Aircraft and space-based hyperspectral imaging (HSI) sensors tailored for the reflective or emissive spectral regimes are being designed and developed for a wide variety of military, civil and science applications. Key sensor-level HSI system performance requirements dictate the optical, spectrometer, focal plane and data processing design parameters for a given choice of spectral instrument design and platform altitude. A detailed understanding of the performance/sensor design trade-space that is available facilitates informed decision making and planning. We have developed a spreadsheet-based sensitivity analysis tool for dispersive HSI sensors that enables rapid and meaningful investigation of candidate sensor designs over a wide variety of parametric conditions at a level of detail consistent with the first-order specification of the instrument subsystems. Our approach also facilitates: assessment of a fixed sensor design against varied atmospheric/target phenomenology assumptions, determination of sensor design drivers, and sensor design optimization. Our parametric analysis capability is illustrated by synthesizing a relatively detailed HSI dispersive design based on optical aperture diameter of 70 cm and an orbital altitude of 690 km. These two parameters are borrowed from the IKONOS commercial remote sensing system. As part of this synthesis, sensitivity enhancement by back-scanning is analyzed for the purpose of deriving both the maximum sensor contiguous scan length and the associated precision line-of-sight pitch angle rate control requirements.

Advancements in Mercury Cadmium Telluride (MCT) focal plane arrays (FPA) in recent years have allowed high performance longwave infrared imagers to prosper. In particular molecular and gas/chemical spectroscopy applications can be vastly advanced with these new products. However, for the transition from single pixel spectrometers to FPA base imaging spectrometers to succeed, a couple of parallel advancements must be made as well. Most Fourier transform spectrometers currently available are designed specifically for a 1 mm single pixel detector. Scientists who try to convert these systems into imaging spectrometers quickly run into throughput issues when FPAs reach sizes of up to 12.5mm, thus limiting the performance and greatly impacting the detection capabilities. Furthermore, for large FPAs the readout time can be significantly longer than the integration time. In turn, this requires slower sweep speeds with a higher degree of control of the scanning mechanism. The benefit of these new technologies in spectroscopy can only be demonstrated with a system optimally designed for imaging spectroscopy. This paper will address the issues of imaging spectroscopy and will show how an instrument designed for specifically imaging applications can dramatically improve the performance of the system and quality of the data acquired.

For the fast acquisition of large amounts of BRDF data over wavelength to be compiled into libraries, a small, fast and rugged spectro-radiometer without moving parts for angular resolution is being developed. The system consists of an elliptical mirror which maps a semi-hemisphere onto a CMOS-detector with a dynamic range of 140dB. The detector has 32887 pixels which are calibrated radiometrically in the range from 10^-5 W/m² to 100 W/m² (7 decades). The system can take 30 semi-hemispherical BRDFs per second, i.e. nearly 1 million solid angles per wavelength per second. The smallest illumination spot has an area of 0.03mm², for spatial averaging the sample is mounted on x-y-stages, so the largest avaraged spot can be about 50.000 mm². Incoherent illumination is provided by a set of assorted LED's. The paper deals with the instrument design, and gives some measurement results.

Airborne surveillance presents challenging target-detection opportunities for optical remote sensors, especially under the constraints of size, weight, and power imposed by small aircraft. We present a spatial-frequency dependent figure-of-merit, called the Detector Quantum Efficiency (DQE), by first tracing its origins in single pixel photon multiplication detectors, where it is shown to be yield (quantum efficiency or QE) divided by the noise factor. We then show the relationship of DQE to several well-known figures-of-merit. Finally we broaden the definition of DQE to include the spatial-frequency dependence on the MTF of the system and the noise power spectrum (NPS) of the detector. We then present the results of the application of this DQE to a hyperspectral camera under development at BAE Systems Spectral Solutions LLC.

Imaging Fourier transform spectrometers (IFTS) are becoming the preferred systems for remote sensing spectral imaging applications because of their ability to provide, simultaneously, both high spatial and spectral resolution images of a scene. IFTS can be operated in either step-and-integrate or rapid-scan modes, where it is common practice to sample interferograms at equal optical path difference intervals. The step-and-integrate mode requires a translation stage with fast and precise point-to-point motion and additional external trigger circuitry for the detector focal plane array (FPA), and produces uniformly position-sampled interferograms which can be analyzed using standard FFT routines. In the rapid-scan mode, the translation stage is continuously moving and interferograms are often acquired at the frame-rate of the FPA. Since all translation stages have associated velocity errors, the resulting interferograms are sampled at non-uniform intervals of optical path difference, which requires more sophisticated analysis. This paper discusses the processing pipeline which is being developed for the analysis of the non-uniform rapid-scan data produced by the Herschel/SPIRE IFTS.

Spectrally tunable quantum-dot infrared photodetectors (QDIPs) can be used to approximate multiple spectral responses with the same focal-plane array. Hence, they exhibit the potential for real time adaptive detection/classification. In the present study, it is shown that we can perform the detection/classification operation at the adaptive focal-plane array (AFPA) based on QDIPs by fitting the QDIP's response to the correspondent operators. With a new understanding of spectral signature in the sensor space, the best fitting can be achieved. Our simulation results show how well QDIPs perform in different regions of the spectrum in the mid- and long wave infrared. The results indicate that the AFPA performance does not match that of the ideal filtering operators, but reliable measurement can be accomplished.

