FPGA-based architecture for hyperspectral endmember extraction

Hyperspectral instruments have been incorporated in satellite missions, providing data of high spectral resolution of the Earth. This data can be used in remote sensing applications, such as, target detection, hazard prevention, and monitoring oil spills, among others. In most of these applications, one of the requirements of paramount importance is the ability to give real-time or near real-time response. Recently, onboard processing systems have emerged, in order to overcome the huge amount of data to transfer from the satellite to the ground station, and thus, avoiding delays between hyperspectral image acquisition and its interpretation. For this purpose, compact reconfigurable hardware modules, such as field programmable gate arrays (FPGAs) are widely used. This paper proposes a parallel FPGA-based architecture for endmember’s signature extraction. This method based on the Vertex Component Analysis (VCA) has several advantages, namely it is unsupervised, fully automatic, and it works without dimensionality reduction (DR) pre-processing step. The architecture has been designed for a low cost Xilinx Zynq board with a Zynq-7020 SoC FPGA based on the Artix-7 FPGA programmable logic and tested using real hyperspectral data sets collected by the NASA’s Airborne Visible Infra-Red Imaging Spectrometer (AVIRIS) over the Cuprite mining district in Nevada. Experimental results indicate that the proposed implementation can achieve real-time processing, while maintaining the methods accuracy, which indicate the potential of the proposed platform to implement high-performance, low cost embedded systems, opening new perspectives for onboard hyperspectral image processing.

On independent component analysis applied to unmixing hyperspectral data

One of the most challenging task underlying many hyperspectral imagery applications is the spectral unmixing, which decomposes a mixed pixel into a collection of reectance spectra, called endmember signatures, and their corresponding fractional abundances. Independent Component Analysis (ICA) have recently been proposed as a tool to unmix hyperspectral data. The basic goal of ICA is to nd a linear transformation to recover independent sources (abundance fractions) given only sensor observations that are unknown linear mixtures of the unobserved independent sources. In hyperspectral imagery the sum of abundance fractions associated to each pixel is constant due to physical constraints in the data acquisition process. Thus, sources cannot be independent. This paper address hyperspectral data source dependence and its impact on ICA performance. The study consider simulated and real data. In simulated scenarios hyperspectral observations are described by a generative model that takes into account the degradation mechanisms normally found in hyperspectral applications. We conclude that ICA does not unmix correctly all sources. This conclusion is based on the a study of the mutual information. Nevertheless, some sources might be well separated mainly if the number of sources is large and the signal-to-noise ratio (SNR) is high.

Signal subspace identification in hyperspectral imagery

Terrestrial remote sensing imagery involves the acquisition of information from the Earth's surface without physical contact with the area under study. Among the remote sensing modalities, hyperspectral imaging has recently emerged as a powerful passive technology. This technology has been widely used in the fields of urban and regional planning, water resource management, environmental monitoring, food safety, counterfeit drugs detection, oil spill and other types of chemical contamination detection, biological hazards prevention, and target detection for military and security purposes [2-9]. Hyperspectral sensors sample the reflected solar radiation from the Earth surface in the portion of the spectrum extending from the visible region through the near-infrared and mid-infrared (wavelengths between 0.3 and 2.5 µm) in hundreds of narrow (of the order of 10 nm) contiguous bands [10]. This high spectral resolution can be used for object detection and for discriminating between different objects based on their spectral xharacteristics [6]. However, this huge spectral resolution yields large amounts of data to be processed. For example, the Airbone Visible/Infrared Imaging Spectrometer (AVIRIS) [11] collects a 512 (along track) X 614 (across track) X 224 (bands) X 12 (bits) data cube in 5 s, corresponding to about 140 MBs. Similar data collection ratios are achieved by other spectrometers [12]. Such huge data volumes put stringent requirements on communications, storage, and processing. The problem of signal sbspace identification of hyperspectral data represents a crucial first step in many hypersctral processing algorithms such as target detection, change detection, classification, and unmixing. The identification of this subspace enables a correct dimensionality reduction (DR) yelding gains in data storage and retrieval and in computational time and complexity. Additionally, DR may also improve algorithms performance since it reduce data dimensionality without losses in the useful signal components. The computation of statistical estimates is a relevant example of the advantages of DR, since the number of samples required to obtain accurate estimates increases drastically with the dimmensionality of the data (Hughes phnomenon) [13].

Spatial Technologies: What use are they to managers of environmental lands?

