Parallel unmixing of remotely sensed hyperspectral images on commodity graphics processing units

Hyperspectral imaging instruments are capable of collecting hundreds of images, corresponding to different wavelength channels, for the same area on the surface of the Earth. One of the main problems in the analysis of hyperspectral data cubes is the presence of mixed pixels, which arise when the spatial resolution of the sensor is not enough to separate spectrally distinct materials. Hyperspectral unmixing is one of the most popular techniques to analyze hyperspectral data. It comprises two stages: (i) automatic identification of pure spectral signatures (endmembers) and (ii) estimation of the fractional abundance of each endmember in each pixel. The spectral unmixing process is quite expensive in computational terms, mainly due to the extremely high dimensionality of hyperspectral data cubes. Although this process maps nicely to high performance systems such as clusters of computers, these systems are generally expensive and difficult to adapt to real‐time data processing requirements introduced by several applications, such as wildland fire tracking, biological threat detection, monitoring of oil spills, and other types of chemical contamination. In this paper, we develop an implementation of the full hyperspectral unmixing chain on commodity graphics processing units (GPUs). The proposed methodology has been implemented, using the CUDA (compute device unified architecture), and tested on three different GPU architectures: NVidia Tesla C1060, NVidia GeForce GTX 275, and NVidia GeForce 9800 GX2, achieving near real‐time unmixing performance in some configurations tested when analyzing two different hyperspectral images, collected over the World Trade Center complex in New York City and the Cuprite mining district in Nevada. Copyright © 2011 John Wiley & Sons, Ltd.     

A quantitative and comparative analysis of different preprocessing implementations of DPSO: a robust endmember extraction algorithm

Linear spectral unmixing is a very important technique in hyperspectral image analysis. It contains two main steps. First, it finds spectrally unique signatures of pure ground components (called endmembers); second, it estimates their corresponding fractional abundances in each pixel. Recently, a discrete particle swarm optimization (DPSO) algorithm was introduced to accurately extract endmembers with high optimal performance. However, because of its limited feasible solution space, DPSO necessarily needs a small amount of candidate endmembers before extraction. Consequently, how to provide a suitable candidate endmember set, which has not been analyzed yet, is a critical issue in using DPSO for unmixing problem. In this study, three representative pure pixel-based methods, pixel purity index, vertex component analysis (VCA), and N-FINDR, are quantitatively compared to provide candidate endmembers for DPSO. The experiments with synthetic and real hyperspectral images indicate that VCA is the most reliable preprocessing implementation for DPSO. Further, it can be concluded that DPSO with the proposed preprocessing implementations given in this paper is robust for endmember extraction.     

A novel FPGA-based architecture for the estimation of the virtual dimensionality in remotely sensed hyperspectral images

A challenging problem in spectral unmixing is how to determine the number of endmembers in a given scene. One of the most popular ways to determine the number of endmembers is by estimating the virtual dimensionality (VD) of the hyperspectral image using the well-known Harsanyi–Farrand–Chang (HFC) method. Due to the complexity and high dimensionality of hyperspectral scenes, this task is computationally expensive. Reconfigurable field-programmable gate arrays (FPGAs) are promising platforms that allow hardware/software codesign and the potential to provide powerful onboard computing capabilities and flexibility at the same time. In this paper, we present the first FPGA design for the HFC-VD algorithm. The proposed method has been implemented on a Virtex-7 XC7VX690T FPGA and tested using real hyperspectral data collected by NASA’s Airborne Visible Infra-Red Imaging Spectrometer over the Cuprite mining district in Nevada and the World Trade Center in New York. Experimental results demonstrate that our hardware version of the HFC-VD algorithm can significantly outperform an equivalent software version, which makes our reconfigurable system appealing for onboard hyperspectral data processing. Most important, our implementation exhibits real-time performance with regard to the time that the hyperspectral instrument takes to collect the image data.     