Prototype 2-, 3-, and 4- band long wave infrared (LWIR) focal plane arrays (FPA) for missile defense applications have recently been constructed to enhance target discrimination in space-based interceptor seekers. To address issues related to target identification such as algorithm choice and band number, this study created synthesized, optimized (using a genetic algorithm) image cubes (8- 12 mm) of targets and backgrounds compatible with expected mid-course defense scenarios and current multicolor sensors. Each candidate band was weighted using an interacting band edge model for 2-, 3- or 4- band sensors, consistent with a DRS multi-color HgCdTe LWIR FPA. Whitening the binned cubes and assigning red, green, blue colors directly to the whitened data set can prominently display and identify targets. Modified target signatures applied in matched filters searches and spectral angle maps autonomously searched for targets in the synthetic binned image cubes. Target discrimination diminished with decreasing target temperature and/or increasing distance between sensor and targets due to mixing subpixel target spectra with noise background. Spectral angle maps identified target temperatures and materials substantively better than the matched filter in this particular study. Target material and temperature identification improved by increasing number of bands, with greatest improvement for 3 bands relative to 2 bands. Extending detector sensitivity to 6-14 mm failed to improve target identification. This is the first study to systematically examine target identification in synthetic images cubes, consistent with missile defense scenarios and current multi-sensor technology.

Several remote sensing platforms have been developed that include boresighted hyperspectral and panchromatic imaging sensors. The NASA EO-1 platform is a prime example, and includes the hyperspectral Hyperion sensor and multispectral Advanced Land Imager (ALI). Typically in these cases, the panchromatic imagery that is produced is of higher spatial resolution than the hyperspectral imagery. In the NASA EO-1 case, for example, Hyperion exhibits a 30 meter ground sample distance (GSD) and the ALI includes a 10 meter GSD panchromatic band. This paper addresses the issue of combining concurrent imagery from both sources with the goal of deriving a hyperspectral image with the spectral resolution of the hyperspectral data source and the spatial resolution of the panchromatic data source. Specifically, the use of a stochastic mixing model (SMM) along with MAP estimation is extended to the case where the point spread function of the hyperspectral sensor is not assumed to be detector-limited. This case is addressed by using an iterative optimization strategy based on a parametric description of the point spread function of the hyperspectral sensor. Results indicate that the iterative approach appears to find the optimal MAP solution. This paper summarizes the MAP/SMM enhancement method, the iterative optimization strategy, and enhancement results.

Binary coding is one of simplest ways to characterize spectral features. One commonly used method is a binary coding-based image software system, called Spectral Analysis Manager (SPAM) for remotely sensed imagery developed by Mazer et al. For a given spectral signature, the SPAM calculates its spectral mean and inter-band spectral difference and uses them as thresholds to generate a binary code word for this particular spectral signature. Such coding scheme is generally effective and also very simple to implement. This paper revisits the SPAM and further develops three new SPAM-based binary coding methods, called equal probability partition (EPP) binary coding, halfway partition (HP) binary coding and median partition (MP) binary coding. These three binary coding methods along with the SPAM well be evaluated for spectral discrimination and identification. In doing so, a new criterion, called a posteriori discrimination probability (APDP) is also introduced for performance measure.

The Remote Sensing Group of the Optical Sciences Center at the University of Arizona has a history of collecting ground-based atmospheric data in support of calibration/validation and for atmospheric correction. This work has included the determination of columnar water vapor based on measurements of direct solar irradiance. In the past, the conversion of these data to transmittance and then column water vapor has relied upon a modified Langley approach and two-parameter band model of absorption developed in the 1980s at the University of Arizona. More recently, the RSG has used the well-known MODTRAN code for its prediction of at-sensor radiance and atmospheric correction. This work examines the use of the same MODTRAN code for the retrieval of column water vapor to simplify the overall processing approach of the RSG, as well as providing consistency between the measurements and the predicted at-sensor radiance. The water vapor retrieval using MODTRAN follows the same basic approach as the previous method except that the water vapor absorption parameters are obtained from MODTRAN. A sensitivity analysis is performed to examine the influence of the MODTRAN input parameters on the retrieval. The results of the new method are compared with results from GPS-derived column water vapor. These preliminary results show that the MODTRAN-based values have an accuracy of 10% and agreement with the GPS-derived results is better than 10%.

This paper describes a Compact High Resolution Imaging Spectrometer (CHRIS) developed at Sira. The imaging spectrometer is flying on PROBA, a small agile satellite, which was launch in 2001. This paper provides details of the instrument design and performance. The main purpose of the instrument is to provide images of land areas, particularly the measurement of the Bi-directional Reflectance Distribution Function (BRDF) properties for selected targets on the Earth surface using multi-angle observations. The platform provides pointing in both across-track and along-track directions, for target acquisition, BRDF and aerosol measurements and slow pitch during imaging in order to increase the integration time of the instrument. This increase in integration time enhances the target radiometric resolution. The spectral range covered by the instrument extends from 400nm to 1050nm. The platform orbits the Earth with an apogee of 673km and a perigee of 560km. The spatial sampling interval at apogee is approximately 17m. In this mode it is possible to readout 19 spectral bands. The location and width of the spectral bands are programmable. Selectable on-chip integration can increase the number of bands to 63 for a spatial sampling interval of 34m. The swath width imaged is 13km at perigee.