This paper reviews the potential use of three types of spatial technology to land managers, namely satellite imagery, satellite positioning systems and supporting computer software. Developments in remote sensing and the relative advantages of multispectral and hyperspectral images are discussed. The main challenge to the wider use of remote sensing as a land management tool is seen as uncertainty whether apparent relationships between biophysical variables and spectral reflectance are direct and causal, or artefacts of particular images. Developments in satellite positioning systems are presented in the context of land managers’ need for position estimates in situations where absolute precision may or may not be required. The role of computer software in supporting developments in spatial technology is described. Spatial technologies are seen as having matured beyond empirical applications to the stage where they are useful and reliable land management tools. In addition, computer software has become more user-friendly and this has facilitated data collection and manipulation by semi-expert as well as specialist staff.

Bi-directional reflectance of corals

An assessment of the bi-directional reflectance distribution function (BRDF) of corals with different morphologies was undertaken using hyperspectral reflectance measurements. The highest variance in reflectance obtained at different viewing angles was found for the open branching Acropora colony, possibly attributed to branch orientation and internal shadow distribution. Spectral separation within and between features at the nominated sensor-viewing angles was greatest in the near infrared portion of the spectrum. The analysis of coral reef bi-directional reflectance properties and degree of internal shadowing holds potential for future assessment and information extraction relating to coral structural characteristics.

GPU implementation of the simplex identification via split augmented Lagrangian

Hyperspectral imaging can be used for object detection and for discriminating between different objects based on their spectral characteristics. One of the main problems of hyperspectral data analysis is the presence of mixed pixels, due to the low spatial resolution of such images. This means that several spectrally pure signatures (endmembers) are combined into the same mixed pixel. Linear spectral unmixing follows an unsupervised approach which aims at inferring pure spectral signatures and their material fractions at each pixel of the scene. The huge data volumes acquired by such sensors put stringent requirements on processing and unmixing methods. This paper proposes an efficient implementation of a unsupervised linear unmixing method on GPUs using CUDA. The method finds the smallest simplex by solving a sequence of nonsmooth convex subproblems using variable splitting to obtain a constraint formulation, and then applying an augmented Lagrangian technique. The parallel implementation of SISAL presented in this work exploits the GPU architecture at low level, using shared memory and coalesced accesses to memory. The results herein presented indicate that the GPU implementation can significantly accelerate the method's execution over big datasets while maintaining the methods accuracy.

Urban image classification with semisupervised multiscale cluster kernels

This paper presents a semisupervised support vector machine (SVM) that integrates the information of both labeled and unlabeled pixels efficiently. Method's performance is illustrated in the relevant problem of very high resolution image classification of urban areas. The SVM is trained with the linear combination of two kernels: a base kernel working only with labeled examples is deformed by a likelihood kernel encoding similarities between labeled and unlabeled examples. Results obtained on very high resolution (VHR) multispectral and hyperspectral images show the relevance of the method in the context of urban image classification. Also, its simplicity and the few parameters involved make the method versatile and workable by unexperienced users.

Cluster kernels for semisupervised classification of VHR urban images

In this paper, we present and apply a semisupervised support vector machine based on cluster kernels for the problem of very high resolution image classification. In the proposed setting, a base kernel working with labeled samples only is deformed by a likelihood kernel encoding similarities between unlabeled examples. The resulting kernel is used to train a standard support vector machine (SVM) classifier. Experiments carried out on very high resolution (VHR) multispectral and hyperspectral images using very few labeled examples show the relevancy of the method in the context of urban image classification. Its simplicity and the small number of parameters involved make it versatile and workable by unexperimented users.

Modelling plant species distribution in alpine grasslands using airborne imaging spectroscopy.

Remote sensing using airborne imaging spectroscopy (AIS) is known to retrieve fundamental optical properties of ecosystems. However, the value of these properties for predicting plant species distribution remains unclear. Here, we assess whether such data can add value to topographic variables for predicting plant distributions in French and Swiss alpine grasslands. We fitted statistical models with high spectral and spatial resolution reflectance data and tested four optical indices sensitive to leaf chlorophyll content, leaf water content and leaf area index. We found moderate added-value of AIS data for predicting alpine plant species distribution. Contrary to expectations, differences between species distribution models (SDMs) were not linked to their local abundance or phylogenetic/functional similarity. Moreover, spectral signatures of species were found to be partly site-specific. We discuss current limits of AIS-based SDMs, highlighting issues of scale and informational content of AIS data.

