The detection of land mines and other ordnance on the battlefield has grown in importance with their increased use, not only for military personnel, but for civilians after hostilities have ceased. The need for new approaches and sensors to increase the speed and efficiency of methods to clear mines is an issue that must be addressed. A method to detect metal mines, on top of or buried under dry sand, is demonstrated using the passive detection of naturally occurring millimeter wave radiation (at 44 GHz) emanating from the scene. Measurements will be shown that indicate the feasibility of detection of metal under at least 3 inches of dry sand.

Preliminary results have shown that cross polarization enhances mine detection capabilities. This paper contains data from experimental and simulated testing, analysis of the data, and conclusions on effects of polarization techniques for mine detection. Cross polarization of the transmit and receive antennas significantly diminishes the surface reflection integrated at the receive antenna. The angle of the transmit and receive antennas in a cross polarization configuration is inherently 90 degrees; at different azimuth angles, a contrast in the backscattered energy from a mine and from the air/ground interface above the mine has been observed. To further understand this phenomenon, Kaman Sciences has configured a bistatic, dipole, impulse system operating at 2.5 GHz center frequency. The transmitter and receiver are positioned at specific down-looking angles to the ground. The receiver is stepped around the azimuth of the transmitter for co- and cross-polarizations. The Kaman Sciences system and experimental data collection will be discussed in the paper. Specifically, Kaman Sciences collected data with both metallic and nonmetallic mine simulants or targets. The targets were placed on the surface and buried 2 inches in 3 different types of soils: sand, sand mixed with gravel, and loam-rich top soil. Free space measurements of mines were modelled using a method of moments code that was specifically modified to account for the ultra wideband nature of the transmitter waveform. The resulting data was analyzed for the optimum polarization angle according to the target and compared with experimental results. The paper will contain the experimental and model data, results, and conclusions with respect to the benefits to mine detection.

Microwave imaging has recently been presented as a promising technique for detecting buried and foliage-obscured objects. In preparation of a project for the detection of anti-personnel mines (APMs), first test measurements have been performed at the European Microwave Signature Laboratory (EMSL), which is a measurement system based on an ultrawide bandwidth polarimetric SAR. The followed approach for the detection of small plastic mines together with some examples of the resulting radar images are presented in this paper.

A novel polarimetric ultra-wideband radar system operating in the 1-2 GHz frequency for subsurface probing applications is currently under development at the University of Nebraska. The radar system transmits white Gaussian noise. Detection and localization of buried objects is accomplished by correlating the reflected waveform with a time-delayed replica of the transmitted waveform. Broadband dual-polarized log-periodic antennas are used for transmission and reception. A unique signal processing scheme is used to obtain the target's polarimetric amplitude and phase response by frequency translation of the high depth resolution, low bandwidth-duration product, as well as simplified signal processing. This paper describes the unique design features of the radar system, develops the theoretical foundations of noise polarimetry, and provides experimental evidence of the polarimetric and resolution capabilities of the system.

The Balanced Bridge (BB) detection concept was developed just after the end of WWII. It has been researched for many years since then but it has never truly overcome the following inherent problems: sensitivity to antenna height and tilt variations, detectability of flush mines, sensitivity to soil moisture content, high false alarms, and most importantly, the inability to detect small anti-personnel (AP) mines. Even with all of these shortcomings, the BB sensor technology is still one of the most promising electrmagnetic mine detection systems. This paper will address a new BB detector and its preliminary field performance compared to earlier BB research. The new BB detector has superior capabilities compared to earlier BB efforts involving single frequency or single octave excitation because the new BB operates over a multi-octave bandwidth. The new BB detector also incorporates audio and visual presentations of digitally processed signals where earlier versions only had an audible announcement derived from a simple thresholding algorithm. New BB designs addressing previous BB deficiencies will also be discussed. Design changes include using a broadband printed circuit board antenna, RF transmit and receive components, and a digital signal processor. This new BB detector will be tested at an Advanced Technology Demonstration (ATD) evaluation in FY95. The ATD exit criteria will be discussed and compared to recent field testing of the new BB detector. Preliminary results with the new BB system have demonstrated encouraging results which will be incorporated in this paper.

Over the years, many different sensor types have been evaluated in an attempt to satisfy the need to detect and discriminate tactical and strategic targets concealed in foliage or underground. In large measure these early efforts were disappointing because of the lack of appropriate technologies. Today, by taking advantage of commercial off-the-shelf processors, an advanced analog-to-digital (A/D) converter, and lessons learned, a highly capable impulse radar has been designed and assembled to investigate an ultra-wideband (UWB) radar approach for ground penetration (GPEN) radar studies. The testbed consists of several major subsystems that are modular to allow for the evaluation of alternate approaches. The testbed radar subsystem consist of the antenna, the transmitter, the A/D converter, the processor/data storage system, the timing and control assembly, the positioning subsystem, and the operator interface computer. Many of the subassemblies exist as standard 19 inch rackmount units or as VME-compatible printed circuit assemblies. Much of the system operation is controlled by software, allowing easy modifications or other future upgrades. Data collected with this upgraded system will be used for measuring and analyzing the basic phenomenology of radar propagation through the ground and the response of targets, clutter, and targets embedded in clutter. One important aspect of basic phenomenology studies is validation of models with data. Range profiles of synthetic aperature radar (SAR) processed data from the Army Research Laboratory UWB radar is compared to 3D method of moments models for similar targets. In this paper, a mix of canonical and mine-like targets are examined and compared. Comparison between data and models shows some correlation, thus validating the need for further investigation.

The Lincoln Laboratory ground-based UWB Rail SAR was used to collect UHF and L-band data on a variety of mine-like targets. The target set consisted of metal pipes, bomb fragments, and M-20 metallic anti-tank mines, above and below ground. Mostly co-polarized data was collected for depression angles between 10 and 30 degrees. Imagery of the targets in different frequency sub-bands are shown and RCS characteristics are quantified.

The objective of the Advanced Mine Detection Radar Program is to demonstrate the capability of the vehicular-mounted, frequency agile radar to detect, identify and locate metallic and nonmetallic, buried and surface land mines at a stand-off distance of about 10 meters (m). This wideband, operating frequency system consists of one transmit and two receive horns, transmit and receive circuitries, an IBM-compatible computer (PC), and a Sun computer. The transmit horn generates a train of pulsed continuous-wave (CW) signals in 36 stepped operating frequencies. These frequencies are used to resonate all nonmetallic and metallic mines. At each cycle of these pulses, while the signal of the transmit horn is off, the receive horns receive the echo, the signal of soil, roots, rocks, and targets. This echo is then mixed with the two I/Q circuits to produce the in-phase and quadrature-phase signals. These signals are then low-pass filtered, amplified, and digitized signals while the PC is acquiring a new data set at the next operating frequency. At each frequency, the system noise and the clutter of the digitized signals will be reduced by the averaging and smoothing algorithms. After all 36 frequencies have been transmitted and the data preprocessed, an anomaly will be detected, located and identifies by the data processing algorithms. The results from the final field test of this program shows a 100 percent detection with an average of 36 percent identification and 17 false alarms per 78 m² in a condition of nearly 5 percent of the moisture content in the soil.

The measurement results from the Technology Assessment for the Close-in Man Portable Mine Detection (CIMMD) program at Fort A.P. Hill, Virginia, of the Hand Buried Ordanance and Mine Detection System (HBO-MDS) are presented. The HBO-MDS was developed by Coleman Research Corportion (CRC) under a contract with the US Army Night Vision & Electronic Sensors Directorate, Mine Detection Division.