PROBA-1 is a technology demonstration mission of the European Space Agency's General Support Technology Programme. It was launched on October, 22nd, 2001 in a LEO, Sun-synchronous, 681x561 km orbit. The spacecraft mass is 94 kg, with 25 kg dedicated to scientific and Earth observation instruments, in addition to the technology demonstration payloads. The principal objective is the in-orbit evaluation of new spacecraft technologies. PROBA-1, however, has also been intended as a flight opportunity for Earth observation instruments that can benefit from the agile pointing capabilities and the autonomy features of the satellite. PROBA-1 onboard automatic functions include all payload operations scheduling and execution, target fly-by prediction and control of camera pointing and scanning from raw user inputs (target latitude, longitude and altitude). The point and stare requirements of the High Resolution Camera (HRC), as well as the multiple image scan requirement to support Bi-directional Reflectance Distribution Function (BRDF) measurements with the Compact High Resolution Imaging Spectrometer (CHRIS) are satisfied with the specified accuracy, by this small and agile gyro-less platform, whose attitude determination is based on autonomous star tracker only. The main Earth imaging payload, CHRIS, weighing only 14 kg, is used to measure directional spectral reflectance. The instrument is capable of imaging up to 200 spectral bands simultaneously at full resolution with a spatial resolution of 20m at nadir and swath width of 15 km. The HRC is a black and white camera with a miniaturised Cassegrain telescope providing 5m geometrical resolution images. Each image covers a ground area of approximately 4 km by 4 km. The pointing agility of the spacecraft allows both cameras to take multiple images of the same target area at different viewing angles on the same orbital pass. This paper covers the spacecraft design and in-flight performance, as well as a description of the enabling technologies.

The CHRIS (Compact High Resolution Imaging Spectrometer) instrument was launched onboard the European Space Agency (ESA) PROBA satellite on 22 October 2001. CHRIS can acquire up to 63 bands of hyperspectral data at a ground spatial resolution of 36m. Alternatively, the instrument can be configured to acquire 18 bands of data with a spatial resolution of 17m. PROBA, by virtue of its agile pointing capability, enables CHRIS to acquire five different angle images of the selected site. Two sites can be acquired every 24 hours. The hyperspectral and multi-angle capability of CHRIS makes it an important resource for stydying BRDF phenomena of vegetation. Other applications include coastal and inland waters, wild fires, education and public relations. An effective data acquisition planning procedure has been implemented and since mid-2002 users have been receiving data for analysis. A cloud prediction routine has been adopted that maximises the image acquisition capacity of CHRIS-PROBA. Image acquisition planning is carried out by RSAC Ltd on behalf of ESA and in co-operation with Sira Technology Ltd and Redu, the ESA ground station in Belgium, responsible for CHRIS-PROBA.

To perform point target acquisition in multispectral and hyperspectral images, it is often advantageous to compare the signature of the investigated pixel to a known target signature. To do this properly, it is necessary to estimate the expected mean and covariance matrix of an investigated pixel in a particular location, based on its local surroundings. The degree to which this pixel signature differs from the estimated background then becomes the data, which is matched to the desired target signature. The standard method for such an analysis is the RX algorithm of Reed and Yu. The mean is normally estimated from the local environment of the pixel; the covariance matrix can either be estimated globally or in some local window. In recent research, we have considered how to improve the algorithm by eliminating edge points as potential false alarms. In the present work, a prior segmentation of the image before processing is utilized. While our estimate for the mean continues to be based on the immediate neighbors of the investigated pixel, our estimate of the covariance matrix is now based on the covariance matrix of the segment to which the adjacent pixels belong. In this way, we get a more accurate estimate of the covariance matrix. Results on real multispectral and hyperspectral images with embedded targets in several spectral regions are presented and improvement is demonstrated.

Anomaly detection in hyperspectral imagery seeks to identify a small subset of pixels whose spectra differ most significantly from the background. The challenge is to characterize the background and noise well enough to recognize which observations are truly distinct and not simply noise outliers. The covariance-based RXD operator was developed to select low-probability pixel spectra and is therefore sensitive to noise. We compare the RXD operator to a Euclidean metric weighted by the inverse of the estimated spectral noise variance. We then combine the weighted Euclidean metric with RXD using a Lagrange multiplier and demonstrate that this formulation retains RXD's emphasis on small clusters while controlling the impact of noise. An optimum value of the Lagrange multiplier is determined based on the number of bands. We explore the utility of normalizing the pixel spectra as a step in anomaly detection. Results for the RXD, weighted-Euclidean, and Lagrange approach are presented using AVIRIS and HYDICE imagery. Based on these results, we conclude that the Euclidean, although robust to noise, does little more than emphasize the brightest pixels. The Lagrange detector selects the same regions as RXD while significantly reducing the impact of noise.

The purpose of this paper is to introduce a new anomaly detection algorithm for application to hyperspectral imaging (HSI) data. The algorithm uses characterisations of the joint (among wavebands) probability density function (pdf) of HSI data. Traditionally, the pdf has been assumed to be multivariate Gaussian or a mixture of multivariate Gaussians. Other distributions have been considered by previous authors, in particular Elliptically Contoured Distributions (ECDs). In this paper we focus on another distribution, which has only recently been defined and studied. This distribution has a more flexible and extensive set of parameters than the multivariate Gaussian does, yet the pdf takes on a relatively simple mathematical form. The result of all this is a model for the pdf of a hyperspectral image, consisting of a mixture of these distributions. Once a model for the pdf of a hyperspectral image has been obtained, it can be incorporated into an anomaly detector. The new anomaly detector is implemented and applied to some medium wave infra-red (MWIR) hyperspectral imagery. Comparison is made with a well-known anomaly detector, and it will be seen that the results are promising.

