KEYWORDS: Transform theory, Image fusion, Target detection, Simulation of CCA and DLA aggregates, Sensors, Near infrared, Hyperspectral imaging, Vegetation, Cesium, Hyperspectral target detection
This study examines normalizing the imagery and the optimization metrics to enhance anomaly and change detection,
respectively. The RX algorithm, the standard anomaly detector for hyperspectral imagery, more successfully extracts
bright rather than dark man-made objects when applied to visible hyperspectral imagery. However, normalizing the
imagery prior to applying the anomaly detector can help detect some of the problematic dark objects, but can also miss
some bright objects. This study jointly fuses images of RX applied to normalized and unnormalized imagery and has a
single decision surface. The technique was tested using imagery of commercial vehicles in urban environment gathered
by a hyperspectral visible/near IR sensor mounted in an airborne platform. Combining detections first requires
converting the detector output to a target probability. The observed anomaly detections were fitted with a linear
combination of chi square distributions and these weights were used to help compute the target probability. Receiver
Operator Characteristic (ROC) quantitatively assessed the target detection performance. The target detection
performance is highly variable depending on the relative number of candidate bright and dark targets and false alarms
and controlled in this study by using vegetation and street line masks. The joint Boolean OR and AND operations also
generate variable performance depending on the scene. The joint SUM operation provides a reasonable compromise
between OR and AND operations and has good target detection performance. In addition, new transforms based on
normalizing correlation coefficient and least squares generate new transforms related to canonical correlation analysis
(CCA) and a normalized image regression (NIR). Transforms based on CCA and NIR performed better than the standard
approaches. Only RX detection of the unnormalized of the difference imagery in change detection provides adequate
change detection performance.
Finding, tracking and monitoring events and activities of interest on a continuous basis remains one of our highest
Intelligence Surveillance and Reconnaissance (ISR) requirements. Unmanned Aerial Systems (UAS) serve as one of the
warfighter's primary and most responsive means for surveillance and gathering intelligence information and are
becoming vital assets in military operations. This is demonstrated by their significant use in Afghanistan during
Operation Enduring Freedom and in Iraq as part of Operation Iraqi Freedom. Lessons learned from these operations
indicate that UAVs provide critical capabilities for enhancing situational awareness, intelligence gathering and force
protection for our military forces. Current UAS high resolution electro-optics offers a small high resolution field of
view (FOV). This narrow FOV is a limiting factor on the utility of the EO system. The UAS that are available offer
persistence; however, the effectiveness of the EO system is limited by the sensors and available processing.
DARPA is addressing this developing the next generation of persistent, very wide area surveillance with the
Autonomous Real-time Ground Ubiquitous Surveillance - Imaging System (ARGUS-IS). The system will be capable of
imaging an area of greater than 40 square kilometers with a Ground Space Distance (GSD) of 15 cm at video rates of
greater than 12 Hz. This paper will discuss the elements of the ARGUS-IS program.
Objects shielded from direct illumination, or lying in shadows, can be difficult to detect using airborne hyperspectral
sensors. Diminished illumination of objects reduces the signal contrast with respect to the background and shade shifts
the spectral signature distribution. Supervised detection of shadowed objects is therefore confounded from implementing
the simplest approach, namely inserting signatures trained on fully illuminated objects into target searches. Previously
developed statistical temporal transformation ("whitening/dewhitening") of target signatures and target covariance
matrices has been adapted to convert fully illuminated signatures to the more appropriate shadowed signatures for target
detection. The choice of areas to transform the signatures must include dimilar background composition under full
illumination and shadow conditions. A new search algorithm, Regularized Maximum Likelihood Clustering (RMLC),
uses pixels for the CV computation associated with the object. "Regularizing" the object's covariance matrix avoids non-singularities from the CV computation and mitigates statistical degradation for the covariance matrix calculation due to
undersampling of the small number of pixels. To accurately compute the required covariance matrices from imagery of
small open and shadowed areas, "regularization" is also applied to the covariance matrices associated with those areas.
The searches are applied to visible/near IR data collected from forested areas. Inserting the transformed signatures into
RMLC and the adaptive cosine estimator (ACE) achieved higher target detection for fixed alarm rate, relative to the
matched filter. The temporal transform of the signatures was compared to a scaling approach using mean signatures from
the open and shadowed areas. This study successfully extracted targets from shadows by using a sensitive target search
and through transforming signatures collected from fully illuminated conditions into shadowed spectra.