Evaluating performance of microwave-based handheld mine detection sensors has until recently been limited to using tethered test carts instrumented with a network analyzer and computer controller. The Army's Night Vision and Electronic Sensors Directorate's (NVESD) Mine Detection Division has completed the development of a versatile portable data acquisition system (DAS) enabling researchers to simulate handheld mine detection systems and to evaluate mine detection antennas from 500 MHz to 2000 MHz. The data that is collected by the DAS is in the form of the complex S₂₁ and S₁₁ scattering parameters and includes sensor head position. System software is in a data collection instrument useful for evaluating the performance of detection algorithms and antennas, or as a fully operational handheld mine detector brassboard. Described in this paper are the technologies comprising the DAS, a detailed system description, and results from an experimental effort using the separated aperture antenna mine detection technique to measure the performance of the DAS in an outdoor environment.

This paper details the background and status of the unexploded ordance (UXO) problem. Included are a brief history of the problem; a discussion of government initiatives; and a description of UXO detection technologies, their effectiveness, and potential future uses. This paper focuses on domestic UXO contamination primarily resulting from training and testing exercises; nevertheless, the discussion pertains to UXO contamination that has resulted from wartime activities throughout the world. Currently, there is no safe, technically feasible method to accurately detect, locate, and identify surface and subsurface UXO.

Ground Penetrating Radar (GPR) has become widely accepted as a major technique for subsoil investigations over the recent years, mainly in civil engineering. Another field of application, on a global scale, is the pollution of vast areas with land mines, especially in countries of former armed conflicts. According to UN estimates, the number of buried anti-personnel mines exceeds 100 million, with 15,000 people killed every year. The rate of new mines being layed is about one million per year and surpasses the number of mines cleared by a factor of twenty. This demonstrates the need to develop new technologies to increase the efficiency of mine clearing operations. The intension of this paper is to give a short review of the underlying principles and limitations of the GPR-technique. The advantage of 3D versus 2D image processing techniques to enhance data quality and thus detection probability is demostrated, using measured data from sandbox experiments with buried plastic mines. The processed data presented show vertical and horizontal planes through the subsurface and give a clear indication of the buried objects. Factors determining the resolution of the method are discussed. Measurements taken from stones are compared with data obtained from buried mines. The mine data exhibit specific resonances, which is probably due to a minor metal content.

A small minefield was deployed in the desert near Yuma, Arizona in June of 1993. Radar data of this minefield was collected by ground-based and airborne radar sensors. The minefield consists of M-20 metal and M-80 plastic anti-armor mines and Valmara-69 antipersonnel mines. The mines were deployed on the surface and buried at three different depths. Images and analysis of the minefield, which are derived from data collected by the SRI FOLPEN II synthetic aperture radar, are presented here. The minefield was imaged over three bands from 100 to 500 MHz and at various depression angles with this radar sensor. The image analysis is compared to the modeling results of surface and buried mine-like objects. We also show the results of a new radio frequency interference (RFI) rejection algorithm and the image quality improvement we achieved.

The advanced technology demonstration program in 1994 was initiated by the US Army Environmental Center, at Jefferson Proving Ground, Madison, Indiana, to assess the state-of- the-art technologies applicable to detection and remediation of unexploded ordnance (UXO). A multifaceted methodology was developed for evaluating the demonstrators' results in detecting and identifying UXO targets, utilizing both spatial and nonspatial information. The evaluation process involves two major steps. The first step, called target matching, is to compare the reported targets with the known, emplaced targets, in terms of target location, depth, type, class, and other ordnace-specific attributes. The purpose of the first step is to measure a demonstrator's detection and identification capability in general, while the purpose of the second step is to analyze the matching results using statistical methods.

This paper describes imaging of buried objects and fluids. The motivations are to locate pipe leakage and unexploded ordnance. The method is to radiate and receive continuous, discrete frequency radio waves with antennas near the ground, to synthesize sampled area arrays of reflectance data, and to process the data into images with an algorithm based on angular spectrum diffraction theory. Experimental results are presented for three setups. An initial, laboratory setup had a single, spatially scanned antenna; it was used to image buried mud. The second with an array of five antennas on a vehicle, images a buried creosote pit. The third, with a vehicular array of seven antennas, imaged buried metallic objects and depressions in the soil surface.

This report presents findings based on an analysis of the thermal characteristics of live US Army anti-tank mines and concrete slurry-filled M75 surrogates. The US Army's Airborne Standoff Minefield Detection System program relies on the use of surrogate mines to provide their prime contractors with targets to test and develop their systems. Analysis of 8-12 mm sensor image data collected over a period of days at Ft. A.P. Hill, Virginia indicates that the concrete slurry-filled M75 surrogates have diurnal thermal infrared signatures that are very similar to those of live M75 mines, and are therefore good mine surrogates.

The US Army's Communications and Electronics Command through the auspices of its Night Vision and Electronics Sensors Directorate (CECOM-NVESD) is actively applying multisensor techniques to the detection of mine targets. This multisensor research results from the 'detection activity' with its broad range of operational conditions and targets. Multisensor operation justifies significant attention by yielding high target detection and low false alarm statistics. Furthermore, recent advances in sensor and computing technologies make its practical application realistic and affordable. The mine detection field-of-endeavor has since its WWI baptismal investigated the known spectra for applicable mine observation phenomena. Countless sensors, algorithms, processors, networks, and other techniques have been investigated to determine candidacy for mine detection. CECOM-NVESD efforts have addressed a wide range of sensors spanning the spectrum from gravity field perturbations, magentic field disturbances, seismic sounding, electromagnetic fields, earth penetrating radar imagery, and infrared/visible/ultraviolet surface imaging technologies. Supplementary analysis has considered sensor candidate applicability by testing under field conditions (versus laboratory), in determination of fieldability. As these field conditions directly effect the probability of detection and false alarms, sensor employment and design must be considered. Consequently, as a given sensor's performance is influenced directly by the operational conditions, tradeoffs are necessary. At present, mass produced and fielded mine detection techniques are limited to those incorporating a single sensor/processor methodology such as, pulse induction and megnetometry, as found in hand held detectors. The most sensitive fielded systems can detect minute metal components in small mine targets but result in very high false alarm rates reducing velocity in operation environments. Furthermore, the actual speed of advance for the entire mission (convoy, movement to engagement, etc.) is determined by the level of difficulty presented in clearance or avoidance activities required in response to the potential 'targets' marked throughout a detection activity. Therefore the application of fielded hand held systems to convoy operations in clearly impractical. CECOM-NVESD efforts are presently seeking to overcome these operational limitations by substantially increasing speed of detection while reducing the false alarm rate through the application of multisensor techniques. The CECOM-NVESD application of multisensor techniques through integration/fusion methods will be defined in this paper.

The use of hyperspectral visible and infrared sensors is being explored under an ARPA program to provide a means for the detection of buried mines. The purpose of this paper is to summarize the status of the phenomenology of the detection of buried mines using hyperspectral IR detection mechanisms. Both spectral and temperature phenomena related to buried mines will be investigated in the paper. Concepts using the midwave IR (3 to 5 micrometers ), the longwave IR (8 to 12 micrometers ) and the reflection IR (from 1.1 to 2.5 micrometers ) are emphasized in this current effort, although the full IR and visible spectra is considered. Thermally dominated IR is emphasized because of the desire for day/night operation. The program is initially focusing on nonimaging spectrometer measurements of top layers of soil and subsoil, to determine the presence of spectral differences that can be an indicator of mine placement. These spectrometer measurements will be followed by measurements with hyperspectral imaging sensors. While many broad measurements have been made in the MWIR and LWIR, few measurements have been made with an imaging spectrometer. The ARPA/University of Hawaii Spatially Modulated Imaging Fourier Tranform Spectrometer (SMIFTS) can provide such data in the 1.1 to 5.0 micrometers band and the Lawrence Livermore National Laboratory's Livermore Imaging Fourier Transform Infrared Spectrometer (LIFTIRS) will cover the 8-12 micrometers region. The sensors will be deployed in the field from an elevated platform to acquire data in support of both the phenomenology verification and the development of algorithms.

