Imaging spectrometers measure electromagnetic energy scattered in their instantaneous field view in hundreds or thousands of spectral channels with higher spectral resolution than multispectral cameras. Imaging spectrometers are therefore often referred to as hyperspectral cameras (HSCs). Higher spectral resolution enables material identification via spectroscopic analysis, which facilitates countless applications that require identifying materials in scenarios unsuitable for classical spectroscopic analysis. Due to low spatial resolution of HSCs, microscopic material mixing, and multiple scattering, spectra measured by HSCs are mixtures of spectra of materials in a scene. Thus, accurate estimation requires unmixing. Pixels are assumed to be mixtures of a few materials, called endmembers. Unmixing involves estimating all or some of: the number of endmembers, their spectral signatures, and their abundances at each pixel. Unmixing is a challenging, ill-posed inverse problem because of model inaccuracies, observation noise, environmental conditions, endmember variability, and data set size. Researchers have devised and investigated many models searching for robust, stable, tractable, and accurate unmixing algorithms. This paper presents an overview of unmixing methods from the time of Keshava and Mustard's unmixing tutorial to the present. Mixing models are first discussed. Signal-subspace, geometrical, statistical, sparsity-based, and spatial-contextual unmixing algorithms are described. Mathematical problems and potential solutions are described. Algorithm characteristics are illustrated experimentally.

/pdf/hyperspectral-unmixing-overview-geometrical-statistical-and-2lb11fr9l4.pdf

Hyperspectral Unmixing Overview: Geometrical, Statistical, and Sparse Regression-Based Approaches

Geophysical applications of radar interferometry to measure changes in the Earth's surface have exploded in the early 1990s. This new geodetic technique calculates the interference pattern caused by the difference in phase between two images acquired by a spaceborne synthetic aperture radar at two distinct times. The resulting interferogram is a contour map of the change in distance between the ground and the radar instrument. These maps provide an unsurpassed spatial sampling density (∼100 pixels km−2), a competitive precision (∼1 cm), and a useful observation cadence (1 pass month−1). They record movements in the crust, perturbations in the atmosphere, dielectric modifications in the soil, and relief in the topography. They are also sensitive to technical effects, such as relative variations in the radar's trajectory or variations in its frequency standard. We describe how all these phenomena contribute to an interferogram. Then a practical summary explains the techniques for calculating and manipulating interferograms from various radar instruments, including the four satellites currently in orbit: ERS-1, ERS-2, JERS-1, and RADARSAT. The next chapter suggests some guidelines for interpreting an interferogram as a geophysical measurement: respecting the limits of the technique, assessing its uncertainty, recognizing artifacts, and discriminating different types of signal. We then review the geophysical applications published to date, most of which study deformation related to earthquakes, volcanoes, and glaciers using ERS-1 data. We also show examples of monitoring natural hazards and environmental alterations related to landslides, subsidence, and agriculture. In addition, we consider subtler geophysical signals such as postseismic relaxation, tidal loading of coastal areas, and interseismic strain accumulation. We conclude with our perspectives on the future of radar interferometry. The objective of the review is for the reader to develop the physical understanding necessary to calculate an interferogram and the geophysical intuition necessary to interpret it.

/pdf/radar-interferometry-and-its-application-to-changes-in-the-2xi2itqotn.pdf

Radar interferometry and its application to changes in the Earth's surface

Due to the advantages of deep learning, in this paper, a regularized deep feature extraction (FE) method is presented for hyperspectral image (HSI) classification using a convolutional neural network (CNN). The proposed approach employs several convolutional and pooling layers to extract deep features from HSIs, which are nonlinear, discriminant, and invariant. These features are useful for image classification and target detection. Furthermore, in order to address the common issue of imbalance between high dimensionality and limited availability of training samples for the classification of HSI, a few strategies such as L2 regularization and dropout are investigated to avoid overfitting in class data modeling. More importantly, we propose a 3-D CNN-based FE model with combined regularization to extract effective spectral-spatial features of hyperspectral imagery. Finally, in order to further improve the performance, a virtual sample enhanced method is proposed. The proposed approaches are carried out on three widely used hyperspectral data sets: Indian Pines, University of Pavia, and Kennedy Space Center. The obtained results reveal that the proposed models with sparse constraints provide competitive results to state-of-the-art methods. In addition, the proposed deep FE opens a new window for further research.