On Endmember Identification in Hyperspectral Images Without Pure Pixels: A Comparison of Algorithms

Hyperspectral imaging is an active area of research in Earth and planetary observation. One of the most important techniques for analyzing hyperspectral images is spectral unmixing, in which mixed pixels (resulting from insufficient spatial resolution of the imaging sensor) are decomposed into a collection of spectrally pure constituent spectra, called endmembers weighted by their correspondent fractions, or abundances. Over the last years, several algorithms have been developed for automatic endmember extraction. Many of them assume that the images contain at least one pure spectral signature for each distinct material. However, this assumption is usually not valid due to spatial resolution, mixing phenomena, and other considerations. A?recent trend in the hyperspectral imaging community is to design endmember identification algorithms which do not assume the presence of pure pixels. Despite the proliferation of this kind of algorithms, many of which are based on minimum enclosing simplex concepts, a rigorous quantitative and comparative assessment is not yet available. In this paper, we provide a comparative analysis of endmember extraction algorithms without the pure pixel assumption. In our experiments we use synthetic hyperspectral data sets (constructed using fractals) and real hyperspectral scenes collected by NASA’s Jet Propulsion Laboratory.     

Kernel sparse representation for hyperspectral unmixing based on high mutual coherence spectral library

ABSTRACT

Sparse regression is now a popular method for hyperspectral unmixing relying on a prior spectral library. However, it is limited by the high mutual coherence spectral library which contains high similarity atoms. In order to improve the accuracy of sparse unmixing with a high mutual coherence spectral library, a new algorithm based on kernel sparse representation unmixing model with total variation constraint is proposed in this paper. By constructing an appropriate kernel function to expand similarity measure scale, library atoms and hyperspectral data are mapped to kernel space where sparse regression algorithms are then applied. Experiments conducted with both simulated and real hyperspectral data sets indicate that the proposed algorithm effectively improves the unmixing performance when using a high mutual coherence spectral library because of its ability to precisely extract endmembers in hyperspectral images. Compared with other state-of-the-art algorithms, the proposed algorithm obtains low reconstruction errors in pixels with different mixed degree.     

Implementation of real-time constrained linear discriminant analysis to remote sensing image classification

In this paper, we investigate the practical implementation issues of the real-time constrained linear discriminant analysis (CLDA) approach for remotely sensed image classification. Specifically, two issues are to be resolved: (1) what is the best implementation scheme that yields lowest chip design complexity with comparable classification performance, and (2) how to extend CLDA algorithm for multispectral image classification. Two limitations about data dimensionality have to be relaxed. One is in real-time hyperspectral image classification, where the number of linearly independent pixels received for classification must be larger than the data dimensionality (i.e., the number of spectral bands) in order to generate a non-singular sample correlation matrix R for the classifier, and relaxing this limitation can help to resolve the aforementioned first issue. The other is in multispectral image classification, where the number of classes to be classified cannot be greater than the data dimensionality, and relaxing this limitation can help to resolve the aforementioned second issue. The former can be solved by introducing a pseudo inverse initiate of sample correlation matrix for R^-1 adaptation, and the latter is taken care of by expanding the data dimensionality via the operation of band multiplication. Experiments on classification performance using these modifications are conducted to demonstrate their feasibility. All these investigations lead to a detailed ASIC chip design scheme for the real-time CLDA algorithm suitable to both hyperspectral and multispectral images. The proposed techniques to resolving these two dimensionality limitations are instructive to the real-time implementation of several popular detection and classification approaches in remote sensing image exploitation.     