This study adapts a variety of multi-spectral image classification techniques to generate supervised object detection algorithms for hyperspectral imagery, and compares and quantitatively tests them against the Adaptive Cosine Estimator (ACE) and the standard, matched filter (MF). A new search algorithm, Regularized Maximum Likelihood Clustering (RMLC), uses only pixels for the covariance matrix (CV) computation associated with the object after "regularizing" the matrix to avoid singularities and mitigate statistical degradation due to undersampling for small objects. The searches are applied to both visible/near IR and short wave IR data collected from forested areas. This study tests the detection sensitivity by using object signatures and CVs taken directly from the scene and from temporally transformed signatures and object CVs. This study adds simple, high performing algorithms to the small object search arsenal.
A previous study adapted a variety of techniques derived from multi-spectral image classification to find objects amid
cluttered backgrounds in hyperspectral imagery. That study quantitatively compared the adapted algorithms against a
standard object search, the matched filter (MF) and a recently developed object detector, Adaptive Cosine Estimator
(ACE) and found substantial reduction in false alarm rates for a given target detection probability. One adapted object
search, Regularized Maximum Likelihood Classifier (RMLC), requires calculating the covariance matrix involving the
average object spectral signature and the target pixels. The object covariance matrix requires a relatively large number of
pixels to generate a non-singular, accurate covariance matrix. This study examines the robustness of the RMLC
algorithm on number of training pixels, the optimal mixing covariance matrices, and choice of object subspaces for the
ACE algorithm. The tests were applied to visible/near IR data collected from forest and desert environments. This study
finds that high detection performance standards for RMLC are invariant for pixel number for homogenous targets, down
to two pixels. Regularization is relatively unaffected by the choice of areas to optimize the object covariance matrix
although targets that mix background appear to be more sensitive to choice of covariance matrix. Reducing the object
subspace dimensions by using the average target signature or choosing the first principle component enhances ACE
performance relative to using the entire object space.
This study adapts a variety of techniques derived from multi-spectral image classification to find objects amid cluttered backgrounds in hyperspectral imagery. This study quantitatively compares the algorithms against a standard object search, the matched filter (MF) and recently developed object detector, Adaptive Cosine Estimator (ACE). These object searches require calculating the Mahalanobis distance between the average object spectral signature and the test pixel spectrum and needs the computation of a covariance matrix. The covariance matrix is generated using the entire image (Whitened Euclidean Distance, WED) or using pixels associated with the object (Maximum Likelihood Classifier, MLC). The latter computation requires a relatively large number of pixels to generate a non-singular, accurate covariance matrix. Regularizing object pixels via optimally mixing (likelihood maximization) diagonal, object, and entire image covariance matrices to generate the object covariance matrix estimate. This approximation is called the Regularized Maximum Likelihood Classifier (RMLC). The object searches MF, ACE, WED, MLC, and RMLC were applied to visible/near IR data collected from forest and desert environments. This study searched for objects using object signatures and covariance matrices taken directly from the scene and from statistically transformed object signatures and covariance matrices from another time. This study found a substantial reduction in the number of false alarms (factor of 10 to 1000) using WED, ACE, RMLC relative to MF searches for the two independent data collects. The regularization of in-scene and transformed covariance matrices substantially reduced false alarms relative to using unprocessed covariance matrices. This study adds simple, high performing algorithms to the object search arsenal.
We present results from an improved ORASIS (Optical Real-time Adaptive Spectral Identification System) hyperspectral-data compression-algorithm that is being implemented on the Naval EarthMap Observer (NEMO) satellite. The algorithm is shown to produce results that are statistically improved from previous findings. To augment the statistical testing, the re-inflated data are run through analysis programs such as unsupervised classification. ORASIS compression is a series of algorithms. The first algorithm, the exemplar selector process (ESP), is a variation of Learned Vector Quantization (LVQ) that builds up a relatively small set of spectra to represent the full data set. Subsequent algorithms find approximate endmembers for the exemplar set and project the set into the space defined by the endmembers. Both the ESP and the projection process contribute to the compression of the data. The obtainable compression ratios vary with scene content and other factors but ratios between 10:1 and 30:1 are possible. The compressed data format is designed to allow direct access to individual pieces of the data without reinflation of the entire data set. Details of the hardware implementation of the Imagery On-Board Processor (IOBP) of NEMO is discussed, as well as the use of the compressed data on the ground.