When choosing a route to transport troops and equipment in tactical scenarios, one requires a decision-making scheme that can make fast surveys of the possible paths. One of the main threats in this operation is the presence of scattered surface-laid mines. A possible solution would use an airborne long wave infrared (LWIR) active imaging system. In this paper, we report on one such system based on an intensity modulated waveguide CO₂ laser. This system, which provides images in reflected intensity and relative range, has been tested on replica mines in laboratory. A relative range resolution of 2 mm is reported. Evidence of the insensitivity to the contrast in reflection and the absence of speckle noise for the relative range images is shown. A phenomenon associated with the erroneous evaluation of the relative range of inclined surfaces is identified.

The passive multispectral IR polarization signature attributes of mines and background are observable to an appropriately designed detection system. The processes that create signature polarization are spectrally dependent. At shorter wavelengths, reflected solar radiation produces polarization which is perpendicular to the plane of incidence. At long wavelengths, the reflected sunlight is relatively weak and polarization of thermal emissions, which are parallel to the plane of incedence, may dominate. A multispectral polarimetric imaging system could measure a scene's percent and angle of polarization attributes in different spectral regimes. These images can be spatially compared to reveal the presence of manmade polarizing features such as the exposed surfaces of mines or anomalous perturbations to normal background. This information would be processed by suitable discrimination algorithms which might cross-correlate the spatial polarimetric and spectral channels. Aerodyne Research, Inc. and Boeing Defense and Space Group of Seattle, WA have investigated the feasibility of employing passive IR multispectral polarimetry to locate and identify land mines. The results of this investigation, which used a combination of model- based analysis and field measurements, are reported.

Buoys in the open ocean are clearly observed from a range of 90 km with an airborne IR sensor. A simple model which predicts the observed contrast between small floating objects and the sky radiance reflected from the ocean surface at shallow depression angles has been developed. The detectability of the floating object can be predicted as a function of the percentage of the object that is above the water surface and the size of the object. The model treats the surface object as a black-body at the ambient ocean temperature and predicts the Fresnel reflection of the sky radiance from the ocean surface. Model predictions for the radiance contrast between the ocean and the object have been compared to the observations and are in good agreement. Using a time sequence of the IR images, 3D space-time detection processing shows the potential for further improvement of the detectability of these small objects.

Mine detection is essential to the military forces' mobility and survivability on the battlefield and to the deminers' survivability in humanitarian demining operations. Mine detection technologies are the same but the operational requirements and consequently the technical emphases are different. Twenty-four hour detection capability and rate of detection are essential in military operations while 100 percent probability of detection is essential in demining operations to ensure the inhabitants' safety. My presentation focuses primarily on mine detection technologies for military operations; however, these technologies are very applicable for detecting mines and mine-like targets during humanitarian demining operations.

The Navy's Coastal System Station (CSS) at Panama City, Florida has been investigating the use of multispectral, intensified cameras for standoff minefield detection. In support of CSS, Nichols Research Corporation's Shalimar Florida Office has developed a 'minefield image synethesis tool', (MIST), which is capable of simulating UV to near-IR images of minefields. The MIST software is divided into two major modules, an image generator and an intensified camera model. The image generator (IG) software performs 3D graphics rendering of objects in the scene to produce 2D images as an imaging sensor would see them. The IG models diffuse reflection from sunshine, skyshine, and earthshine. Path transmittances and radiances are accounted for. The sensor spectral band is a user input. Other quantities including reflectances and illumination sources are imput spectrally, making it possible to generate images for different spectral bands, such as those being investigated by CSS. Sensor effects including intensifier/detector response, noise, and analog-to-digital conversion are modeled in the intensified camera model (ICM) software. This paper describes the MIST software and tests that have been performed to validate the software.

A statistical parametric multispectral sensor performance model was developed by ERIM to support mine field detection studies, multispectral sensor design/performance trade-off studies, and target detection algorithm development. The overall model incorporates four components; a mission flight model, a multispectral target and background signature model, a multispectral sensor model, and a multispectral target detection model. Emphasis is placed on estimating the effects of mission multispectral target detection algorithms. Thus, the model ideally supports mission and multispectral sensor trade studies which require optimization of the system's overall target detection performance. The model and a typical example of performance prediction results are presented.

A major obstacle to target detection by airborne electro-optic systems in the shallow water environment is distortion due to wave action from the sea surface air-to-water interface. This problem is being studied by the Program Executive Office Mine Warfare (PMO-210) through an ongoing project called Magic Lantern Adaptation. Under funding from this program, a pier test has been planned to collect data to help understand and model the warping effects of the air-to-water interface. To measure the amount of distortion occurring from wave action, a figure of merit was developed by a fuzzy 'ANDing' of 1) a scale and bias invariant normalized correlation, 2) the ratio of signal power to local noise power, and 3) the ratio of signal strength to overall image background variance. Results show that the figure of merit can successfully evaluate the 'goodness' of target distortion. On a scale from 0 to 100, with 100 indicating no distortion, 'good' targets were measured in the 70s, 'fair' targets in the 50s, and 'poor' targets in the 20s and 30s. The choice of scale and bias invariant correlation was derived after evaluating the problems encountered with standard and scale invariant correlation techniques. This paper contains a discussion of this evaluation.

Land mines usually appear in a wide-field airborne camera as small, low contrast objects. Factors such as the transfer function at relatively high spatial frequencies (as compared with half the pixel frequency) become critically important for detection as well as predicting performance for a notional detection system. An experimental method is described (using a general-purpose filter wheel) for obtaining better resolution and contrast values. Mathcad^TM documents that automate this process somewhat will also be presented.

A multispectral modeling methodology has been developed by Nichols Research Corporation (NRC) in support of the Joint Mine Detection Technology Program at Coastal Systems Station, Panama City, FL. The modeling technique is an extension of NRC's Minefield Image Synthesis Tool. Synthetic imagery generated using this methodology has been compared to measured imagery with excellent results. Multispectral images from an intensified multispectral camera being flown over a minefield have been synthetically generated. This sequence of images was processed by NRC's baseline mine detection algorithm. The results were compared to images generated during a flight test. The conclusion is that the model can predict the automatic target recognition (ATR) algorithm's performance. This paper describes the minefield signature modeling methodology, presents a case study of the comparison of a synthetic scene and a measured scene, and indicates the lessons learned about the phenomenological causes of the mine and background signatures and their effects on a particular class of mine detection algorithms.

This paper describes a novel method for detection of object characteristics using coherent imaging methods. Experimental results are presented. The technique is particularly useful for identifying known or desired features of the object being illuminated and is suited for detection under adverse conditions. A system was demonstrated that generates real-time surface contours with zoom capability using a spatial modulation approach. The system is unique in that it uses video technology to generate patterns from a single variable spacing projection grating. In the demonstrated projection system, the distorted target grating is processed with a spatially matched filter. This is a binary optical filter that passes information selectively. By using a special video processor, this situation may be duplicated electronically. The output of the video processor circuit can be an image showing equal depth contours superimposed on the target or may be designed to produce different electronic signature information such as texture information. In addition, noise reduction techniques have been used to reduce or remove the regular and distorted grating features, while retaining object contours. The adaptation of this method to turbid or scattering media is an object of current research. The approach relies upon coherent signal processing techniques to recover generated patterns broadcast from the illumination device against the background.