马尔可夫随机场的空间相关模型在非负矩阵分解线性解混中的应用

针对基于非负矩阵分解（NMF）的高光谱解混存在的初始化与"局部极小"等问题,提出一种基于马尔可夫随机场（MRF）的空间相关约束NMF线性解混算法（MRF-NMF）。首先,通过基于最小误差的高光谱信号识别（HySime）法估算端元数量,同时利用顶点成分分析（VCA）和全约束最小二乘法（FCLS）初始化端元矩阵与丰度矩阵;其次,利用MRF模型建立描述地物空间分布规律的能量函数,以此描述地物分布的空间相关特征;最后,将基于MRF的空间相关约束函数与NMF标准目标函数以交替迭代的形式参与解混,得出高光谱数据的端元信息与丰度分解结果。理论分析和真实数据实验结果表明,在高光谱数据空间相关程度较低的情况下,相比最小体积约束的NMF （MVC-NMF）、分段平滑和稀疏约束的NMF （PSNMFSC）和交互投影子梯度非负矩阵分解（APS-NMF）三种参考算法,所提算法的端元分解精度仍分别提高了7.82%、12.4%和10.1%,其丰度分解精度仍分别提高了8.34%、12.6%和9.87%。MRF-NMF能够弥补NMF对于空间相关特征描述能力的不足,减小解混结果中地物的空间能量分布误差。     

Unmixing-based content retrieval system for remotely sensed hyperspectral imagery on GPUs

This paper presents a new unmixing-based retrieval system for remotely sensed hyperspectral imagery. The need for this kind of system is justified by the exponential growth in the volume and number of remotely sensed data sets from the surface of the Earth. This is particularly the case for hyperspectral images, which comprise hundreds of spectral bands at different (almost contiguous) wavelength channels. To deal with the high computational cost of extracting the spectral information needed to catalog new hyperspectral images in our system, we resort to efficient implementations of spectral unmixing algorithms on commodity graphics processing units (GPUs). Spectral unmixing is a very popular approach for interpreting hyperspectral data with sub-pixel precision. This paper particularly focuses on the design of the proposed framework as a web service, as well as on the efficient implementation of the system on GPUs. In addition, we present a comparison of spectral unmixing algorithms available in the system on both CPU and GPU architectures.     

单形体体积最小化的差分进化光谱解混算法

目的 光谱解混是高光谱遥感图像处理的核心技术。当图像不满足纯像元假设条件时,传统算法难以适用,基于(单形体)体积最小化方法提供了一种有效的解决途径。然而这是一个复杂的约束最优化问题,更由于图像噪声等不确定性因素的存在,导致算法容易陷入局部解。方法 引入一种群智能优化技术-差分进化算法(DE),借助其较强的全局搜索能力以及优越的处理高维度问题的能力,并通过对问题编码,提出了一种体积最小化的差分进化(VolMin-DE)光谱解混算法。结果 模拟数据和真实数据实验的结果表明,与现有算法相比,该算法在15端元时精度(光谱角距离)可提高7.8%,当端元数目少于15个时,其精度普遍可以提高15%以上,特别是10端元时精度可以提高41.3%;在20~50 dB的噪声范围内,精度变化在1.9~3.2(单位:角度)之间,传统算法在2.2~3.5之间,表明该算法具有相对较好的噪声鲁棒性。结论 本文算法适用于具有纯像元以及不存在纯像元(建议最大纯度不低于0.8)这两种情况的高光谱遥感图像,并可在原始光谱维度进行光谱解混,从而避免降维所带来的累计误差,因此具有更好的适应范围和应用前景。     

结合空间光谱预处理和约束非负矩阵分解的高光谱图像混合像元分解

目的 混合像元问题在高光谱遥感图像处理分析中普遍存在,非负矩阵分解的方法被引入到高光谱图像解混中。本文提出结合空间光谱预处理和约束非负矩阵分解的混合像元分解流程。方法 结合空间光谱预处理的约束非负矩阵分解,如最小体积约束、流行约束等,通过加入邻域的空间和光谱信息进行预处理获得更优的预选端元,从而对非负矩阵分解的解混结果进行优化。结果 在5组不同信噪比的模拟数据实验中,空间预处理（SPP）和空间光谱预处理（SSPP）均能够有效提高约束非负矩阵分解（最小体积约束的非负矩阵分解和图正则非负矩阵分解）的解混结果,其中SPP在不同信噪比的情况下都能优化约束非负矩阵分解的结果,而SSPP在低信噪比的情况下,预处理效果更佳。利用美国内华达州Cuprite矿区数据进行真实数据实验,SPP提高了约束非负矩阵分解的解混精度,而SSPP在复杂场景下,解混精度更佳。模拟数据和真实数据的实验均表明,空间光谱预处理能够有效地提高约束非负矩阵分解的解混精度,特别是对于信噪比较低的情况下,融合空间和光谱信息对噪声有很好的鲁棒性。结论 本文对约束非负矩阵分解的解混算法添加空间光谱预处理,利用高光谱遥感数据的空间和光谱信息,优化预选端元,加入空间光谱预处理的非负矩阵解混实验流程,在复杂场景情况下,对噪声具有较好的鲁棒性。     

