Minshan Cui

Spectral–Spatial Preprocessing Using Multihypothesis Prediction for Noise-Robust Hyperspectral Image Classification

Spectral-spatial preprocessing using multihypothesis prediction is proposed for improving accuracy of hyperspectral image classification. Specifically, multiple spatially collocated pixel vectors are used as a hypothesis set from which a prediction for each pixel vector of interest is generated. Additionally, a spectral-band-partitioning strategy based on inter-band correlation coefficients is proposed to improve the representational power of the hypothesis set. To calculate an optimal linear combination of the hypothesis predictions, a distance-weighted Tikhonov regularization to an ill-posed least-squares optimization is used. The resulting predictions effectively integrate spectral and spatial information and thus are used during classification in lieu of the original pixel vectors. This processed hyperspectral image dataset has less intraclass variability and more spatial regularity as compared to the original dataset. Classification results for two hyperspectral image datasets demonstrate that the proposed method can enhance the classification accuracy of both maximum-likelihood and support vector classifiers, especially under small sample size constraints and noise corruption.

Class-Dependent Sparse Representation Classifier for Robust Hyperspectral Image Classification

Sparse representation of signals for classification is an active research area. Signals can potentially have a compact representation as a linear combination of atoms in an overcomplete dictionary. Based on this observation, a sparse-representation-based classification (SRC) has been proposed for robust face recognition and has gained popularity for various classification tasks. It relies on the underlying assumption that a test sample can be linearly represented by a small number of training samples from the same class. However, SRC implementations ignore the Euclidean distance relationship between samples when learning the sparse representation of a test sample in the given dictionary. To overcome this drawback, we propose an alternate formulation that we assert is better suited for classification tasks. Specifically, class-dependent sparse representation classifier (cdSRC) is proposed for hyperspectral image classification, which effectively combines the ideas of SRC and K-nearest neighbor classifier in a classwise manner to exploit both correlation and Euclidean distance relationship between test and training samples. Toward this goal, a unified class membership function is developed, which utilizes residual and Euclidean distance information simultaneously. Experimental results based on several real-world hyperspectral data sets have shown that cdSRC not only dramatically increases the classification performance over SRC but also outperforms other popular classifiers, such as support vector machine.

Locality Preserving Genetic Algorithms for Spatial-Spectral Hyperspectral Image Classification

Recent developments in remote sensing technologies have made hyperspectral imagery (HSI) readily available to detect and classify objects on the earth using pattern recognition techniques. Hyperspectral signatures are composed of densely sampled reflectance values over a wide range of the spectrum. Although most of the traditional approaches for HSI analysis entail per-pixel spectral classification, spatial-spectral exploitation of HSI has the potential to further improve the classification performance-particularly when there is unique class-specific textural information in the scene. Since the dimensionality of such remotely sensed imagery is often very large, especially in spatial-spectral feature domain, a large amount of training data is required to accurately model the classifier. In this paper, we propose a robust dimensionality reduction approach that effectively addresses this problem for hyperspectral imagery (HSI) analysis using spectral and spatial features. In particular, we propose a new dimensionality reduction algorithm, GA-LFDA where a Genetic Algorithm (GA) based feature selection and Local-Fishers Discriminant Analysis (LFDA) based feature projection are performed in a raw spectral-spatial feature space for effective dimensionality reduction. This is followed by a parametric Gaussian mixture model classifier. Classification results with experimental data show that our proposed method outperforms traditional dimensionality reduction and classification algorithms in challenging small training sample size and mixed pixel conditions.

Segmented Mixture-of-Gaussian Classification for Hyperspectral Image Analysis

The same high dimensionality of hyperspectral imagery that facilitates detection of subtle differences in spectral response due to differing chemical composition also hinders the deployment of traditional statistical pattern-classification procedures, particularly when relatively few training samples are available. Traditional approaches to addressing this issue, which typically employ dimensionality reduction based on either projection or feature selection, are at best suboptimal for hyperspectral classification tasks. A divide-and-conquer algorithm is proposed to exploit the high correlation between successive spectral bands and the resulting block-diagonal correlation structure to partition the hyperspectral space into approximately independent subspaces. Subsequently, dimensionality reduction based on a graph-theoretic locality-preserving discriminant analysis is combined with classification driven by Gaussian mixture models independently in each subspace. The locality-preserving discriminant analysis preserves the potentially multimodal statistical structure of the data, which the Gaussian mixture model classifier learns in the reduced-dimensional subspace. Experimental results demonstrate that the proposed system significantly outperforms traditional classification approaches, even when few training samples are employed.

Decision Fusion of Textural Features Derived From Polarimetric Data for Levee Assessment

Texture features derived from Synthetic Aperture Radar (SAR) imagery using grey level co-occurrence matrix (GLCM) can result in very high dimensional feature spaces. Although this high dimensional texture feature space can potentially provide relevant class-specific information for classification, it often also results in over-dimensionality and ill-conditioned statistical formulations. In this work, we propose a polarization channel based feature grouping followed by a multi-classifier decision fusion (MCDF) framework for a levee health monitoring system that seeks to detect landslides in earthen levees. In this system, texture features derived from the SAR imagery are partitioned into small groups according to different polarization channels. A multi-classifier system is then applied to each group to perform classification at the subspace level (i.e., a dedicated classifier for every subspace). Finally, a decision fusion system is employed to fuse decisions generated by each classifier to make a final classification decision (healthy levee versus landslide in this work). The resulting system can handle the high dimensionality of the problem very effectively, and only needs a few training samples for training and optimization.