The Ocean Portable Hyperspectral Imager for Low-Light spectroscopy (Ocean PHILLS), is a new hyperspectral imager specifically designed for imaging the coastal ocean. It uses a thinned, backside illuminated CCD for high sensitivity, and an all-reflective spectrograph with a convex grating in an Offner configuration to produce a distortion free image. Here we describe the instrument design and present the results of laboratory calibration and characterization and example results from a two week field experiment imaging the coastal waters off Lee Stocking, Island, Bahamas.
KEYWORDS: Data processing, Satellites, Data compression, Spectrographs, Short wave infrared radiation, Sensors, Hyperspectral imaging, Space operations, Signal to noise ratio, Signal processing
The primary mission of the Naval EarthMap Observer (NEMO) is to demonstrate the importance of hyperspectral imagery in characterizing the littoral battlespace environment and littoral model development. NEMO will demonstrate real time on-board processing and compression of hyperspectral data with real-time tactical downlink of ocean and surveillance products directly from the spacecraft to the field. The NRL's Optical Real-time Adaptive Spectral Identification System (ORASIS) will be deployed on a 3.8 Gflop multiprocessing computer, the Imagery On-Board Processor (IOBP), for automated data analysis, feature extraction and compression. NEMO's wide area coverage (106 km2 imaged per day), as well as power and cost constraints require data compression between 10:1 and 20:1. The NEMO Sensor Imaging Payload (SIP) consists of two primary sensors: first, the Coastal Ocean Imaging Spectrograph (COIS) is a hyperspectral imager which records 60 spectral bands in the VNIR (400 to 1000 nm) and 150 bands in the SWIR (1000 to 2500 nm), with a GSD of either 30 or 60 meters; and second, the 5 m GSD Panchromatic Imaging Camera (PIC). This paper describes the design and implementation of the data processing hardware and software for the NEMO satellite.
Recent advances in large format detector arrays and holographic diffraction gratings have made possible the development of imaging spectrographs with high sensitivity and resolution, at relatively low component cost. Several airborne instruments have been built for the visible and near IR spectral band with 10-nm resolution, and SNR of 200:1. Three instruments are compared, an all-reflective spectrography using a convex grating in an Offner configuration, and two off-the-shelf transmission grating spectrographs using volume holograms. The camera is a 1024 X 1024 frame transfer, back-thinned CCD, with four taps for obtaining high frame rates. Performance and scan data is presented and compared to the design for image quality, distortion, and throughput.
We compare the results produced by the NRL ORASIS algorithm with those produced by ENVI's Pixel Purity Index. Both procedures attempt to find appropriate estimations of the constituent endmembers.
Recent advances in large format detector arrays and holographic diffraction gratings have made possible the development of imaging spectrographs with high sensitivity and resolution, ideally suited for space-based remote sensing of earth resources. An optical system composed of dual spectrographs and a common fore-optic has been designed for the visible-near infrared (VNIR) and shortwave bands with 10-nm spectral resolution, providing 30-meter ground resolution from an altitude of 605 km. The spectrograph designs are based on a modified Offner 1-X relay with spherical mirrors and a convex spherical holographic grating for the secondary mirror. The fore-optic is a three-mirror anastigmatic telescope with a 360-mm focal length to match the pixel pitch of the respective 1024 X 1024 visible silicon CCD and SWIR HgCdTe FPAs. The primary advantages of this design are the relatively low f-number (f/3), large flat field (18 mm), and low distortion. Preliminary performance results of a VNIR testbed grating and spectrograph are presented and compared to the design predictions.
In this work, we generate ROC curves on real and synthetic scenes and develop scoring methods to evaluate the performance of the ORASIS hyperspectral algorithm. The goal of this effort is to improve the overall performance of ORASIS, focusing on the endmember selection methods. ROC curve evaluations have been performed on hyperspectral data sets from different scenes. We have scored by target and by target pixel. A scene generator has been developed allowing many features: combination of real or synthetic background and multiple, distinct targets; user-defined angle of target spectrum to background subspace; and user-specified non-uniform target/background transparency.