/pdf/deep-feature-extraction-and-classification-of-hyperspectral-4p21mou7rr.pdf

Deep Feature Extraction and Classification of Hyperspectral Images Based on Convolutional Neural Networks

Imaging spectrometers measure electromagnetic energy scattered in their instantaneous field view in hundreds or thousands of spectral channels with higher spectral resolution than multispectral cameras. Imaging spectrometers are therefore often referred to as hyperspectral cameras (HSCs). Higher spectral resolution enables material identification via spectroscopic analysis, which facilitates countless applications that require identifying materials in scenarios unsuitable for classical spectroscopic analysis. Due to low spatial resolution of HSCs, microscopic material mixing, and multiple scattering, spectra measured by HSCs are mixtures of spectra of materials in a scene. Thus, accurate estimation requires unmixing. Pixels are assumed to be mixtures of a few materials, called endmembers. Unmixing involves estimating all or some of: the number of endmembers, their spectral signatures, and their abundances at each pixel. Unmixing is a challenging, ill-posed inverse problem because of model inaccuracies, observation noise, environmental conditions, endmember variability, and data set size. Researchers have devised and investigated many models searching for robust, stable, tractable, and accurate unmixing algorithms. This paper presents an overview of unmixing methods from the time of Keshava and Mustard's unmixing tutorial [1] to the present. Mixing models are first discussed. Signal-subspace, geometrical, statistical, sparsity-based, and spatial-contextual unmixing algorithms are described. Mathematical problems and potential solutions are described. Algorithm characteristics are illustrated experimentally.

Signal subspace identification is a crucial first step in many hyperspectral processing algorithms such as target detection, change detection, classification, and unmixing. The identification of this subspace enables a correct dimensionality reduction, yielding gains in algorithm performance and complexity and in data storage. This paper introduces a new minimum mean square error-based approach to infer the signal subspace in hyperspectral imagery. The method, which is termed hyperspectral signal identification by minimum error, is eigen decomposition based, unsupervised, and fully automatic (i.e., it does not depend on any tuning parameters). It first estimates the signal and noise correlation matrices and then selects the subset of eigenvalues that best represents the signal subspace in the least squared error sense. State-of-the-art performance of the proposed method is illustrated by using simulated and real hyperspectral images.

/pdf/hyperspectral-subspace-identification-4cvaygfbm7.pdf

Hyperspectral Subspace Identification

The authors propose a new method for unsupervised classification of terrain types and man-made objects using polarimetric synthetic aperture radar (SAR) data. This technique is a combination of the unsupervised classification based on polarimetric target decomposition, S.R. Cloude et al. (1997), and the maximum likelihood classifier based on the complex Wishart distribution for the polarimetric covariance matrix, J.S. Lee et al. (1994). The authors use Cloude and Pottier's method to initially classify the polarimetric SAR image. The initial classification map defines training sets for classification based on the Wishart distribution. The classified results are then used to define training sets for the next iteration. Significant improvement has been observed in iteration. The iteration ends when the number of pixels switching classes becomes smaller than a predetermined number or when other criteria are met. The authors observed that the class centers in the entropy-alpha plane are shifted by each iteration. The final class centers in the entropy-alpha plane are useful for class identification by the scattering mechanism associated with each zone. The advantages of this method are the automated classification, and the interpretation of each class based on scattering mechanism. The effectiveness of this algorithm is demonstrated using a JPL/AIRSAR polarimetric SAR image.