Parallel processing of remotely sensed hyperspectral imagery: full‐pixel versus mixed‐pixel classification

The rapid development of space and computer technologies allows for the possibility to store huge amounts of remotely sensed image data, collected using airborne and satellite instruments. In particular, NASA is continuously gathering high‐dimensional image data with Earth observing hyperspectral sensors such as the Jet Propulsion Laboratory's airborne visible–infrared imaging spectrometer (AVIRIS), which measures reflected radiation in hundreds of narrow spectral bands at different wavelength channels for the same area on the surface of the Earth. The development of fast techniques for transforming massive amounts of hyperspectral data into scientific understanding is critical for space‐based Earth science and planetary exploration. Despite the growing interest in hyperspectral imaging research, only a few efforts have been devoted to the design of parallel implementations in the literature, and detailed comparisons of standardized parallel hyperspectral algorithms are currently unavailable. This paper compares several existing and new parallel processing techniques for pure and mixed‐pixel classification in hyperspectral imagery. The distinction of pure versus mixed‐pixel analysis is linked to the considered application domain, and results from the very rich spectral information available from hyperspectral instruments. In some cases, such information allows image analysts to overcome the constraints imposed by limited spatial resolution. In most cases, however, the spectral bands collected by hyperspectral instruments have high statistical correlation, and efficient parallel techniques are required to reduce the dimensionality of the data while retaining the spectral information that allows for the separation of the classes. In order to address this issue, this paper also develops a new parallel feature extraction algorithm that integrates the spatial and spectral information. The proposed technique is evaluated (from the viewpoint of both classification accuracy and parallel performance) and compared with other parallel techniques for dimensionality reduction and classification in the context of three representative application case studies: urban characterization, land‐cover classification in agriculture, and mapping of geological features, using AVIRIS data sets with detailed ground‐truth. Parallel performance is assessed using Thunderhead, a massively parallel Beowulf cluster at NASA's Goddard Space Flight Center. The detailed cross‐validation of parallel algorithms conducted in this work may specifically help image analysts in selection of parallel algorithms for specific applications. Copyright © 2008 John Wiley & Sons, Ltd.     

高光谱图像非线性解混方法的研究进展

由于空间分辨率的限制,高光谱遥感图像中存在大量混合像元,对混合像元的解混是实现地物精确分类和识别的前提。与传统的线性解混方法相比,非线性解混方法在寻找组成混合像元的端元以及每个端元的丰度时具有较高的精度。分析了光谱非线性混合的原理,总结了近年来提出的非线性解混算法,重点对双线性模型、神经网络、基于核函数的非线性解混算法以及基于流形学习的非线性解混算法进行了介绍和分析。最后总结了混合像元非线性解混未来发展的趋势。     

Estimating the coverage of coral reef benthic communities from airborne hyperspectral remote sensing data: multiple discriminant function analysis and linear spectral unmixing

A staged approach for the application of linear spectral unmixing techniques to airborne hyperspectral remote sensing data of reef communities of the Al Wajh Barrier, Red Sea, is presented. Quantification of the percentage composition of four different reef components (live coral, dead coral, macroalgae and carbonate sand) contained within the ground sampling distance associated with an individual pixel is demonstrated. In the first stage, multiple discriminant function analysis is applied to spectra collected in situ to define an optimal subset combination of derivative and raw image wavebands for discriminating reef benthos. In the second phase, unmixing is applied to a similarly reduced subset of pre-processed image data to accurately determine the relative abundance of the reef benthos (R ² > 0.7 for all four components). The result of a phased approach is an increased signal-to-noise ratio for solution of the linear functions and reduction of processing burdens associated with image unmixing.     