A multiprocessor version of the ORASIS hyperspectral analysis program has been implemented in support of the ASRP. In brief, the long-term technical objectives of the ASRP are to demonstrate the feasibility and military utility of real-time target detection from uncrewed air vehicles using hyperspectral data. This paper presents a preliminary assessment of ORASIS performance and describes the ORASIS development effort designed to meet the ASRP goals. Real-time performance of the analysis program and its potential effectiveness as a target detection method are demonstrated.
The covered lantern project was initiated by the central MASINT Technology Coordination Office to demonstrate the tactical use of hyperspectral imagery with real time processing capability. We report on the design and use of the HYCORDER system developed for Covered Lantern that was tested in June 1995. The HYCORDER system consisted of an imaging spectrometer flying in a Pioneer Uncrewed Aeronautical Vehicle and a ground based real time analysis and visualization system. The camera was intensified allowing dawn to dusk operation. The spectral information was downlinked to the analysis system as standard analog video. The analysis system was constructed from 17 Texas Instrument C44 DSPs controlled by a 200 MHz Pentium Pro PC. A real time, parallel version of NRL's optical real-time adaptive spectral identification system algorithm was developed for this system. The system was capable of running continuously, allowing for broad area coverage. The algorithm was adaptive, accommodating changing lighting conditions and terrain. The general architecture of the algorithm will be discussed as well as results from the test.
The ability to detect weak targets of low contrast or signal-to- noise ratio (SNR) is improved by a fusion of data in space and wavelength from multispectral/hyperspectral sensors. It has been demonstrated previously that the correlation of the clutter between multiband thermal infrared images plays an important role in allowing the data collected in one spectral band to be used to cancel the background clutter in another spectral band, resulting in increased SNR. However, the correlation between bands is reduced when the spectrum observed in each pixel is derived from a mixture of several different materials, each with its own spectral characteristics. In order to handle the identification of objects in this complex (mixed) clutter, a class of algorithms have been developed that model the pixels as a linear combination of pure substances and then unmix the spectra to identify the pixel constituents. In this paper a linear unmixing algorithm is incorporated with a statistical hypothesis test for detecting a known target spectral feature that obeys a linear mixing model in a mixture of background noise. The generalized linear feature detector utilizes a maximum likelihood ratio approach to detect and estimate the presence and concentration of one or more specific objects. A performance evaluation of the linear unmixing and maximum likelihood detector is shown by comparing the results to the spectral anomaly detection algorithm previously developed by Reed and Yu.
We report the results of a tradeoff study for the selection of the number of wavelength bands and resolution needed in a hyperspectral data set in order to separate a scene into its constituent features. This separation is accomplished by finding approximate endmembers using convex mixing and shrink-wrapping techniques. This and related techniques are referred to as NRL's Optical Real-time Adaptive Spectral Identification System (ORASIS). ORASIS's algorithms will be briefly described. Once endmembers are found, matched filters are calculated which can then be used to separate (or demix) the scene. We have analyzed synthetic cubes, cubes acquired by NRL's Portable Hyperspectral Imager for Low Light Spectroscopy (PHILLS) sensor, and cubes from other sensors. PHILLS consists of multiple hyperspectral sensors that operate in pushbroom mode. PHILLS has been deployed from aircraft and on the ground in a variety of terrains from the polar icecap to the Florida Keys. The majority of the data were recorded with a 16-bit thermo-electrically cooled camera which records 1024 wavelengths over the range of 200 to 1100 nm. Major features of the scene can be successfully demixed using fewer than 1024 wavelength bands. However, preliminary evidence suggests that finer features require the full wavelength range and resolution.
The portable hyperspectral imager for low light spectroscopy (PHILLS) instrument consists of several modules containing analog and digital imaging spectrometers, two of which are intensified, covering the 200 nm to 1100 nm wavelength range with over 100 wavelength bands. The PHILLS instrument is usually flown aboard P-3 Orion aircraft at altitudes from 500 feet to 10,000 feet. PHILLS ground images are acquired with > 70 degrees FOV and 2-5 meter spatial resolution, and 0.5-1 nm wavelength resolution. Hyperspectral data cubes are processed using a combination of spectral matching techniques and the filter vector algorithm, to produce terrain separation and spectral dimensionality; i.e. variability of the spectral signatures for individual 'substances' in the image. Data obtained on flights over the Florida Keys showing land and underwater features are presented.