Tools that perform pixel-by-pixel classification of multispectral imagery are useful in broad area mapping applications such as terrain categorization, but are less well-suited to the detection of discrete objects. Pixel-by-pixel classifiers, however, have many advantages: they are relatively simple to design, they can readily employ formal machine learning tools, and they are widely available on a variety of platforms. We describe an approach that enables pixel-by-pixel classifiers to be more effectively used in object-detection settings. This is achieved by optimizing a metric which does not attempt to precisely delineate every pixel comprising the objects of interest, but instead focusses the attention of the analyst to these objects without the distraction of many false alarms. The approach requires only minor modification of exisiting pixel-by-pixel classifiers, and produces substantially improved performance. We will describe algorithms that employ this approach and show how they work on a varitety of object detection problems using remotely-sensed multispectral data.

Least square unmixing approach has been successfully applied in hyperspectral image processing for subpixel target detection. It can detect target with size less than a pixel by estimating its abundance fraction resident in each pixel. In order for the this approach to be effective, the number of bands must be larger than or equal to that of signatures to be classified, i.e., the number of equations should be no less than the number of unknowns. This ensures that there are sufficient dimensions to accommodate orthogonal projections resulting from the individual signatures. It is known as band number constraint (BNC). Such inherent constraint is not an issue for hyperspectral images since they generally have hundreds of bands, which is more than the number of signatures resident within images. However, this may not be true for multispectral images where the number of signatures to be classified is greater than the number of bands. This paper presents an extension of the least square approach that relaxes this constraint with a set of least square filters that are nonlinearly combined for endmember detection. The effectiveness of the proposed method is evaluated by SPOT images. The experimental results show significantly improves in classification performance than Orthogonal Subspace Projection (OSP).

The CHRIS sensor, mounted on the PROBA satellite, is one of the first space-borne hyperspectral sensors offering high spatial resolution (18 m x 18 m). This, combined with the possibilty of multi-temporal coverage, makes CHRIS-PROBA exceptionally well suited for lake water monitoring. The concept of the project MEMAMON is to monitor the water quality of lakes in the Mecklenburg (Germany) and Mazurian (Poland) lake districts. Both test sites contain a large number of lakes with high variability and different trophic states. This paper presents a study, which aims to determine the trophic parameter chlorophyll-a using hyperspectral CHRIS-PROBA data. To investigate the seasonal dynamics of lakes, CHRIS-PROBA data were acquired in spring, summer and autumn 2002, 2003 and 2004. A first analysis of the data showed that CHRIS radiance data have strong artefacts in along track direction. Standard destriping techniques were not sufficient to correct the data. Therefore, a novel iterative destriping technique was developed and successfully applied to CHRIS-PROBA data. During the CHRIS-PROBA data recording, spectral field measurements and acquisition of in-situ data took place at several test sites. Using these data, chlorophyll algorithms were developed and optimised to the spectral characteristics of the CHRIS sensor. On the basis of simulated CHRIS-PROBA spectra in mode 2 and mode 3, the effect on the cholorphyll estimation will be discussed.

Measurements of spectro-directional radiances done with the imaging spectrometer CHRIS on-board the agile platform PROBA are being used to determine key properties of terrestrial vegetation at the appropriate spatial resolution. These data on vegetation properties can then be used to improve the accuracy and the parameterizations of models describing biosphere processes, i.e. photosynthesis and water use by irrigated crops and trees. The vegetation properties considered are: albedo, Leaf Area Index (LAI), fractional cover, fraction of absorbed photosynthetically active radiation (fAPAR) and canopy chlorophyll content. The Natural Park of San Rossore (Pisa, Central Italy) is a primary test site for several national and international research projects dealing with forest ecosystem monitoring. In particular, since 1999 measurements of transpiration and ecosystem gas-exchange have been regularly taken in the park pine forest to characterize its main water and carbon fluxes. In the same period, several aerial flights have been carried out with onboard hyper-spectral sensors (MIVIS, VIRS, AISA), while a series of satellite images have been acquired using both conventional (NOAAAVHRR, Landsat-TM/ETM+) and advanced sensors (CHRIS-PROBA). The final objective of these activities is to calibrate and validate methodologies which integrate remotely sensed and ancillary data for monitoring forest ecosystem. More specifically, a major research effort has been focused on evaluating the additional information content provided by advanced hyper-spectral multi-angular sensors about the main parameters needed for forest characterization (species, LAI, pigment content, etc.). These activities are part of projects which are financed by the Italian and European Space Agencies (ASI and ESA, respectively) within the framework of the CHRIS-PROBA and SPECTRA missions. During 2002 and 2003 nine complete multi-angular acquisitions were successfully performed over the San Rossore site. This paper summarizes first results of the evaluation of data acquired so far, particularly forward modeling of Top Of Canopy (TOC) reflectances. The models KUUSK, SAIL and GeoSAIL were used to simulate spectro-directional reflectance of different stands in the forest and compared with PROBA - CHRIS and airborne hyperspectral observations. Deviations of simulated from observed reflectances were significant.