The performance of a temporally and spatially nonscanning imaging spectrometer is examined in the context of mine detection. The instrument is described in terms of computed- tomography concepts, specifically the central-slice theorem. The critical system element is a sequence of three transmission sinusoidal-phase gratings rotated in 60 degree increments which achieve dispersion in multiple directions and into multiple orders. The dispersed images of the system's fieldstop are interpreted as 2D projections of a 3D object cube. Due to finite focal-plane array size, this computed-tomography imaging spectrometer (CTIS) is an example of a limited-view-angle tomographic system. The imaging spectrometer's point-spread- function is measured experimentally as a function of wavelength and position in the field-of- view. Reconstruction of the object cube is then achieved via the Maximum-Likelihood Expectation-Maximization algorithm under the assumption of a Poisson likelihood law. The CTIS was tested in five experiments, each experiment utilizing a different target scene displayed on a video monitor. The five target scenes were: a 'University of Arizona' target, two minefield scenes (based on field data) including a simulated mine-like object produced by a HeNe-laser beam-print, and two minefield scenes including a simulated mine-like object produced by a laser-diode beam-print. In all five cases, the CTIS yielded accurate spatial and spectral signatures of the mine-like targets.

This paper reports an analysis of polarimetric hyperspectral images of an outdoor scene containing a number of inactive mines. The images were taken with a visible/near-IR acousto- optic tunable filter polarimetric hyperspectral imaging system. The result has illustrated that the characteristics of mine smooth surfaces can create high signals in polarization difference images and consequently provide a quick and reliable approach for detection.

Multispectral imaging system analyses have been conducted to develop and evaluate system concepts for airborne cameras capable of detecting surface targets of interest in a shoreline environment. We have applied classifier-based multispectral models to analyze individual image chain components from ground through image acquisition to output image. Using a real or synthetically generated multispectral scene as input, the output metrics describe target and background class separations for a particular imaging scenario. The models provide spectral analyses over wavelengths from 0.3 to 15 micrometers . Illumination and atmospheric transmission are determined using LOWTRAN7. Analysis in the thermal regions includes a radiometrically accurate simulator to produce 3D scenes to which illumination conditions are added. This paper presents a description of these models and analytical results. Optimum detection time-of- day and band combinations are presented. Solar evelation, atmospheric transmission, and target albedo, as well as analytic characterization of the detection system are included. Additional effects such as target characteristics, backgrounds, and depth below the surface are also included.

A visible wavelength imaging method of identifying surface-laid mines from an airborne platform is described. A Compact Airborne Spectrographic Imager (CASI) collects multispectral radiometric images of mines and backgrounds which are converted to reflectance images using an incident light sensor. Mines are identified by classifying reflectance spectra in two ways. The first classifies individual pixels using the linear correlation coefficient as a measure of spectral similarity while the second classifies spectra using a variant of linear spectral unmixing in which the majority spectral members within an image are treated as background. In scanning manlift imagery of replica mines, targets were discriminated from a variety of background types, even when partially obscured by vegetation, for widely varying illuminations caused by diurnal and seasonal variations, sky conditions, and sun angles. In preliminary practical tests, the CASI was flown over various agricultural fields in which subpixel-size mine-like targets were laid. Visually undetectable targets were detected with good results. Comparison of classifiers revealed that the correlation method is better for high spatial resolution data. When the targets were subpixel in size, the end member analysis had a higher probability of detection than the correlation method, but had more false alarms.

Earlier landmine imaging systems used two collimated detectors to image objects. These systems had difficulty in distinguishing between surface features and buried features. Using a combination of collimated and uncollimated detectors in a Compton backscatter imaging (CBI) system, allows the identification of surface and buried features. Images created from the collimated detectors contain information about the surface and the buried features, while the uncollimated detectors respond (approximately 80%) to features on the surface. The analysis of surface features are performed first, then these features can be removed and the buried features can be identified. Separation of the surface and buried features permits the use of a globbing algorithm to define regions of interest that can then be quantified [area, Y dimension, X dimension, and center location (x_o, y_o)]. Mine composition analysis is also possible because of the properties of the four detector system. Distinguishing between a pothole and a mine, that was previously very difficult, can now be easily accomplished.

This paper describes the sea mine countermeasures developmental context, technology goals, and progress to date of the two principal Office of Naval Research exploratory development programs addressing sea mine reconnaissance and minehunting technology development. The first of these programs, High Area Rate Reconnaissance, is developing toroidal volume search sonar technology, sidelooking sonar technology, and associated signal processing technologies (motion compensation, beamforming, and computer-aided detection and classification) for reconnaissance and hunting against volume mines and proud bottom mines from 21-inch diameter vehicles operating in deeper waters. The second of these programs, Amphibious Operation Area Mine Reconnaissance/Hunter, is developing a suite of sensor technologies (synthetic aperture sonar, ahead-looking sonar, superconducting magnetic field gradiometer, and electro-optic sensor) and associated signal processing technologies for reconnaissance and hunting against all mine types (including buried mines) in shallow water and very shallow water from 21-inch diameter vehicles. The technologies under development by these two programs must provide excellent capabilities for mine detection, mine classification, and discrimination against false targets.

The use of the magnetic gradient tensor in the point-by-point localization of a magentic dipole was first demonstrated by Wynn in 1971, with a more explicit solution derived by Frahm in 1972. This algorithm maps the five independent components of the magnetic gradient tensor at a point into the dipole bearing vector and the dipole moment vector scaled by the inverse fourth power of the range to the dipole. This inversion produces four solutions, two of which are reflections through the origin of the bearing and scaled moment vectors of the other two. In the present paper, we describe an algorithm for mapping the time rate of change of the gradient tensor measured by a sensor of known velocity into the dipole bearing vector and the dipole moment vecot scaled by the inverse fifth power of the range. An extensive computer exercise with random position and moment vector geometries consistently produces at least one and at most four distinct solutions, with an equal number of additional solutions related to these by reflection of the bearing vector through the origin, for a total of at least two and at most eight solutions. In the same exercise, the solution common to this algorithm and the gradient equation iversion algorithm is consistently unique, and the two different moment vector scalings allow the range to be determined, resulting in a unique solution for dipole position and moment vectors. A general proof of uniqueness is not yet available.

A very high power, state-of-the-art, scanning x-ray source has been developed for use with an x-ray backscatter system that detects and images buried land mines. This paper describes the distinctive qualities of the x-ray source technology necessary to prove the feasibility of the mine detection technique in the field. The imaging system requires that an x-ray beam, having a nominal illumination area on the ground of two centimeters by two centimeters, sweeps across a width of three meters in a time of 15 milliseconds or less. The source must produce an integrated flux of 10⁶ x-rays (min) at 120 kVp (min) for each pixel. The source technology is based on a plasma-focused electrom beam operating up to 140 kilovolts with a current of 0.7 ampere. The electrom beam is magnetically shaped to form a thin ellipse with dimensions of approximately one millimeter by ten millimeters. The scanner is designed to run continuously with target temperature of 160 degrees F (max). The overall design allows the scanner to run with operational and auxilary power generators in the field. A unique 400 hertz, 440 volt, 3-phase, SCR-controlled, low energy storage DC source, with low ripple and 1% voltage regulation, supplies the scanner with 100 kilowatts of power at up to 160 kilovolts. The uniqueness of the mine detection technique and scanner design limits radiation hazards: 1) focusing and tight collimation minimizes stray x-rays; 2) the x-rays travel directly into the ground and are mostly absorbed; 3) radiation leakage from the source is not permitted; and 4) backscatter radiation is strongly localized around the irradiation area, is directed upward, and has a small angular distribution.

Sensors incorporating Superconducting Quantum Interference Devices (SQUIDs) provide the greatest sensitivity for magnetic anomaly detection available with current technology. During the 1980's, the Naval Surface Warfare Center Coastal Systems Station (CSS) developed a superconducting magnetic sensor capable of operation outside of the laboratory environment. This sensor demonstrated rugged, reliable performance even onboard undersea towed platforms. With this sensor, the CSS was able to demonstrate buried mine detection for the US Navy. Subsequently the sensor was incorporated into a multisensor suite onboard an underwater towed vehicle to provide a robust mine hunting capability for the Magnetic and Acoustic Detection of Mines (MADOM) project. This sensor technology utilized niobium superconducting componentry cooled by liquid helium to temperatures on the order of 4 degrees Kelvin (K). In the late 1980's a new class of superconductors was discovered with critical temperatures above the boiling point of liquid nitrogen (77K). This advance has opened up new opportunities, especially for mine reconnaissance and hunting from small unmanned underwater vehicles (UUVs). This paper describes the magnetic sensor detection and classification concept developed for MADOM. In addition, opportunities for UUV operations made possible with high T_c technology and the Navy's current efforts in this area will be addressed.