Influence of data acquisition geometry on soybean spectral response simulated by the prosail model

View angle and directional effects significantly affect reflectance and vegetation indices, especially when daily images collected by large field-of-view (FOV) sensors like the Moderate Resolution Imaging Spectroradiometer (MODIS) are used. In this study, the PROSAIL radiative transfer model was chosen to evaluate the impact of the geometry of data acquisition on soybean reflectance and two vegetation indices (Normalized Difference Vegetation Index - NDVI and Enhanced Vegetation Index -EVI) by varying biochemical and biophysical parameters of the crop. Input values for PROSAIL simulation were based on the literature and were adjusted by the comparison between simulated and real satellite soybean spectra acquired by the MODIS/Terra and hyperspectral Hyperion/Earth Observing-One (EO-1). Results showed that the influence of the view angle and view direction on reflectance was stronger with decreasing leaf area index (LAI) and chlorophyll concentration. Because of the greater dependence on the near-infrared reflectance, the EVI was much more sensitive to viewing geometry than NDVI presenting larger values in the backscattering direction. The contrary was observed for NDVI in the forward scattering direction. In relation to the LAI, NDVI was much more isotropic for closed soybean canopies than for incomplete canopies and a contrary behavior was verified for EVI.

Développement d'un indice de vulnérabilité à l'érosion éolienne à partir d'images satellitales, dans le Bassin arachidier du Sénégal : cas de la région de Thiès.

L’érosion éolienne est un problème environnemental parmi les plus sévères dans les régions arides, semi-arides et les régions sèches sub-humides de la planète. L’érosion des sols accélérée par le vent provoque des dommages à la fois localement et régionalement. Sur le plan local, elle cause la baisse des nutriments par la mobilisation des particules les plus fines et de la matière organique. Cette mobilisation est une des causes de perte de fertilité des sols avec comme conséquence, une chute de la productivité agricole et une réduction de la profondeur de la partie arable. Sur le plan régional, les tempêtes de poussières soulevées par le vent ont un impact non négligeable sur la santé des populations, et la déposition des particules affecte les équipements hydrauliques tels que les canaux à ciel ouvert ainsi que les infrastructures notamment de transport. Dans les régions où les sols sont fréquemment soumis à l’érosion éolienne, les besoins pour des études qui visent à caractériser spatialement les sols selon leur degré de vulnérabilité sont grands. On n’a qu’à penser aux autorités administratives qui doivent décider des mesures à prendre pour préserver et conserver les potentialités agropédologiques des sols, souvent avec des ressources financières modestes mises à leur disposition. Or, dans certaines de ces régions, comme notre territoire d’étude, la région de Thiès au Sénégal, ces études font défaut. En effet, les quelques études effectuées dans cette région ou dans des contextes géographiques similaires ont un caractère plutôt local et les approches suivies (modèles de pertes des sols) nécessitent un nombre substantiel de données pour saisir la variabilité spatiale de la dynamique des facteurs qui interviennent dans le processus de l’érosion éolienne. La disponibilité de ces données est particulièrement problématique dans les pays en voie de développement, à cause de la pauvreté en infrastructures et des problèmes de ressources pour le monitoring continu des variables environnementales. L’approche mise de l’avant dans cette recherche vise à combler cette lacune en recourant principalement à l’imagerie satellitale, et plus particulièrement celle provenant des satellites Landsat-5 et Landsat-7. Les images Landsat couvrent la presque totalité de la zone optique du spectre exploitable par télédétection (visible, proche infrarouge, infrarouge moyen et thermique) à des résolutions relativement fines (quelques dizaines de mètres). Elles permettant ainsi d’étudier la distribution spatiale des niveaux de vulnérabilité des sols avec un niveau de détails beaucoup plus fin que celui obtenu avec des images souvent utilisées dans des études environnementales telles que AVHRR de la série de satellites NOAA (résolution kilométrique). De plus, l’archive complet des images Landsat-5 et Landsat-7 couvrant une période de plus de 20 ans est aujourd’hui facilement accessible. Parmi les paramètres utilisés dans les modèles d’érosion éolienne, nous avons identifiés ceux qui sont estimables par l’imagerie satellitale soit directement (exemple, fraction du couvert végétal) soit indirectement (exemple, caractérisation des sols par leur niveau d’érodabilité). En exploitant aussi le peu de données disponibles dans la région (données climatiques, carte morphopédologique) nous avons élaboré une base de données décrivant l’état des lieux dans la période de 1988 à 2002 et ce, selon les deux saisons caractéristiques de la région : la saison des pluies et la saison sèche. Ces données par date d’acquisition des images Landsat utilisées ont été considérées comme des intrants (critères) dans un modèle empirique que nous avons élaboré en modulant l’impact de chacun des critères (poids et scores). À l’aide de ce modèle, nous avons créé des cartes montrant les degrés de vulnérabilité dans la région à l’étude, et ce par date d’acquisition des images Landsat. Suite à une série de tests pour valider la cohérence interne du modèle, nous avons analysé nos cartes afin de conclure sur la dynamique du processus pendant la période d’étude. Nos principales conclusions sont les suivantes : 1) le modèle élaboré montre une bonne cohérence interne et est sensible aux variations spatiotemporelles des facteurs pris en considération 2); tel qu’attendu, parmi les facteurs utilisés pour expliquer la vulnérabilité des sols, la végétation vivante et l’érodabilité sont les plus importants ; 3) ces deux facteurs présentent une variation importante intra et inter-saisonnière de sorte qu’il est difficile de dégager des tendances à long terme bien que certaines parties du territoire (Nord et Est) aient des indices de vulnérabilité forts, peu importe la saison ; 4) l’analyse diachronique des cartes des indices de vulnérabilité confirme le caractère saisonnier des niveaux de vulnérabilité dans la mesure où les superficies occupées par les faibles niveaux de vulnérabilité augmentent en saison des pluies, donc lorsque l’humidité surfacique et la végétation active notamment sont importantes, et décroissent en saison sèche ; 5) la susceptibilité, c’est-à-dire l’impact du vent sur la vulnérabilité est d’autant plus forte que la vitesse du vent est élevée et que la vulnérabilité est forte. Sur les zones où la vulnérabilité est faible, les vitesses de vent élevées ont moins d’impact. Dans notre étude, nous avons aussi inclus une analyse comparative entre les facteurs extraits des images Landsat et celles des images hyperspectrales du satellite expérimental HYPERION. Bien que la résolution spatiale de ces images soit similaire à celle de Landsat, les résultats obtenus à partir des images HYPERION révèlent un niveau de détail supérieur grâce à la résolution spectrale de ce capteur permettant de mieux choisir les bandes spectrales qui réagissent le plus avec le facteur étudié. Cette étude comparative démontre que dans un futur rapproché, l’amélioration de l’accessibilité à ce type d’images permettra de raffiner davantage le calcul de l’indice de vulnérabilité par notre modèle. En attendant cette possibilité, on peut de contenter de l’imagerie Landsat qui offre un support d’informations permettant tout de même d’évaluer le niveau de fragilisation des sols par l’action du vent et par la dynamique des caractéristiques des facteurs telles que la couverture végétale aussi bien vivante que sénescente.