Bi-Directional Reflectance Distribution Function (BRDF) of natural targets is a relevant topic to many remote sensing applications. Recent satellite sensors such as the Multi-angle Imaging SpectroRadiometer (MISR) and the Compact High Resolution Imaging Spectrometer (CHRIS) supply experimental data to improve the current understanding of directional properties of reflection from natural surfaces. As a technology demonstrator to evaluate the performance of a compact and versatile system, CHRIS has been mounted on board of the European Space Agency (ESA) PRoject for On Board Autonomy 1 (PROBA-1), a small platform provided with advanced pointing capability, which allows the spectrometer to observe the same target at different viewing angles. In this paper anisotropy factors are retrieved from multiangle CHRIS images acquired over San Rossore (Italy) forestry test site during the last year. The spectral behaviour of retrieved factors for different illumination and viewing angles is addressed and analysed. Particular attention is paid to data quality and geometric and atmospheric effects correction. Finally, the deviation of the retrieved factors from the standard assumption of a Lambertian target is analysed.

The Remote Sensing Group at the University of Arizona has been active in the vicarious calibration of numerous sensors through the use of ground-based test sites. Application of these approaches has been limited in the past by the fact that ground-based personnel must be present at the time of the sensor overpass. This work presents the design and implementation of a set of ground-based, ground-viewing radiometers that are deployed without the need for on-site personnel. The radiometers are based on LED detectors allowing them to be robust and inexpensive and combining the results of these measurements with known calibration of the sensors and a suitable surface BRDF model, allows the surface spectral reflectance of the test site to be determined for the sensor overpass. The at-sensor radiance can be predicted via a radiative transfer code using atmospheric data from a fully-automated solar radiometer. Early results from this approach are presented for the Landsat ETM+ and Terra and Aqua MODIS sensors. These results show that errors are currently larger for this method than those with ground-based personnel, but the increased number of calibration opportunities should improve the overall understanding of the sensor calibration.

A small, fast and rugged spectro-radiometer without moving parts for angular resolution for the fast acquisition of large amounts of BRDF data over wavelength to be compiled into libraries is being developed. The system consists of an elliptical mirror which maps a semi-hemisphere onto a CMOS-detector with a dynamic range of 140dB. Radiometric measurements over a wide spectral range from VIS to NIR over 7 deacades (10^-5 W/m² to 100 W/m²) require a pixelwise calibration taking into account the spectral characteristics of the light sources, potentially used filters and the CMOS-detector itself. Also the elliptical mirror has a reflectivity which is a function of the incident angle, the wavelength and polarization of the collected light. All these influences have to be taken into account, if a proper radiometric measurement shall be conducted. The paper deals with the instrument design and mainly with its calibration, and gives some measurement results.

We present a proof-of-concept optical experiment that demonstrates the ability to record squinted broadband RF images formed by a Fourier beamforming phased-array antenna and subsequent squint correction using spatial spectral holography. A cryogenically cooled inhomogeneously broadened absorber (Tm3+:YAG) acts as a spectrally selective holographic medium which records the squinted RF image, covering a wide RF bandwidth (approaching 20 GHz) with resolution of approximately 1 MHz. Subsequently, a frequency-swept laser can read out the squinted image while a magnification-compensating motorized zoom lens synchronously corrects the magnification due to beam squint. Time-integration the image on a CCD detector array produces a squint-compensated broadband RF image, while detection with a MHz bandwidth detector can produce spectral estimates for all sources recorded with this imaging system.

Monitoring the responsivities of the visible channels of the operational Geostationary Operational Environmental Satellites (GOES) is an on-going effort at NOAA. Various techniques are being used. In this paper we describe the technique based on the analysis of star signals that are used in the GOES Orbit and Attitude Tracking System (OATS) for satellite attitude and orbit determination. Time series of OATS star observations give information on the degradation of the detectors of a visible channel. Investigations of star data from the past three years have led to several modifications of the method we initially used to calculate the exponential degradation coefficient of a star-signal time series. First we observed that different patterns of detector output versus time result when star images drift across the detector array along different trajectories. We found that certain trajectories should be rejected in the data analysis. We found also that some detector-dependent weighting coefficients used in the OATS analysis tend to scatter the star signals measured by different detectors. We present a set of modifications to our star monitoring algorithms for resolving such problems. Other simple enhancements on the algorithms will also be described. With these modifications, the time series of the star signals show less scatter. This allows for more confidence in the estimated degradation rates and a more realistic statistical analysis on the extent of uncertainty in those rates. The resulting time series and estimated degradation rates for the visible channels of GOES-8 and GOES-10 Imagers will be presented.

ALISEO (Aerospace Leap-frog Imaging Static Interferometer for Earth Observation) is a prototype of a new imaging interferometer for Earth Observation from the Space. This instrument has been derived from the so called "static interferometers", which do not employ any moving part to optically scan the instrument field-of-view. The device acquires the image of an object with superimposed a pattern of autocorrelation functions of the electromagnetic field coming from each pixel. The complete interferogram, constituted by a system of vertical fringes, is retrieved by moving the observed target with respect to the imaging device. The dependence of the OPD on the radiation-source spectral content, has been addressed performing a set of measurements by spectrally pre-filtering a 600W halogen lamp by means of interference filters with bandwidth of 10nm. We also describe the procedure of pre-elaboration of the acquired data to retrieve the spectrum of at-sensor radiance (dark signal subtraction, spectral instrument response compensation, effects of vignetting and Fourier transform algorithm). Laboratory measurements obtained by ALISEO are presented and discussed. This instrument was partially developed under a contract of Agenzia Spaziale Italiana (ASI).