In the past, fluxgate magnetometers have been limited to stationary applications due to their vector nature. Because of its small relative size and cost, many additional potential uses and applications for the fluxgate magnetometer would be generated if this inherent motion noise could be eliminated. This paper describes a unique hardware solution that when coupled with a set of robust software algorithms, would allow the use of inexpensive fluxgate magnetometers to collect high quality data even when the sensor system is in motion. A novel three sensor gradiometer configuration patented by IBM makes this noise cancellation possible. When coupled with appropriate algorithms, significant detection and target tracking capabilites have been demonstrated. This technology has obvious advantages and applications in areas where other technologies are limited, such as buried objects including those underwater, and targets intentionally hidden or obscured from visual detection. While the sensor system is still in the prototype development stage, the described techniques and technology are expected to ultimately make a small, fieldable, man-portable, tracking gradiometer a reality. Advantages of this technique are its simple implementation and wide applicability to a variety of targets and scenarios. Early test results from the first diver prototype system are presented. Several potential areas of application are reviewed, and the technological advantages and limitations discussed.

Two mine detection techniques rely on the development of a field-equipped x-ray source. The techniques are the Photon Backscatter Imaging method (based mainly on multiple x-ray backscatter) and the first-scatter differential x-ray method. This paper discusses a new x-ray source technology concept and its application in mine detection for use with these techniques. The concept is based on multiple-cathode technology. The basics of the concept are to use small, fast-switching, grid-controlled, multiple dispenser cathodes that are arranged in a closely-spaced, firing sequence. The unique feature of the technology is the ability to control the operation of a sequence of cathodes, so that each cathode is supplied with 1 ampere of current for a time of 100 microseconds. The basics were demonstrated with three cathodes, operated in sequence. The demonstration included dual-energy voltage switching. 'On' and 'off' switching times of 1 microsecond and 3 microseconds, respectively, have been demonstrated at an anode potential of 105 kilovolts and current of 1 ampere. The technology has the capability to dynamically modulate the voltage of the cathodes by plus/minus 20 kilovolts. The source, utilizing this technology, will be able to scan the width of a vehicle for vehicular land mines. The source is expected to be rugged and affixed to the front of a vehicle. The housing diameter for the scanner will be small, i.e. approximately 20 centimeters. The technology favors the use of modular components that are easily replaceable for simple maintenance. These qualities are all positive considerations for a mine detector for the field.

The Marine Corps is pursuing development of sensor technologies to detect and localize mines and other small targets against highly cluttered backgrounds. The objective is to provide real- time day and night detection, with information conveyed immediately to tactical commanders to facilitate mine avoidance or optimum countermeasured deployment. Sensor choice is severely constrained by available platform space and the small size of the targets. Optical sensors have been perceived as most promising to date. Currently an intensified CCD camera using a spinning filter wheel is being demonstrated. The immediate extension is a multichip imager employing tunable filters. Beyond that, active imagers will be investigated for their ability to function at night. The incentive is strong to seek fundamental advances in imaging technology for their prospective ability to extend capability while preserving system simplicity and reliability.

Today's ATR is constructed via inefficient and suboptimal system configuration and training. The process of configuring an ATR is currently very labor intensive, subjective, and inaccurate, as is the process of training an ATR for a particular mission. To cure this deficiency, what is desired is an automated method of configuration and training which is capable of searching the N-dimensional space of modules, algorithms, and parameter values to produce ATR algorithm suites which perform best in each trained scenario. Also, today's ATR is only capable of a limited amount of adaptation to sensed (or otherwise obtained) changes in the environment. To improve the adaptibility of ATR processing and thereby improve accuracy and robustness, what is desired is a high-level control structure which enables system adaptation via changes in parameter values and changes in algorithms (at the component and at the 'suite' level). The Honeywell effort is producing a system for Adaptive Configuration and Control (ACC) of an ATR system which addresses the above described problems. The software system is using the machine learning technique of Genetic Algorithms to autonomously and optimally perform configuration and training and it is using case-based reasoning to provide run-time configuration and control of the ATR system. This paper provides an overview of the ACC system, describes its operation, and describes the benefits it provides to ATR systems.

An automatic target detection algorithm which exploits spectral and spatial signatures of mines is described. Key features of this approach include the ability to adapt to unknown or changing background statistics and the capability to operate with unknown spectral signatures. Preliminary results of applying this algorithm for surface mine detection in video-based multispectral imagery covering the 400-900 nm region are presented. Tests on actual airborne data collected during 1992, 1993, and 1994 show that at 8-inch ground resolution (with 4x over-sampling), 12-inch diameter circular mines can be discriminated from natural backgrounds with a probability of detection around 85% with 3-4 false alarms per image in a relatively harsh clutter environment. This capability has been shown to be sufficient to meet COBRA minefield requirements during preliminary system testing.

A wavelet processing-based automatic target detection technique has been developed at JPL and demonstrated for mine detection applications. In this approach, first, a multiresolution wavelet decomposition method was utilized to remove clutter from the input scene so that the false alarm rate due to background clutter could be greatly reduced. Second, a shape-specific ternary-valued wavelet filter was used to perform mine detection. This ternary-valued wavelet filter was successfully implemented in an optical wavelet processor and real-time mine detection was demonstrated. Theoretical analysis of this wavelet processing method will be provided. Experimental results illustrating mine detection will also be presented.

Coastal Systems Station has developed an approach for automatic mine detection and classification. The Detection Density ACF Approach was created by integrating the adaptive clutter filter (ACF) developed by Martin Marietta, the specification of target signature suggested by Loral Federal Systems, and the Attracted-Based Neural Network developed at NSWC Coastal Systems Station with a detection density target recognition criterion. The Detection Density ACF Approach consists of eight steps: image normalization, ACF, selecting the largest ACF output pixels, convolving the selected pixels with a minesize rectangular window, applying a Bayesian decision rule to detect minelike pixels, grouping the minelike pixels into objects, extracting object features, and classifying objects as either a mine or a nonmine with a neural network. When trained on features extracted from 30 sonar images and tested against another 30 images, this approach demonstrates very good performance: probability of detection and classification (p_dp_c) of 0.84 with a false alarm rate of 1.4 false calls per image. A performance analysis study shows that the detection density ACF approach performs very well and significantly reduces the false alarm rate.

A fractal-based analysis algorithm has been developed to perform the task of automated recognition of minelike targets in side scan sonar images. Because naturally occurring surfaces, such as the sea bottom, are characterized by irregular textures they are well suited to modeling as fractal surfaces. Manmade structures, including mines, are composed of Euclidean shapes, which makes fractal-based analysis highly appropriate for discrimination of mines from a natural background. To that end, a set of fractal features, including fractal dimension, was developed to classify image areas as minelike targets, nonmine areas, or clutter. Four different methods of fractal dimension calculation were compared and the Weierstrass function was used to study the effect of various signal processing procedures on the fractal qualities of an image. The difference in fractal dimension between different images depends not only on the physical features extant in the images but in the underlying statistical characteristics of the processing procedures applied to the images and the underlying mathematical assumptions of the fractal dimension calculation methods. For the image set studied, fractal-based analysis achieved a classification rate similar to human operators, and was very successful in identifying areas of clutter. The analysis technique presented here is applicable to any type of signal that may be configured as an image, making this technique suitable for multisensor systems.