We report recent progress using a filter vector technique to analyze the data from a hyperspectral image. The filter vector technique finds the optimal filter vectors for demixing the complex patterns found in the hyperspectral image. The method has the potential to be implemented in real time since it is fully parallel. Computation of the filter vectors for a given family of known species vectors is fast and direct and improved algorithms for developing of the algorithm which may be updated as conditions change is possible. Advantages of using the filter vector techniques over the technique of pattern matching will be discussed. The portable hyperspectral images for low light spectroscopy (PHILLS) instrument has been used on a number of depolyments in the last year. Typically, the instrument files on the Naval Research Laboratory's P-3 Orion aircraft. Currently, the PHILLS instrument records over 1000 wavelength bands between UV and near IR. Results from a number of deployment and test situations is shown.
KEYWORDS: Imaging systems, Hyperspectral imaging, Sensors, Optical filters, Analog electronics, Computing systems, Control systems, Cameras, Data analysis, Video
The portable hyperspectral imager for low light spectroscopy (PHILLS) is a modular system incorporating panchromatic and spectroscopic imagers. PHILLS was designed to be a low cost, versatile hyperspectral imager using primarily commercial off-the-shelf parts that could be used as a ground based or airborne imager. The goals fo the PHILLS program are to allow collection of high quality broadband hyperspectral data, to develop methodology for near real time spectral unmixing, to make deployment of hyperspectral imagers feasible in airborne or spaceborne scenarios. Presently PHILLS combines high resolution color video cameras with intensified UV, visible, and NIR spectroscopic modules with wavelength resolution of 0.5-2 nm over the 0.2-1.1 micrometers band. The system offers analog and digital data recording capability with GPS annotation. The instrument has been successfully deployed on NRL's P-3 aircraft in missions from the North Pole to the Florida Keys. The PHILLS system is controlled by a personal computer and is integrated with an adaptive hyperspectral data analysis system that provides very fast end member determination and subpixel spectral unmixing using an orthogonal projection based method known as the filter vector algorithm.
Hyperspectral imaging can be a powerful tool for remote sensing of geologic, biologic, and ocean surfaces of atmospheres, and of rocket plumes. A hyperspectral imager provides a 3D (two spatial and one spectral) description from a sequence of 2D images. Common hyperspectral approaches using narrow band filters or imaging spectrometers are inefficient because photons outside the filter passband or the slit area are not detected. A new imaging technique called spectro-tomography collects all available photons and relies on computer tomography to reconstruct the 3D data cube of the image. A rotational spectro-tomographic (RST) imager is designed with a wide aperture, objective-grating camera, that is rotated in steps around its optical axis. The 2D projections of the object are analyzed using methods based on Fourier transforms. Both direct Fourier methods and filter-backprojection algorithms have been developed for 3D tomographic analysis. Numerical methods are employed to simulate and reconstruct a broad spectrum object with 64 spectral bands and 64 X 64 spatial resolution elements. For this example, the photon flux at the detector of the RST imager is 64 times that of a conventional spectral imager. The rotational spectro-tomographic imager has applications to detection of natural and artificial atmospheric emissions where large photon through-put is required.
SuperIBEX produces 5 MeV electron beams with currents of 5 - 30 kA for atmospheric propagation experiments. Such beams are subject to the resistive hose instability which quickly disrupts propagation unless the beam is properly conditioned prior to injection into the atmosphere. The most successful conditioning scheme to date employs an ion-focused regime (IFR) cell to tailor the beam emittance followed by an active wire cell to center the beam. The emittance tailoring tends to 'detune' the hose instability, while the wire cell damps transverse perturbations which seed the instability. A 17 kA, 1.5 cm radius beam has been propagated in full density air to the end of the 5 m long tank without disruption. The beam has also been propagated in a density channel produced by a laser-guided electric discharge. The channel reduces beam expansion generated by scattering of beam electrons by the atmosphere. Results will be compared with simulation codes and with previous propagation experiments at other laboratories.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.