Genetic algorithms and Linear Discriminant Analysis based dimensionality reduction for remotely sensed image analysis

Remotely sensed data (such as hyperspectral imagery) is typically associated with a large number of features, which makes classification challenging. Feature subset selection is an effective approach to alleviate the curse of dimensionality when the number of features contained in datasets is huge. Considering the merits of genetic algorithms (GA) in solving combinatorial problems, GA is becoming an increasingly popular tool for feature subset selection. Most algorithms presented in the literature using GA for feature subset selection use the training classification accuracy of a specific algorithm as the fitness function to optimize over the space of possible feature subsets. Such algorithms require a large amount of time to search for an optimal feature subset. In this paper, we will present a new approach called Genetic Algorithm based Linear Discriminant Analysis (GA-LDA) to extract features in which feature selection and feature extraction are performed simultaneously to alleviate over-dimensionality and result in a useful and robust feature space. Experimental results with classification tasks involving both hyperspectral imagery and SAR data indicate that GA-LDA can result in very low-dimensional feature subspaces yielding high classification accuracies.

Sparsity promoting dimensionality reduction for classification of high dimensional hyperspectral images

Sparse representation is an active research area in the signal processing and machine learning community in recent years. Recently, sparse representation classifier was proposed for challenging classification tasks - it entails representing a testing sample as a linear combination of all training samples which form an over-complete dictionary. In this paper, we demonstrate that for challenging high-dimensional classification tasks, appropriate dimensionality reduction is beneficial for sparse representation classifiers and its variants - especially when some features are redundant and/or lack discriminatory power. We propose a new dimensionality reduction algorithm to optimize the performance of greedy pursuit algorithms (required in sparse representation classifiers) by projecting the data into a space where the ratio of intra-class to inter class inner products are maximized. We demonstrate the superiority of the proposed method with standard hyperspectral imagery datasets - both in terms of improved classification accuracy and a speed-up in the run-time.

Morphologically Decoupled Structured Sparsity for Rotation-Invariant Hyperspectral Image Analysis

Hyperspectral imagery has emerged as a popular sensing modality for a variety of applications, and sparsity-based methods were shown to be very effective to deal with challenges coming from high dimensionality in most hyperspectral classification problems. In this paper, we challenge the conventional approach to hyperspectral classification that typically builds sparsity-based classifiers directly on spectral reflectance features or features derived directly from the data. We assert that hyperspectral image (HSI) processing can benefit very significantly by decoupling data into geometrically distinct components since the resulting decoupled components are much more suitable for sparse representation-based classifiers. Specifically, we apply morphological separation to decouple data into texture and cartoon-like components, which are sparsely represented using local discrete cosine bases and multiscale shearlets, respectively. In addition to providing a structured sparse representation, this approach allows us to build classifiers with invariance properties specific to each geometrically distinct component of the data. The experimental results using real-world HSI data sets demonstrate the efficacy of the proposed framework for classifying multichannel imagery under a variety of adverse conditions—in particular, small training sample size, additive noise, and rotational variabilities between training and test samples.

Angular Discriminant Analysis for Hyperspectral Image Classification

Hyperspectral imagery consists of hundreds or thousands of densely sampled spectral bands. The resulting spectral information can provide unique spectral “signatures” of different materials present in a scene, which makes hyperspectral imagery especially suitable for classification problems. To fully and effectively exploit discriminative information in such images, dimensionality reduction is typically undertaken as a preprocessing before classification. Different from traditional dimensionality reduction methods, we propose angular discriminant analysis (ADA), which seeks to find a subspace that best separates classes in an angular sense-specifically, one that minimizes the ratio of between-class inner product to within-class inner product of data samples on a unit hypersphere in the resulting subspace. In this paper, we also propose local angular discriminant analysis (LADA), which preserves the locality of data in the projected space through an affinity matrix, while angularly separating different class samples. ADA and LADA are particularly useful for classifiers that rely on angular distance, such as the cosine angle distance based nearest neighbor-based classifier and sparse representation-based classifier, in which the sparse representation coefficients are learned via orthogonal matching pursuit. We also show that ADA and LADA can be easily extended to their kernelized variants by invoking the kernel trick. Experimental results based on benchmarking hyperspectral datasets show that our proposed methods are greatly beneficial as a dimensionality reduction preprocessing to the popular classifiers.

Locality-preserving discriminant analysis for hyperspectral image classification using local spatial information

Locality-preserving projection as well as local Fisher discriminant analysis is applied for dimensionality reduction of hyperspectral imagery based on both spatial and spectral information. These techniques preserve the local geometric structure of hyperspectral data into a low-dimensional subspace wherein a Gaussian-mixture-model classifier is then considered. In the proposed classification system, local spatial information-which is expected to be more multimodal than strictly spectral features-is used. Results with experimental hyperspectral data demonstrate that this system outperforms traditional classification approaches.