盲信号分离技术在高光谱混合像元分解中的应用

盲信号分离(BSS)是现代信号处理的一种前沿基础技术,近年来在高光谱混合像元分解领域展示了很好的应用前景。通过比较系统地介绍独立成分分析(ICA)、非负矩阵分解(NMF)、复杂度分析(CA)和稀疏成分分析(SCA) 4种BSS方法的基本原理、基本概念及数学模型,重点阐述4种方法在高光谱混合像元分解中的应用现状及各自的优缺点,旨在进一步探讨BSS技术应用于高光谱混合像元分解面临的挑战与存在的潜力。     

Retrieval of subpixel Tamarix canopy cover from Landsat data along the Forgotten River using linear and nonlinear spectral mixture models

Repeatable approaches for mapping saltcedar (Tamarix spp.) at regional scales, with the ability to detect low density stands, is crucial for the species' effective control and management, as well as for an improved understanding of its current and potential future dynamics. This study had the objective of testing subpixel classification techniques based on linear and nonlinear spectral mixture models in order to identify the best possible classification technique for repeatable mapping of saltcedar canopy cover along the Forgotten River reach of the Rio Grande. The suite of methods tested were meant to represent various levels of constraints imposed in the solution as well as varying levels of classification details (species level and landscape level), sources for endmembers (space-borne multispectral image, airborne hyperspectral image and in situ spectra measurements) and mixture modes (linear and nonlinear). A multiple scattering approximation (MSA) model was proposed as a means to represent canopy (image) reflectance spectra as a nonlinear combination of subcanopy (field) reflectance spectra. The accuracy of subpixel canopy cover was assessed through a 1-m spatial-resolution hyperspectral image and field measurements. Results indicated that: 1) When saltcedar was represented by one single image spectrum (endmember), the unconstrained linear spectral unmixing with post-classification normalization produced comparable accuracy (OA = 72%) to those delivered by partially and fully constrained linear spectral unmixing (63-72%) and even by nonlinear spectral unmixing (73%). 2) The accuracy of the fully constrained linear spectral unmixing method increased (from 67% to 77%) when the classes were represented with several image spectra. 3) Saltcedar canopy reflectance showed the strongest nonlinear relationship with respect to subcanopy reflectance, as indicated through a range of estimated canopy recollision probabilities. 4) Despite the considerations of these effects on canopy reflectance, the inversion of the nonlinear spectral mixing model with subcanopy reflectance (field) measurements yielded slightly lower accuracy (73%) than the linear counterpart (77%). Implications of these results for region-wide monitoring of saltcedar invasion are also discussed.     

基于稀疏和正交约束非负矩阵分解的高光谱解混

针对基于非负矩阵分解（NMF）的高光谱解混存在的容易陷入局部极小值和受初始值影响较大的问题，提出一种稀疏和正交约束相结合的NMF的线性解混算法SONMF。首先，从传统的基于NMF的高光谱线性解混方法出发，分析高光谱数据本身的理化特性；然后，结合丰度的稀疏性和端元的独立性两个方面，将稀疏非负矩阵分解（SNMF）和正交非负矩阵分解（ONMF）两种方法结合应用到高光谱解混当中。模拟数据和真实数据实验表明，相比顶点成分分析法（VCA）、SNMF和ONMF这三种参考解混算法，所提算法提高了线性解混的性能；其中，评价指标光谱角距离（SAD）降低了0.012~0.145。SONMF能够结合两种约束条件的优势，弥补传统基于NMF线性解混方法对高光谱数据表达的不足，取得较好的效果。     