The use of earth observation techniques to improve catchment-scale pollution predictions

Remote sensing can potentially provide information useful in improving pollution transport modelling in agricultural catchments. Realisation of this potential will depend on the availability of the raw data, development of information extraction techniques, and the impact of the assimilation of the derived information into models. High spatial resolution hyperspectral imagery of a farm near Hereford, UK is analysed. A technique is described to automatically identify the soil and vegetation endmembers within a field, enabling vegetation fractional cover estimation. The aerially-acquired laser altimetry is used to produce digital elevation models of the site. At the subfield scale the hypothesis that higher resolution topography will make a substantial difference to contaminant transport is tested using the AGricultural Non-Point Source (AGNPS) model. Slope aspect and direction information are extracted from the topography at different resolutions to study the effects on soil erosion, deposition, runoff and nutrient losses. Field-scale models are often used to model drainage water, nitrate and runoff/sediment loss, but the demanding input data requirements make scaling up to catchment level difficult. By determining the input range of spatial variables gathered from EO data, and comparing the response of models to the range of variation measured, the critical model inputs can be identified. Response surfaces to variation in these inputs constrain uncertainty in model predictions and are presented. Although optical earth observation analysis can provide fractional vegetation cover, cloud cover and semi-random weather patterns can hinder data acquisition in Northern Europe. A Spring and Autumn cloud cover analysis is carried out over seven UK sites close to agricultural districts, using historic satellite image metadata, climate modelling and historic ground weather observations. Results are assessed in terms of probability of acquisition probability and implications for future earth observation missions. (C) 2003 Elsevier Ltd. All rights reserved.