While hyperspectral imaging systems are increasingly used in remote sensing and offer enhanced scene characterization relative to univariate and multispectral technologies, it has proven difficult in practice to extract all of the useful information from these systems due to overwhelming data volume, confounding atmospheric effects, and the limited a priori knowledge regarding the scene. The need exists for the ability to perform rapid and comprehensive data exploitation of remotely sensed hyperspectral imagery. To address this need, this paper describes the application of a fast and rigorous multivariate curve resolution (MCR) algorithm to remotely sensed thermal infrared hyperspectral images. Employing minimal a priori knowledge, notably non-negativity constraints on the extracted endmember profiles and a constant abundance constraint for the atmospheric upwelling component, it is demonstrated that MCR can successfully compensate thermal infrared hyperspectral images for atmospheric upwelling and, thereby, transmittance effects. We take a semi-synthetic approach to obtaining image data containing gas plumes by adding emission gas signals onto real hyperspectral images. MCR can accurately estimate the relative spectral absorption coefficients and thermal contrast distribution of an ammonia gas plume component added near the minimum detectable quantity.

An ongoing problem for feature extraction in hyperspectral imagery is that such data consumes large amounts of memory and transmittance bandwidth. In many applications, especially on space based platforms, fast, low power feature extraction algorithms are necessary, but not feasible. To overcome many of the problems due to the large volume of hyperspectral data we have developed a fast, low complexity feature extraction algorithm that is a combination of a fast integer-valued hyperspectral discrete wavelet transform (HSDWT) using a specialized implementation of the Haar basis and an improved implementation of linear spectral unmixing. The Haar wavelet transform implementation involves a simple weighted sum and a weighted difference between pairs of numbers. Features are found by using a small subset of the transform coefficients. More refined spatial and/or spectral identifications can then be made by localized fast inverse Haar transforms using very small numbers of additional coefficients in the spatial or spectral directions. The computational overhead is reduced further since much of the information used for linear spectral unmixing is precomputed and can be stored using a very small amount of additional memory.

The linear mixing model is widely used in hyperspectral imaging applications to model the reflectance spectra of mixed pixels in the SWIR atmospheric window or the radiance spectra of plume gases in the LWIR atmospheric window. In both cases it is important to detect the presence of materials or gases and then estimate their amount, if they are present. The detection and estimation algorithms available for these tasks are related but they are not identical. The objective of this paper is to theoretically investigate how the heavy tails observed in hyperspectral background data affect the quality of abundance estimates and how the F-test, used for endmember selection, is robust to the presence of heavy tails when the model fits the data.

Traditionally, synthetic imagery has been constructed to simulate images captured with low resolution, nadir-viewing sensors. Advances in sensor design have driven a need to simulate scenes not only at higher resolutions but also from oblique view angles. The primary efforts of this research include: real image capture, scene construction and modeling, and validation of the synthetic imagery in the reflective portion of the spectrum. High resolution imagery was collected of an area named MicroScene at the Rochester Institute of Technology using the Chester F. Carlson Center for Imaging Science's MISI and WASP sensors using an oblique view angle. Three Humvees, the primary targets, were placed in the scene under three different levels of concealment. Following the collection, a synthetic replica of the scene was constructed and then rendered with the Digital Imaging and Remote Sensing Image Generation (DIRSIG) model configured to recreate the scene both spatially and spectrally based on actual sensor characteristics. Finally, a validation of the synthetic imagery against the real images of MicroScene was accomplished using a combination of qualitative analysis, Gaussian maximum likelihood classification, and the RX algorithm. The model was updated following each validation using a cyclical development approach. The purpose of this research is to provide a level of confidence in the synthetic imagery produced by DIRSIG so that it can be used to train and develop algorithms for real world concealed target detection.

Detection and neutralization of surface-laid and buried landmines has been a slow and dangerous endeavor for military forces and humanitarian organizations throughout the world. In an effort to make the process faster and safer, scientists have begun to exploit the ever-evolving passive electro-optical realm, both from a broadband perspective and a multi or hyperspectral perspective. Carried with this exploitation is the development of mine detection algorithms that take advantage of spectral features exhibited by mine targets, only available in a multi or hyperspectral data set. Difficulty in algorithm development arises from a lack of robust data, which is needed to appropriately test the validity of an algorithm’s results. This paper discusses the development of synthetic data using the Digital Imaging and Remote Sensing Image Generation (DIRSIG) model. A synthetic landmine scene has been modeled after data collected at a US Army arid testing site by the University of Hawaii’s Airborne Hyperspectral Imager (AHI). The synthetic data has been created and validated to represent the surrogate minefield thermally, spatially, spectrally, and temporally over the 7.9 to 11.5 micron region using 70 bands of data. Validation of the scene has been accomplished by direct comparison to the AHI truth data using qualitative band to band visual analysis, Rank Order Correlation comparison, Principle Components dimensionality analysis, and an evaluation of the R(x) algorithm's performance. This paper discusses landmine detection phenomenology, describes the steps taken to build the scene, modeling methods utilized to overcome input parameter limitations, and compares the synthetic scene to truth data.