A novel anomaly screening algorithm is described which makes use of a regression diagnostic associated with the fitting of Markov Random Field (MRF) models. This regression diagnostic quantifies the extent to which a given neighborhood of pixels is atypical, relative to local background characteristics. The screening algorithm consists first in the calculation of an MRF-based anomoly statistic values. Next, 'blob' features, such as pixel count and maximal pixel intensity are calculated, and ranked over the image, in order to 'filter' the blobs to some final subset of most likely candidates. Receiver operating characteristics obtained from applying the above described screening algorithm to the detection of minelike targets in high- and low-frequency side-scan sonar imagery are presented together with results obtained from other screening algorithms for comparison, demonstrating performance comparable to trained human operators. In addition, real-time implementation considerations associated with each algorithmic component of the described procedure are identified.

A need exists to develop robust automatic techniques for discriminating between minelike target and clutter returns in sonar imagery. To address this need, an adaptive clutter suppression linear FIR filtering technique has been developed and applied to side scan sonar imagery data. The adaptive filtering procedure consists of four stages. First, a normalized average target signature (shape) within the filter window is computed using training set data. Second, the background clutter covariance matrix is computed by scanning the filter window over the data. Third, following substitutions of the average target signature and covariance expressions into a set of normal equations, an adaptive filter is computed which simultaneously suppresses the background clutter while preserving the peak of the average target signature. Finally, the data is filtered using the 2D adaptive range-crossrange filter. The overall mine detection processing string includes automatic gain control, data decimation, adaptive clutter filtering (ACF), 2D normalization, thresholding, exceedance clustering, limiting the number of exceedances and secondary thresholding processing blocks. The utility of the ACF processing string was demonstrated with three side scan sonar datasets. The ACF algorithm provided average probability of detection and false alarm rate performance similar to that obtained when utilizing an expert sonar operator.

The Coastal Battlefield Reconnaissance and Analysis (COBRA) program is a US Marine Corps Advanced Technology Demonstration (ATD). The objective is to design, develop, and demonstrate an unmanned aerial vehicle (UAV) based passive multispectral video sensor subsystem, to detect and locate obstacles and minefields before and during an amphibious assault, and land combat operations in littoral areas. The COBRA ATD system consisting of an airborne sensor subsystem and a ground station subsystem is described along with the testing program.

Minefields have become an increasing threat in modern warfare due to new developments in electronic and explosive technologies. Therefore the planning and realization of troop movements need a far-reaching, fast running detection of minefields and a determination of their precise extension and location. For this task, an airborne reconnaissance system (AAMIS) will be realized by German industry. The technical concept of AAMIS comprises the multisensorial equipment, the carrier, the data transmission, the ground station, and the data processing. Within the scope of AAMIS, the problem of higher, sophisticated data processing will be addressed by FIM. Based on the long experience in image processing, our task hereby is focusing on the following main topics: (1) detection of individual (buried or unburied) mines as cues to minefields using iconic data processing, model-based data analysis, and sensor fusion; (2) determination of minefields and their outlines by methods of structure analysis and grouping based on the detected individual mines; and (3) extraction of background structures and environmental objects to support the detection of individual mines and the determination of minefields.

Detecting minefields in the presence of clutter is an important challenge for the Navy. Minefields have point patterns that tend to exhibit regularity such as equal-spacing and collinearity that provide potentially valuable discriminants against natural occuring clutter. These tendencies arise because of a variety of compelling factors including strategic doctrine, safety, tactical and economic efficiency, and perhaps most intriguing, the human element. In this paper, we introduce several simple procedures to detect regularity in point proceses including the empty boxes test (EBT) and its extensions, the skeptical likelihood test (SLT), and a Fourier-based method. Several possible methods to specifically detect collinearity are also discussed. The preliminary detection performance of a variety of these minefield detection methods are investigated using simulated data and a point pattern extracted from real sensor data.

An inhomogeneous point process is assumed to govern the observation of a minefield embedded in background clutter. Basic assumptions are that the clutter process is distributed as an inhomogeneous Poisson process with a smooth intensity function and that the minefield process is spatially bounded. This leads to a scheme involving estimation of the underlying intensity of the observed process and detection of intensity discontinuities to locate minefield boundaries. A tensor product cubic spline with small bandwidth is used for original intensity estimation. Subsequently, an oversampled spline-wavelet decomposition is applied to the estimated intensity and the maximum modulus of the wavelet transform is used to detect 'discontinuities' in the minefield boundaries when the original smoothness assumptions about the clutter process are valid. Some simulated results are presented for several background clutter processes at different modulus of continuity.

The performance of optimum minefield detection tests for unpatterned minefields is determined in this paper. A replacement model for an unpatterned (scatterable) minefield is defined. The log-likelihood ratio for minefield detection is then derived from the model, and detection performance (receiver operating characteristics) determined. The model is validated by a comparison with empirical results for data acquired with the Marine Corps Coastal Battlefield Reconnaissance and Analysis (COBRA) passive multispectral electro-optical sensor. Work with data from the US Army Remote Minefield Detection System (REMIDS) program is also discussed. The replacement model is extended to model the effect of clutter on minefield detection performance; clutter has a substantial impact on performance. Several issues for minefield detection are discussed in detail. First, the notion of jointly sufficient statistics for unpatterned minefield detection is defined. This leads to the categorization of previous work in minefield detection into hard decision and soft decision techniques. The soft decision technique is based on the idea of aggregating individual mine likelihoods over a candidate minefield region, as opposed to hard thresholding the individual mine calls and using the density of detected mines to determine the presence or absence of a minefield. The performance analysis is based on an application of the Central Limit Theorem; these results are validated by simulation, and in a particular case, by an exact computation of the relevant false alarm probabilities.

Emperical Bayes classification rules are derived for minefield detection. Laplace approximations for the likelihood function and for the posterior density are used to construct approximate Bayes rules for classifying each unit or region as belonging to one of the two possible types, indicating the presence or absence of mines. Approximate maximum likelihood estimation is proposed assuming that repeated observations are available. An application to a multivariate Poisson log normal model is discussed briefly.

Detecting spatial regularity in images arises in military applications, computer vision, scene analysis, and other areas. In this paper we give a O(n^5/2)-time algorithm that for a given pointset finds all maximal approximately equally-spaced collinear subsets. Our algorithm is robust in that it can tolerate noise or imprecision that may be inherent in the measuring process. Thus, our algorithm is applicable in real-world arenas, such as landmine detection from infrared images.

Statistical problems in pattern or structure recognition for a random multidimensional point set may be addressed by variations on the random graph model of Erdos and Renyui. The imposition of graph structure with a variable edge criterion on a large random point set allows a search for signature quantities or behavior under the given distributional hypothesis. The work is motivated by the question of how to make statistical inferences from sensed mine field data. This article describes recent results obtained in the following special cases. On independent random points U₁,...,U_n distributed uniformly on [0,1]^d, a random graph G_n(x) is constructed in which two distinct such points are joined by an edge if the l_(infinity )-distance between them is at most some prescribed value 0 n, the smallest x such that G_n(x) is connected, and the largest nearest neighbor link d_n, the smallest x such that G_n(x) has no vertices of degree zero, are asymptotic in ratio, as n becomes large, for d >= 2.

Given a set of {(x,y)} coordinates, some corresponding to mine locations and the rest corresponding to the locations of minelike clutter, an algorithm is developed which attempts to recognize linear patterns in the data, to filter out clutter, and declare a region as being a minefield or not a minefield. A linear density is computed for each observation at multiple directions. High densities as well as frequently occurring directions are statistics computed for minefield detection as well as pattern recognition for locating minelines. Significance and power curves are developed by Monte Carlo simulation under the assumption that the observed clutter is distributed uniformly over the area scanned. Some limited results on real minefield data are then presented.