Unsupervised classification using polarimetric decomposition and the complex Wishart classifier

The Lee sigma filter was developed in 1983 based on the simple concept of two-sigma probability, and it was reasonably effective in speckle filtering. However, deficiencies were discovered in producing biased estimation and in blurring and depressing strong reflected targets. The advancement of synthetic aperture radar (SAR) technology with high-resolution data of large dimensions demands better and efficient speckle filtering algorithms. In this paper, we extend and improve the Lee sigma filter by eliminating these deficiencies. The bias problem is solved by redefining the sigma range based on the speckle probability density functions. To mitigate the problems of blurring and depressing strong reflective scatterers, a target signature preservation technique is developed. In addition, we incorporate the minimum-mean-square-error estimator for adaptive speckle reduction. Simulated SAR data are used to quantitatively evaluate the characteristics of this improved sigma filter and to validate its effectiveness. The proposed algorithm is applied to spaceborne and airborne SAR data to demonstrate its overall speckle filtering characteristics as compared with other algorithms. This improved sigma filter remains simple in concept and is computationally efficient but without the deficiencies of the original Lee sigma filter.

Improved Sigma Filter for Speckle Filtering of SAR Imagery

A new algorithm for exploiting the nonlinear structure of hyperspectral imagery is developed and compared against the de facto standard of linear mixing. This new approach seeks a manifold coordinate system that preserves geodesic distances in the high-dimensional hyperspectral data space. Algorithms for deriving manifold coordinates, such as isometric mapping (ISOMAP), have been developed for other applications. ISOMAP guarantees a globally optimal solution, but is computationally practical only for small datasets because of computational and memory requirements. Here, we develop a hybrid technique to circumvent ISOMAP's computational cost. We divide the scene into a set of smaller tiles. The manifolds derived from the individual tiles are then aligned and stitched together to recomplete the scene. Several alignment methods are discussed. This hybrid approach exploits the fact that ISOMAP guarantees a globally optimal solution for each tile and the presumed similarity of the manifold structures derived from different tiles. Using land-cover classification of hyperspectral imagery in the Virginia Coast Reserve as a test case, we show that the new manifold representation provides better separation of spectrally similar classes than one of the standard linear mixing models. Additionally, we demonstrate that this technique provides a natural data compression scheme, which dramatically reduces the number of components needed to model hyperspectral data when compared with traditional methods such as the minimum noise fraction transform.

Exploiting manifold geometry in hyperspectral imagery

This paper addresses the noise filtering problem for synthetic aperture radar (SAR) interferometric phase images. The phase noise is characterized by an additive noise model. The model is verified with an L-band shuttle imaging radar (SIR)-C interferogram. An adaptive filtering algorithm based on this noise model is developed. It emphasizes filtering noise adaptively according to the local noise level and filtering along fringes using directionally dependent windows. This algorithm is effective, especially for the tightly packed fringes of X-band interferometry. Using simulated and SIR-C/X-SAR repeat-pass generated interferograms, the effectiveness of this filter is demonstrated by its capabilities in residue reduction, adaptive noise filtering, and its ability to filter areas with high fringe rates. In addition, a scheme of incorporating this filtering algorithm in iterative phase unwrapping using a least-squares method is proposed.

A new technique for noise filtering of SAR interferometric phase images

This paper addresses the problem of polarimetric SAR (POLSAR) data correction for changes in radar cross sections, which are caused by azimuth slopes. Most radiometric slope corrections remove slope effects to account for the effective scattering pixel area. However, few studies address the slope effect on the radar cross section as a function of polarization states. The authors propose two approaches to compensate for this azimuth slope effect for POLSAR data. In the first approach, the digital elevation model (DEM), obtained from interferometric SAR or from other means, is used to estimate orientation angles, and in the second approach, orientation angles are derived directly from the POLSAR data. They have developed three new methods to implement this second approach. One is based on the right-right and left-left circular polarizations, and the other two are based on Cloude's and Huynelt's decompositions. The results are compared with the original polarization signature method using the DEM-generated orientation angles as the reference. POLSAR data is then compensated using the derived orientation angles. Significant changes were observed in many elements of the coherency matrix. The compensated POLSAR data should improve the accuracy of geophysical parameter estimation techniques. NASA/JPL AIRSAR data for an area in central California is used for illustration.

Thomas L. Ainsworth

Papers

Unsupervised classification using polarimetric decomposition and the complex Wishart classifier

Improved Sigma Filter for Speckle Filtering of SAR Imagery

Exploiting manifold geometry in hyperspectral imagery

A new technique for noise filtering of SAR interferometric phase images

Polarimetric SAR data compensation for terrain azimuth slope variation