In an attempt to support the demonstration application of "Spectral Library Of Chinese representative surface substances", an in-situ campaign of spectral measurements was carried out during the summer of 2003 for monitoring the crop stress and planting area in North China Plain. The experiment sites, based on Luancheng Agricultural ecosystem experiment station, Chinese Academy of Sciences, was a 5×5 km area located just south-east of Shijiazhuang, Hebei Province, China. The spectral instrument used is ASD FieldSpecFR with wavelength of 0.35~2.5μm. The experimental achievement includes: soil moisture, corn physiological & biochemical parameters, corn leaf water & chlorophyll-a,b contents, corn structure parameters, field farmland microclimate, aerological exploration data of atmosphere, corn spectra of component, corn crown & background spectra cross the growth period of crop. The paper details the whole experimental scheme and design and partial representation of results of data analysis.

A recently-developed technique of histogram-based segmentation of hyperspectral data allows for a plethora of segmentations. The user can specify the desired number of levels of segmentations, minimum number of pixels defining a peak, and degree of non-linearity in mapping from principal component floating values to histogram bins, all of which affect the derived segmentation. In the present work, we seek to extend previous work which arrives at a small range of clusters or segmentation levels from the image itself. We seek within this range to find "better" segmentations or possibly a unique representative segmentation. The method employed to achieve this goal starts with an over-fine segmentation, i.e. more segmentation levels than needed, and uses quantitative metrics to measure the "quality" of that segmentation and to guide a compression into a reduced segmentation. If the method has merit, different starts should compress down into comparable segmentations. Therefore a measure to establish the similarity of two or more segmentations was developed. Different quantitative metrics were studied and several modes of compression were examined. Some impressive results are presented but the methods are still not robust with respect to segmentation starts and are image dependent as to the best modes of compression.

Fisher's linear discriminant analysis (FLDA) has been widely used in pattern classification due to its criterion, called Fisher's ratio, based on the ratio of between-class variance to within-class variance. Recently, a linear constrained discriminant analysis (LCDA) was developed for huperspectral image classification where Fisher's ratio was replaced with the ratio of inter-distance to intra-distance and the target signatures were constrained to orthogonal directions. This paper directly extends the FLDA to constrained Fisher's linear discriminant analysiss (CFLDA), which uses Fisher's ratio as a classification criterion. Since CFLDA is supervised which requires a set of training samples, this paper further extends the CFLDA to an unsupervised CFLDA (UCFLDA) by including a new unsupervised training sample generation algorithm to automatically produce a sample pool of training data to be used for CFLDA. In order to determine the number of classes, p, to be classified, a newly developed concept, called virtual dimensionality (VD) is used to estimate the p where a Neyman-Pearson-based eigen-analysis approach developed by Harsanyi, Farrand and Chang, called noise-whitened HFC (NWHFC)'s method, is implemented to find the VD. The experimental results have shown that the proposed UCFLDA perform effectively for HYDICE data and provides a promising unsupervised classification technique for hyperspectral imagery.

Unsupervised image classification for remotely sensed imagery is very challenging due to the fact that the unknown image background generally varies with a wide range of spectral deviations. Additionally, spectral similarity among subtle small calsses also causes tremendous difficulty in classification. This paper investigates three major issues, (1) image background removal, (2) generation of training sample data, (3) determination of the number of classes to be classified, p which are encountered in unsupervised image classification. The study on these three issues is conducted via a well-known Airborne Visible InfraRed Imaging Spectrometer (AVIRIS) image scene, Indiana Pine test site available online at Purdue University's website. Since image background varies with different applications, it is generally difficult to perform background removal without prior knowledge. In order for unsupervised classification to be effective, a good set of training data is also necessary. These training samples must be generated directly from the image data in an unsupervised manner. This paper develops an unsupervised training sample generation algorithm (UTSGA) that can generate a good sample pool of training data for supervised classification. In determining p, a newly developed concept, called virtual dimensionality (VD) is used to estimate the p where a Neyman-Pearson-based eigen-analysis approach developed by Harsanyi, Farrand and Chang, called noise-whitened HFC (NWHFC)'s method, is implemented to find the VD to be used for the p. Finally, an unsupervised image classification algorithm can be derived by implementing a supervised classifier in conjunction with teh UTSGA algorithm and NWHFC's method.

Independent component analysis (ICA) is a popular approach to blind source separation. In this paper, we investigate its application to hyperspectral image classification. In particular, the performance of Joint Approximate Diagonalization of Eigenmatrices (JADE) algorithm is studied. The major advantage of using ICA is its capability of classifying objects with unkown spectral signatures in an unkown image scene, i.e., unsupervised classification. However, ICA suffers from computational expensiveness, which limits its application to high dimensional data analysis. In order to make it applicable to hyperspectral image classification, a data preprocessing procedure is employed to select the most important bands based on the band image quality. The number of bands ought to be selected is predetermined by an estimation method. The preliminary results from experiments demonstrate the potential of ICA in conjunction with band selection to unsupervised hyperspectral image classification.

