Hyperspectral imaging is a powerful technology that is plagued by large dimensionality. Our study explores a way to combat that hindrance via noncontiguous and contiguous (simpler to realize sensor) band grouping for dimensionality reduction. Our approach is different in the respect that it is flexible and it follows a well-studied process of visual clustering in high-dimensional spaces. Specifically, we extend the improved visual assessment of cluster tendency and clustering in ordered dissimilarity data unsupervised clustering algorithms for supervised hyperspectral learning. In addition, we propose a way to extract diverse features via the use of different proximity metrics (ways to measure the similarity between bands) and kernel functions. The discovered features are fused with ℓ ∞ -norm multiple kernel learning. Experiments are conducted on two benchmark data sets and our results are compared to related work. These data sets indicate that contiguous or not is application specific, but heterogeneous features and kernels usually lead to performance gain.
Collaborative representation has been a popular classifier for hyperspectral image classification because it can offer excellent classification accuracy with a closed-form solution. Collaborative representation can be implemented using a dictionary with training samples of all-classes, or using class-specific sub-dictionaries. In either case, a testing pixel is assigned to the class whose training samples offer the minimum representation residual. The Collaborative Representation Optimized Classifier with Tikhonov regularization (CROCT) was developed to combine these two types of collaborative representations to achieve the balance for optimized performance. The class-specific collaborative representation involves inverse operation of matrices constructed from class-specific samples, and the all-class version requires inversion operation of the matrix constructed from all samples. In this paper, we propose a low-complexity CROCT to avoid redundant operations in all-class and class-specific collaborative representations. It can further reduce the computational cost of CROCT while maintaining its excellent classification performance.
Levee slides may result in catastrophic damage to the region of failure. Remote sensing data, such as synthetic aperture radar (SAR) images, can be useful in levee monitoring. Because of the long length of a levee, the image size may become too large to use computationally expensive methods for quick levee monitoring, so time-efficient approaches are preferred. The popular support vector machine classifier does not work well on the original three polarized SAR magnitude bands without spatial feature extraction. Gray-level co-occurrence matrix is one of the most common methods for extracting textural information from gray-scale images, but it may not be practically useful for a big data in terms of calculation time. In this study, very efficient feature extraction methods with spatial low-pass filtering are proposed, including a weighted average filter and a majority filter in conjunction with a nonlinear band normalization process. Experimental results demonstrated that these filters can provide comparable results with much lower computational cost.
Spectral mixture analysis is one of the major techniques in hyperspectral remote sensing image analysis. Endmember extraction for spectral mixture analysis is a necessary step when endmember information is unknown. If endmembers are assumed to be pure pixels present in an image scene, endmember extraction is to search the most distinct pixels. Popular algorithms using the criteria of simplex volume maximization (e.g., N-FINDR) and spectral signature similarity (e.g., Vertex Component Analysis) belong to this type. N-FINDR is a parallel-searching method, where all the endmembers are determined simultaneously. VCA is a sequential-searching method, finding endmembers one after another, which can greatly save computational cost. In this paper, we focus on VCA-based endmember extraction. In particular, we propose a new searching approach that makes the extracted endmembers more distinct. Real data experiments show that it can improve the quality of extracted endmembers.
The widely used principal component analysis (PCA) is implemented in nonlinear by an auto-associative neural network. Compared to other nonlinear versions, such as kernel PCA, such a nonlinear PCA has explicit encoding and decoding processes, and the data can be transformed back to the original space. Its data compression performance is similar to that of PCA, but data analysis performance such as target detection is much better. To expedite its training process, graphics computing unit (GPU)-based parallel computing is applied.
Band selection is a common technique to reduce the dimensionality of hyperspectral imagery. When the desired object information is known, the reduction process can be achieved by selecting the bands that contain the most object information. It is expected that these selected bands can offer an overall satisfactory detection and classification performance. In this paper, we propose a new particle swarm optimization (PSO) based supervised band-selection algorithm that uses the known class signatures only without examining the original bands or the need of class training samples. Thus, this method requires much less computing time than other traditional methods. However, the PSO process itself may introduce additional computation cost. To tackle this problem, we propose parallel implementations via emerging general-purpose graphics processing units that can provide satisfactory results in speedup when compared to the cluster-based parallel implementation.
In this paper, the original N-FINDR algorithm is implemented in four different versions: 1) parallel mode using the
data after dimensionality reduction; 2) parallel mode using all the original bands; 3) sequential mode using the data
after dimensionality reduction; 4) sequential mode using all the original bands. We will analyze the performance
discrepancy among these versions. Based on experimental evaluation and comparison, instructive recommendations in
implementation strategy for practical applications are provided.
This research study presents a novel global index based on harmonic mean theory to jointly evaluate the performance of pan-sharpening algorithms without using a reference image. The harmonic mean of relative spatial information gain and relative spectral information preservation provides a unique global index to compare the performance of different methods. The presented index also facilitates in assigning relevance to either the spectral or spatial quality of an image. The information divergence between the multispectral (MS) bands at lower resolutions and the pan-sharpened image provides a measure of the spectral fidelity and mean-shift. Mutual information between the original pan and the synthetic pan images generated from the MS and pan-sharpened images is used to calculate the relative gain. The relative gain helps to quantify the amount of spatial information injected by the method. A trend comparison of the presented approach with other quality indices using well-known pan-sharpening methods on high resolution and medium resolution datasets reveals that the new index can be used to evaluate the quality of pan-sharpened images at the resolution of the pan image without the availability of a reference image.