利用稳健非负矩阵分解实现无监督高光谱解混

目的 基于非负矩阵分解的高光谱图像无监督解混算法普遍存在着目标函数对噪声敏感、在低信噪比条件下端元提取和丰度估计性能不佳的缺点。因此,提出一种基于稳健非负矩阵分解的高光谱图像混合像元分解算法。方法 首先在传统基于非负矩阵分解的解混算法基础上,对目标函数加以改进,用更加稳健的L₁范数作为重建误差项,提高算法对噪声的适应能力,得到新的无监督解混目标函数。针对新目标函数的非凸特性,利用梯度下降法对端元矩阵和丰度矩阵交替迭代求解,进而完成优化求解,得到端元和丰度估计值。结果 分别利用模拟和真实高光谱数据,对算法性能进行定性和定量分析。在模拟数据集中,将本文算法与具有代表性的5种无监督解混算法进行比较,相比于对比算法中最优者,本文算法在典型信噪比20 dB下,光谱角距离（spectral angle distance,SAD）增大了10.5%,信号重构误差（signal to reconstruction error,SRE）减小了9.3%;在真实数据集中,利用光谱库中的地物光谱特征验证本文算法端元提取质量,并利用真实地物分布定性分析丰度估计结果。结论 提出的基于稳健非负矩阵分解的高光谱无监督解混算法,在低信噪比条件下,能够获得较好的端元提取和丰度估计精度,解混效果更好。     

A survey of methods incorporating spatial information in image classification and spectral unmixing

Over the past decade, the incorporation of spatial information has drawn increasing attention in multispectral and hyperspectral data analysis. In particular, the property of spatial autocorrelation among pixels has shown great potential for improving understanding of remotely sensed imagery. In this paper, we provide a comprehensive review of the state-of-the-art techniques in incorporating spatial information in image classification and spectral unmixing. For image classification, spatial information is accounted for in the stages of pre-classification, sample selection, classifiers, post-classification, and accuracy assessment. With regards to spectral unmixing, spatial information is discussed in the context of endmember extraction, selection of endmember combinations, and abundance estimation. Finally, a perspective on future research directions for advancing spatial-spectral methods is offered.     

Abundance estimation of crossover double particle swarms optimization for hyperspectral remote sensing imagery

ABSTRACT

Hyperspectral unmixing (HU) is an important technique for extracting materials and their abundance in hyperspectral remote sensing imagery. The presence of nonlinear mixing of light on the ground poses a difficult problem when estimating abundance fractions of all pixels. This problem makes the foundation of algorithms that can adapt all types of nonlinear mixing on the ground more complex and challenged. In this paper, a new bionic intelligent algorithm named crossover double particle swarms optimization (CDPSO) has been presented to estimate abundance for hyperspectral remote sensing imagery. The reconstruction error is used as the objective function for HU based on multilinear mixing model, and the nonlinear unmixing is transformed into an optimization problem. By improving the optimization performance of PSO for HU, we embed two types of new strategies, including double particle swarms crossover and swarm re-initialization, respectively. Our experiments, conducted using both synthetic and real hyperspectral data, demonstrate that the proposed CDPSO algorithm can outperform other state-of-the-art unmixing methods.     

Postfire soil burn severity mapping with hyperspectral image unmixing

Burn severity is mapped after wildfires to evaluate immediate and long-term fire effects on the landscape. Remotely sensed hyperspectral imagery has the potential to provide important information about fine-scale ground cover components that are indicative of burn severity after large wildland fires. Airborne hyperspectral imagery and ground data were collected after the 2002 Hayman Fire in Colorado to assess the application of high resolution imagery for burn severity mapping and to compare it to standard burn severity mapping methods. Mixture Tuned Matched Filtering (MTMF), a partial spectral unmixing algorithm, was used to identify the spectral abundance of ash, soil, and scorched and green vegetation in the burned area. The overall performance of the MTMF for predicting the ground cover components was satisfactory (r² = 0.21 to 0.48) based on a comparison to fractional ash, soil, and vegetation cover measured on ground validation plots. The relationship between Landsat-derived differenced Normalized Burn Ratio (dNBR) values and the ground data was also evaluated (r² = 0.20 to 0.58) and found to be comparable to the MTMF. However, the quantitative information provided by the fine-scale hyperspectral imagery makes it possible to more accurately assess the effects of the fire on the soil surface by identifying discrete ground cover characteristics. These surface effects, especially soil and ash cover and the lack of any remaining vegetative cover, directly relate to potential postfire watershed response processes.