In this paper we inestigate the application of independent component analysis (ICA) to multispectral image classification. In particular, the performance of Joint Approximate Diagonalization of Eigenmatrices (JADE) algorithm is studied. ICA is particularly useful for classifying objects with unknown spectral signatures in an unkown image scene, i.e., unsupervised classification, because it does not require any prior information about class signatures. Since the weight matrix in ICA is a square matrix for the purpose of mathematical tractability, the number of objects that can be classified is equal to the data dimensionality, i.e., the number of spectral bands. When the number of spectral bands is very small (e.g., 3-band CIR photograph), it is impossible to classify all the different objects present in an image scene with the original data. In order to solve this problem, we present a data dimensionality expansion technique to generate artificial bands for additional spectral measurements. The results from such a nonlinear band generation approach are compared with a linear band generation method using cubic spline interpolation of pixel spectral signatures. The experiments demonstrate that such a nonlinear band generation approach can expand the applicability of ICA and improve the classification accuracy.

Current image quality approaches are designed to assess the utility of single band images by trained image analysts. While analysts today are certainly involved in the exploitation of spectral imagery, automated tools are generally used as aids in the analysis and offer hope in the future of significantly reducing the analysis timeline and analyst work load. Thus, there is a recognized need for spectral image quality metrics that include the effects of automated algorithms. Previously, we have reported on candidate approaches for spectral quality metrics in the context of unresolved object detection. We have continued these efforts through the use of empirical trade studies in the context of ground cover terrain classification. HYDICE airborne hyperspectral imagery have been analyzed for the effects on scene classification accuracy of spatial resolution, signal-to-noise ratio, and number of spectral channels. Various classification algorithms including Gaussian maximum likelihood, spectral angle mapper, and Euclidean minimum distance have been considered. Performance metrics included classification accuracy, confusion matrices, and the Kappa coefficient. An extension of the previously developed Spectral Quality Equation (SQE) has been developed for the terrain classification application. As expected, the accuracy of terrain classification shows only modest sensitivity to the parameters considered, except at the extreme cases of high noise, few bands, and small ground resolution. However, these results are useful in continuing to develop the quantitative relationships necessary for characterizing the quality of spectral imagery in various applications.

Basing ourselves on a novel segmentation algorithm for multi-spectral images, we have considered how to detect multi-pixel anomalous objects in image cubes with a spectral component. In particular, we have developed several filters to compensate for noise which may be present in the initial cube. We show that for speckle noise, a modification of our morphology technique allows us to detect targets without an enhanced false alarm result.

The ellipsometry is known as high precision metrology for thin film thickness measurements and its optical properties by measuring ellipsometric parameters, ψ and Δ, defined as amplitude and phase values of the ratio of Fourier reflection coefficients for p- and s-polarized light. With conventional ellipsometers, we can get average values of ellipsometric parameters in the region of interest determined by spot size of measurement beam. However, we can expand the measurement scheme to two dimensional spectral imaging with additional imaging spectrograph compatible to the structure of ellipsometer. That is, we can simultaneously get spatial and spectroscopic ellipsometric parameters using two dimensional imaging detectors. Using this type of ellipsometers, polarization state dependent response of imaging spectrograph must be considered carefully during azimuth calibration procedures as well as ellipsometric parameters measurement. In this paper, we suggest Jones calculus model for ellipsometer with considering dichroic response in spectrograph and background signal levels in detector. And we show experimental calibration results comparison with that of simulation using suggested Jones calculus model.

The greedy modular eigenspaces (GME) has shown effective in hyperspectral feature extraction. The GME was developed by grouping highly correlated hyperspectral bands into a smaller subset of band modular regardless of the original order in terms of wavelengths. It utilizes the inherent separability of different classes in hyperspectral images to reduce dimensionality and further to generate a unique GME feature. This paper takes advantage of the GME to develop a GME-based band selection (GMEBS) for hyperspectral imagery. It selects a subset of non-correlated hyperspectral bands for hyperspectral images using the unique ability of the GME in class separability. The proposed GMEBS algorithm provides a fast procedure to select the most significant features and speeds up the distance decomposition compared to GME features. It also avoids the bias problems of transforming the information into linear combinations of bands as does the traditional principal components analysis (PCA). The proposed GMEBS approach selects each band by a simple logical operation, call GME feature scale uniformity transformation (GME/FSUT), to include different classes into the most common feature modular subset of bands. Interestingly, experimental results show that this simple GMEBS approach is very effective and can be used as an alternative to other band selection algorithms.

Hyperspectral spectroscopy can be used remotely to measure emitted radiation from minerals and rocks at a series of narrow and continuous wavelength bands resulting in a continuous spectrum for each pixel, thereby providing ample spectral information to identify and distinguish spectrally unique materials. Linear mixture modeling ("spectral unmixing"), a commonly used method, is based on the theory that the radiance in the thermal infrared region (8-12 μm) from a multi-mineral surface can be modeled as a linear combination of the endmembers. A linear mixture model can thus potentially model the minerals present on planetary surfaces. It works by scaling the endmember spectra so that the sum of the scaled endmember spectra matches the measured spectrum with the smallest "error" (difference). But one of the drawbacks of this established method is that mathematically, a fit with an inverted spectrum is valid, which effectively returns a negative abundance of a material. Current models usually address the problem by elimination of endmembers that have negative scale factors. Eliminating the negative abundance problem is not a major issue when the endmembers are known. However, identifying unknown target composition (like on Mars) can be a problem. The goal of this study is to improve the understanding and find a subsequent solution of the negative abundance problem for Mars analog field data obtained from airborne and ground spectrometers. We are using a well-defined library of spectra to test the accuracy of hyperspectral analysis for the identification of minerals on planetary surfaces.