KEYWORDS: Distributed computing, Data modeling, Knowledge management, Computing systems, Data archive systems, Data integration, Satellites, Sensors, Standards development, Systems modeling
The Global Earth Observation System of Systems (GEOSS) is a distributed system of systems built on current
international cooperation efforts among existing Earth observing and processing systems. The goal is to formulate an
end-to-end process that enables the collection and distribution of accurate, reliable Earth Observation data, information,
products, and services to both suppliers and consumers worldwide. One of the critical components in the development
of such systems is the ability to obtain seamless access of data across geopolitical boundaries. In order to gain support
and willingness to participate by countries around the world in such an endeavor, it is necessary to devise mechanisms
whereby the data and the intellectual capital is protected through procedures that implement the policies specific to a
country. Earth Observations (EO) are obtained from a multitude of sources and requires coordination among different
agencies and user groups to come to a shared understanding on a set of concepts involved in a domain. It is envisaged
that the data and information in a GEOSS context will be unprecedented and the current data archiving and delivery
methods need to be transformed into one that allows realization of seamless interoperability. Thus, EO data integration
is dependent on the resolution of conflicts arising from a variety of areas. Modularization is inevitable in distributed
environments to facilitate flexible and efficient reuse of existing ontologies. Therefore, we propose a framework for
modular ontologies based knowledge management approach for GEOSS and present methods to enable efficient
reasoning in such systems.
A pixel in multispectral images is highly correlated with the neighboring pixels both spatially and spectrally. Hence, data
transformation is performed before performing pan-sharpening. Principal component analysis (PCA) has been a popular
choice for spectral transformation of low resolution multispectral images. Current PCA-based pan-sharpening methods
make an assumption that the first principal component (PC) of high variance is an ideal choice for replacing or injecting
it with high spatial details from the high-resolution histogram-matched panchromatic (Pan) image. However, this paper,
using the statistical measures on the datasets, shows that the low-resolution first PC component is not always an ideal
choice for substitution. This paper presents a new method to improve the quality of the resultant images that are obtained
using the PCA-based pan-sharpening methods. This approach is based on adaptively selecting the PC component
required to be replaced or injected with high spatial details. The pan-sharpened image obtained by the proposed method
is evaluated using well-known quality indexes. Results show that the proposed method increases the quality of the
resultant fused images when compared to the standard approach.
Pan-sharpened images are useful in a wide variety of applications. Hence, giving quantitative importance to image
quality, depending on the nature of target application, may be required to yield maximum benefit. Current techniques for
joint evaluation of spatial and spectral quality without reference do not allow to quantitatively associate importance to
the image quality. This work proposes a novel global index based on harmonic mean theory to jointly evaluate the
performance of pan-sharpening algorithms without using a reference image. The harmonic mean of relative spatial
information gain and relative spectral information preservation provides a unique global index to compare the
performance of different algorithms. The proposed approach also facilitates in assigning relevance to either the spectral
or spatial quality of an image. The information divergence between the MS bands at lower resolutions and the pansharpened
image provides a measure of the spectral fidelity and mean-shift. Mutual information between the original pan
and synthetic pan images generated from the MS and pan-sharpen images is used to calculate the relative gain. The
relative gain helps to quantify the amount of spatial information injected by the algorithm. A trend comparison of the
proposed approach with other quality indexes using well-known pan-sharpening algorithms on high resolution (IKONOS
and Quickbird) and medium resolution (LandSat7 ETM+) datasets reveals that the new index can be used to evaluate the
quality of pan-sharpened image at the resolution of the pan image without the availability of a reference image.
In general, reflectance and spatial patterns characterize geospatial data. Current semantic-enabled framework image
retrieval systems for geospatial data extract primitive features based on color, texture (Spatial Gray Level Dependency -
SGLD matrices), and shape from the segmented homogenous region. However, the form of extracting textural
information is computationally expensive. The state-of-the-art image mining system for multimedia image archives uses
the wavelet transform for feature extraction to quickly and efficiently capture color and texture information. Since an
image consists of three bands, color information is captured by converting the RGB space into HSV space. Thus, a new
approach is required to capture the complete reflectance pattern, an important characteristic of geospatial data. This work
proposes a new method to perform fast coarse image segmentation using descriptors obtained by combining the 2Dwavelet
transform along the spatial axis and the Fourier transform along the spectral axis to capture color and texture
information for segmentation. These features are later on used for region-based retrieval in Earth observation data
archives. Compared to traditional techniques, result shows that the proposed method provides good retrieval accuracy in
terms of F-measure for land cover classes.
Pansharpening is a pixel-level fusion technique used to increase the spatial resolution of the multispectral image using spatial information from the high-resolution panchromatic image, while preserving the spectral information in the multispectral image. Various pansharpening algorithms are available in the literature, and some have been incorporated in commercial remote sensing software packages such as ERDAS Imagine® and ENVI®. The demand for high spatial and spectral resolutions imagery in applications like change analysis, environmental monitoring, cartography, and geology is increasing rapidly. Pansharpening is used extensively to generate images with high spatial and spectral resolution. The suitability of these images for various applications depends on the spectral and spatial quality of the pansharpened images. Hence, the evaluation of the spectral and spatial quality of the pansharpened images using objective quality metrics is a necessity. In this work, quantitative metrics for evaluating the quality of pansharpened images are presented. A performance comparison, using the intensity-hue-saturation (IHS)-based sharpening, Brovey sharpening, principal component analysis (PCA)-based sharpening, and a wavelet-based sharpening method, is made to better quantify their accuracies.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.