with greatly improved computational performance. With modern multi-core CPUs, good computational speed-up can be achieved even without additional hardware support. The CPU-GPU hybrid framework rendered the otherwise computationally expensive approach suitable for time-constrained applications.

A Spectral-Spatial Classification of Hyperspectral Images Based on the Algebraic Multigrid Method and Hierarchical Segmentation Algorithm

The algebraic multigrid (AMG) method is used to solve linear systems of equations on a series of progressively coarser grids and has recently attracted significant attention for image segmentation due to its high efficiency and robustness. In this paper, a novel spectral-spatial classification method for hyperspectral images based on the AMG method and hierarchical segmentation (HSEG) algorithm is proposed. Our method consists of the following steps. First, the AMG method is applied to hyperspectral imagery to construct a multigrid structure of fine-to-coarse grids based on the anisotropic diffusion partial differential equation (PDE). The vertices in the multigrid structure are then considered as the initial seeds (markers) for growing regions and are clustered to obtain a sequence of segmentation results. In the next step, a maximum vote decision rule is employed to combine the pixel-wise classification map and the segmentation maps. Finally, a final classification map is produced by choosing the optimal grid level to extract representative spectra. Experiments based on three different types of real hyperspectral datasets with different resolutions and contexts demonstrate that our method can obtain 3.84%–13.81% higher overall accuracies than the SVM classifier. The performance of our method was further compared to several marker-based spectral-spatial classification methods using objective quantitative measures and a visual qualitative evaluation.

Graph-based semi-supervised learning for spectral-spatial hyperspectral image classification

Spectral and spatial regularized semi-supervised learning is investigated for classification of hyperspectral images. In spectral regularizer, sum of minimum distance (SMD) is introduced to determine graph adjacency relationships and local manifold learning (LML) is employed for edge weighting. The resulted SMD_LML regularizer is able to constrain the prediction vectors to preserve the local geometry of each neighborhood. In spatial regularizer, spatial neighbors with similar spectra are enforced to have similar predictions. The two regularizers capture different relations of data points and play complementary roles. By combining SMD_LML regularizer and spatial regularizer in graph based semi-supervised learning, the local properties of both spectral neighborhood and spatial neighborhood can be preserved in the prediction domain. Experiments with AVIRIS and ROSIS hyperspectral images demonstrated that SMD can produce more accurate adjacency relations than the other two popular distance measurements. The classification with SMD_LML and spatial regularizer achieved significant improvements than several spectral and spatial based methods.

A Discontinuity Preserving Relaxation Scheme for Spectral–Spatial Hyperspectral Image Classification

In remote sensing image processing, relaxation is defined as a method that uses the local relationship among neighboring pixels to correct spectral or spatial distortions. In recent years, relaxation methods have shown great success in classification of remotely sensed data. Relaxation, as a preprocessing step, can reduce noise and improve the class separability in the spectral domain. On the other hand, relaxation (as a postprocessing approach) works on the label image or class probabilities obtained from pixelwise classifiers. In this work, we develop a discontinuity preserving relaxation strategy, which can be used for postprocessing of class probability estimates, as well as preprocessing of the original hyperspectral image. The newly proposed method is an iterative relaxation procedure, which exploits spatial information in such a way that it considers discontinuities existing in the data cube. Our experimental results indicate that the proposed methodology leads to state-of-the-art classification results when combined with probabilistic classifiers for several widely used hyperspectral data sets, even when very limited training samples are available.

Spectral–Spatial Hyperspectral Image Classification Based on KNN

Fusion of spectral and spatial information is an effective way in improving the accuracy of hyperspectral image classification. In this paper, a novel spectral–spatial hyperspectral image classification method based on K nearest neighbor (KNN) is proposed, which consists of the following steps. First, the support vector machine is adopted to obtain the initial classification probability maps which reflect the probability that each hyperspectral pixel belongs to different classes. Then, the obtained pixel-wise probability maps are refined with the proposed KNN filtering algorithm that is based on matching and averaging nonlocal neighborhoods. The proposed method does not need sophisticated segmentation and optimization strategies while still being able to make full use of the nonlocal principle of real images by using KNN, and thus, providing competitive classification with fast computation. Experiments performed on two real hyperspectral data sets show that the classification results obtained by the proposed method are comparable to several recently proposed hyperspectral image classification methods.