An algorithm for real-time minefield detection from monochromatic airborne imagery must analyze the shape and spatial inter-relationship between compact, several pixel wide, regions that contrast with background. The regions are sparsely distributed over large areas and input data rates are very high. A hierarchical algorithm is described which meets these severe requirements. In progressing from lower to higher levels, computational operations become more complex but the volume of data to be analyzed decreases. At the lowest level of hierarchy, nonsuspect regions are rejected to drastically reduce the data rate. At increasingly higher levels, suspect regions are segmented into homogeneous subregions, morphological features of the subregions are extracted and subregions are classified based on extracted features. At the top level, spatial relationships between 'minelike' regions are determined and are used by statistical and knowledge-based methods to classify the imaged area as being a minefield with an estimated likelihood. An expert system performs syntactic pattern recognition, coordinates the algorithm, and integrates external information. The algorithm is being implemented on a distributed computing system consisting of workstation, vector processors, and transputers. Processor requirements, data rates, estimated probability of detection and false alarm rates and expected system performance are discussed.

In recent years, MCM has moved to the forefront of the Navy's attention. This paper describes the general problems that drive the technology requirements of classical sea mine countermeasure (MCM) sensors for those working outside of this specialized area. Sensor requirements for MCM are compared with those for antisubmarine warfare. This highlights the unique environmental issues and crucial false target problems. The elimination of false targets, not mine detection, is the principal driver of MCM sensor requirements and places special emphasis on the technologies needed for the sequential operations of detection, classification, and identification.

As a precursor to hardware implementation, we have performed mine detection functions in simulation on a polarimetric hyperspectral imaging (PHI) dataset collected by using a technique of acousto-optic tunable filter (AOTF) camera. In principle, PHI data of an image containing objects of interest in a cluttered background provide significant information about the objects, including their relative sizes, shapes, orientation, and other characteristics such as light reflectance and polarization signatures based on their material properties. The present study was, however, limited only to the direct spectral data for the object of interest. A feedforward artifical neural network (ANN) architecture was programmed to recognize predefined spectral 'templates' by using a well-known, hardware implementable inner-product matching scheme. This scheme is particulartly suited to the problem of spectral discrimination where the spectra to be examined or the objects to be discriminated are uncorrelated, as in the present case. In this paper, we describe the ANN architecture and discuss its hardware implementation issues. In addition, we provide the results of our simulation study performed along with suitable preprocessing steps with various window sizes from 1 X 1 to 50 X 50 pixels, leading to an unambiguous detection of the position of mines in test runs without false alarms.

Detection of objects in images in an automated fashion is necessary for many applications, including automated target recognition. In this paper, we present results of boundary detection using Markov random fields. Once the boundaries of regions are detected, object recognition can be conducted to classify the regions within the boundaries. Thus, an approach that gives good boundary detection is very important in many automated target recognition systems. Our algorithm for boundary detection combines Bayesian approach with a histogram specification technique to locate edges of objects that have a closed-loop boundary. The boundary image is modeled by a Markov random field. The method is relatively insensitive to the input parameters required by the user and provides a fairly robust automated detection procedure that produces an image with closed one-pixel-wide boundaries. We apply our method to mine data with very good results.

Center-surround (CS) fields have long been identified in the human visual system as having properties of edge enhancement that facilitate the detection, location, and tracking of small objects. Unfortunately, the automatic digital implementation of such capabilities is not straightforward, since digital computers lack the neural functionality and connectivity inherent in the retina, optic chiasm, geniculate nuclei, and visual cortex. In this paper, we discuss efficient techniques for implementing CS detectors in terms of a class of nonlinear filters. In particular, we emphasize image enhancement schemes that cue the registration of CS detectors over the target, thus rendering our filters implementationally feasible in terms of hit rate and computational cost. Examples are given of the CS detectors that we call the adaptive Double- Gated and Triple-Gated filters (DGF and TGF). Such filters account for spatial nonstationary input. We further show how the DGF and TGF are useful for object detection in multispectral imagery. Analyses emphasize error and computational cost in practical sensing scenarios.

Between August 1993 and December 1994, the Army Environmental Center conducted a congressionally mandated demonstration of systems for the detection, identification, and remediation of unexploded ordance. Two sites were prepared at Jefferson Proving Ground with emplaced inert ordance of known type in recorded locations and orientations to provide ground truth against which demonstrator performance could be evaluated. Uncertainties due to the sensor, as well as surveying errors on the part of the demonstrators, make matching the demonstrator declarations with the emplaced items on a nontrivial exercise. At the same time, an accurate evaluation of system performance requires that this matching be done in a fair and objective fashion. The matching procedure uses a 'critical distance' to determine whether a demonstrator declaration matches an emplaced item and is, therefore, counted as a detection, or does not and is counted as a false alarm. Declarations that are within the critical distance from an emplaced item are candidates for matches, and those outside the critical distance are false alarms. As expected, the number of matches is a function of the choice of critical distance. Therefore, this distance was varied and the probability of detection was determined as a dependent variable in an attempt to separate true detections from random matches of false alarms to undetected baseline items. As a result of this and other tests, we have gained confidence that relative performance rankings are not dependent on an arbitrary choice of cut- off distance and that the evaluation procedure accurately reflects demonstrator performance at the Jefferson Proving Ground demonstration. In general, detection capabilites were lower than 65 percent and most demonstrators reported multiple false alarms per ordnance item detected.

This paper concerns general statistical properties of mine detection systems utilizing high (e.g. acoustic) returns in the presence of reverberation, modeled as a (background) random field. Recently developed extensions of the 1D theory of high level stochastic excursions are used to describe the occurrences of high peaks of a 2D background reverberation field by a (theoretically justified) Poisson model. This model and its further refinements are then used in discussing false alarm, detection, and classification probabilites.

This paper proposes a new system for real-time detection and classification of arbitrarily scattered surface-laid mines. The system consists of six channels which use various neural network structures for feature extraction, detection and classification of targets in six different optical bands ranging from near UV to near IR. A single-layer auto-associative neural network trained by the recursive least square (RLS) learning rule was employed in each channel to perform target feature extraction. The detection/classification based upon the extracted features was accomplished by a three-layer back-propagation neural network with 11-25-10-1 architecture. The outputs of the detector/classifier network in all the channels are fused together in a final decision making system. Simulations were performed on real data for six bands. Forty-eight different images were used in order to account for the variations in size, shape, and contrast of the targets and also the signal-to-clutter ratio. The overall results for the combined system showed a detection rate of approximately 97%, with less than 3% false alarm rate.

This paper presents a new method for detection/classification of surface-laid land mines fom infrared imagery. This is a multistage process of preprocessing, feature extraction, neural network detection/classification, and path finding utilizing the modified high order correlation (MHOC) method. The preprocessing consists of remapping the image distribution such that the conspicuity of targets are enhanced and the background noise/clutter is suppressed as much as possible. The target feature extraction is accomplished by evaluating the principal component (PC) of blocks of data from the image. The benfit to this feature extraction is that the PCs are decorrelated. A recursive least square (RLS) algorithm is implemented in training of the PC extraction network that performs the feature extraction. Once the PCs are found, they are then used to train and test a three-layer back-propagation neural network to detect and classify the targets. The MHOC method is then applied to the resultant image to further reduce the false positives in the image. This method forms a sequence of cross-correlations and determines the consistency of correlations for path finding. The MHOC method can also be realized in a neural network structure. The simulation results, some of which are included, clearly show the detected mine paths with only a small number of false positives.

The problem of landmine detection has been studied for decades. Mine detection systems have typically been developed by first identifying a sensor technology, then testing on particular manmade testbeds, then deploying the sensor on a vehicle or manportable device. Despite much effort, current systems still exhibit gaps between existing and desired capability, e.g., in terms of rate of advance, detection rate, and false alarm rate within demonstration testbeds. In this paper, we propose a new system-level approach to landmine detection. We argue that 'the landmine detection problem' cannot be attacked in a piecewise fashion: system-level solutions must simultaneously consider functional requirements, sensor technologies, models of sensors, the method of sensor application, and the platforms from which sensors are applied. This perspective allows us to shift our focus from the previous emphasis on novel sensor technology, and to go somewhat beyond traditional doctrines governing standoff or manportable detection. We first propose a new theory of geometric sensing and probing in the mine detection context. Specifically, we propose new formulations of 'object identification by probing' which correspond to various sensing modalities. We demonstrate that multiple agents can achieve probe classes that are not serializable for emulation by a single probe agent. With this in mind, our second main contribution lies in proposing a new paradigm for landmine detection, based on (i) close-in observation with simple spectra, and (ii) small, inexpensive, networkable robotic sensing platforms which can act in a cooperative fashion to implement powerful multi-agent probing strategies.

The need is described for a system-level integrated treatment of compression and detection methods and several issues are raised. Compression detection examples are provided as a first step in this direction and to illustrate the concepts.

For an automatic target recognition (ATR) technology contract, sponsored by the US Marine Corps Systems Command, and by Coastal Systems Station, Honeywell designed, mapped to Khoros, and evaluated state-of-the-art algorithms for target discrimination from an airborne platform. Honeywell's baseline approach to improve traditional algorithm robustness is to use a functional maximization approach for representations of algorithm performance as a function of image metrics and algorithm parameters. Revised ATR parameter values are established by a hillclimbing algorithm that revises the ATR algorithm parameter values in the direction of the largest gradient of the function, thus attaining improved performance for a greater variety of scenarios than those for which the system was trained. The baseline ATR algorithms implemented for this program are designed to effectively exploit spectral features to enhance target cueing reliability. An innovative approach for the mapping of three of the individual waveband images from an array of multispectral images into a feature map which obtains high target versus background contrast is discussed. Experimental results are shown for flight test imagery.

The detection of hazardous targets frequently requires a multispectral approach to image acquisition and analysis, which we have implemented in a software system called MATRE (multispectral automated target recognition and enhancement). MATRE provides capabilities of image enhancement, image database management, spectral signature extraction and visualization, statistical analysis of greyscale imagery, as well as 2D and 3D image processing operations. Our system is based upon a client-server architecture that is amenable to distributed implementation. In this paper, we discuss salient issues and requirements for multispectral recognition of hazardous targets, and show that our software fulfills or exceeds such requirements. MATRE's capabilities, as well as statistical and morphological analysis results, are exemplified with emphasis upon computational cost, ease of installation, and maintenance on various Unix platforms. Additionally, MATRE's image processing functions can be coded in vector-parallel form, for ease of implementation of SIMD-parallel processors. Our algorithms are expressed in terms of image algebra, a concise, rigorous notation that unifies linear and nonlinear mathematics in the image domain. An image algebra class library for the C + + language has been incorporated into the our system, which facilitates fast algorithm prototyping without the numerous drawbacks of descrete coding.

Alternative algorithms for detecting and classifying mines and minelike objects must be evaluated against common image sets to assess performance. The Khoros Cantata^{TM environment provides a standard interface that is both powerful and user friendly. It provides the image algorithmist with an object oriented graphical programming interface (GPI}. A Khoros user can import 'toolboxes' of specialized image processing primitives for development of high order algorithms. When Khoros is coupled with a high speed single instruction multiple data (SIMD) algorithms. When Khoros is coupled with a high speed single instruction multiple (SIMD) processor, that operates as a co-processor to a Unix workstation, multiple algorithms and images can be rapidly analyzed at high speeds. The Khoros system and toolboxes with SIMD extensions permit rapid description of the algorithm and allow display and evaluation of the intermediate results. The SIMD toolbox extensions mirror the original serial processor's code results with a SIMD drop in replacement routine which is highly accelerated. This allows an algorithmist to develop identical programs/workspace which run on the host workstation without the use of SIMD coprocessor, but of course with a severe speed performance lost. Since a majority of mine detection componenets are extremely 'CPU intensive', it becomes impractical to process a large number of video frames without SIMD assistance. Development of additional SIMD primitives for customized user toolboxes has been greatly simplified in recent years with the advancement of higher order languages for SIMD processors (e.g.: C + +, Ada). The results is a tool that should greatly enhance the scientific productivity of the mine detection community.

In Part I of this two-part series, we described the adaptive double-gated filter (DGF), a nonlinear CS detector that is sensitive to target shape and statistical properties. In particular, we derived the DGFs parameters from knowledge of target shape, size, and statistics, as well as background statistics. Several variants of the DGF were presented different for target recognition regimes (e.g., smooth targets against rough backgrounds, and vice versa). An additional variant of the DGF, known as the adaptive triple-gated-filter (TGF) was presented which aids in target registration. In this paper, we analyze the errors inherent in the DGF and TGF. Our analyses emphasize the performance of the DGF in limiting conditions, including high noise and poorly defined targets. Additionally, we discuss implementational optimization of the DGFs performance and time complexity. Examples of DGFs application to a database of field images presented, with discussion of results stated in terms of probability of hits and rate of false alarms.

A fully automatic algorithm was developed for single-frame detection of minelike objects in realistic shallow water, beach, and nearby land environments. Detection was accomplished in gray scale images, containing representative targets and backgrounds, which had been collected by a down-looking coherent active sensor. The problem was made challenging by low contrast, partly covered targets, and highly cluttered images including beach vegetation and rocks, complicated natural backgrounds, obscuration (replacement noise) by glint from the surface of the water and distortion within it, and 15 kinds of manmade objects. To deal with these challenges, innovations have been made in automatic background cancellation, in the final thresholding to binary, and in shape and veracity clues for the feature vector used in the final classification step. Performance is reported for a representative set of 1024 frames. For the majority of background types, including low conventional signal-to-noise ratio and pervasive instances of clutter and replacement noise patches, the algorithm performed correctly in 92% of the frames. This applies individually to frames identified as 'target' where one or more targets existed, and frames identified as 'notarget' where no targets existed. Target-field detection over multiple frames depends upon reliable single-frame target detection. Despite challenging images, performance of our single-frame algorithm appears sufficient for multiframe target-field detection to proceed with acceptable error rates for the majority of background types encountered in the tests conducted.

A statistical parametric multispectral sensor performance model was developed by ERIM to support mine field detection studies, multispectral sensor design/performance trade-off studies, and target detection algorithm development. The model assumes target detection algorithms and their performance models which are based on data assumed to obey multivariate Gaussian probability distribution functions (PDFs). The applicability of these algorithms and performance models can be generalized to data having non-Gaussian PDFs through the use of transforms which convert non-Gaussian data to Gaussian (or near-Gaussian) data. An example of one such transform is the Box-Cox power law transform. In practice, such a transform can be applied to non-Gaussian data prior to the introduction of a detection algorithm that is formally based on the assumption of multivariate Gaussian data. This paper presents an extension of these techniques to the case where the joint multivariate probability density function of the non-Gaussian input data is known, and where the joint estimate of the multivariate Gaussian statistics, under the Box-Cox transform, is desired. The jointly estimated multivariate Gaussian statistics can then be used to predict the performance of a target detection algorithm which has an associated Gaussian performance model.

Coastal System Station, Panama City, Florida, is developing an airborne multispectral visible mine detection system for detecting minefields in littoral regions. Nichols Research Corporation (NRC) in support of CSS has been tasked to develop mine detection algorithms. NRC developed a generic baseline mine detection algorithm in order to provide a stable reference by which to measure other algorithms and enhancements. This algorithm has also served as a baseline to be modified and enhanced as the nature of the data is better understood and discrimination methodology is refined. The baseline algorithm is described in this paper along with modifications which significantly improve its robustness and with the analyses which motivated those modifications.

Modifying of the initial prerequisites of the classical detection theory allows to synthesize the generalized signal detection algorithm. The optimal signal detection algorithms with a priori known and unknown amplitude-phase structure are particular cases of the generalized algorithm.

Automatic target detection is the primary goal of many imaging systems both in defense and manufacturing industries. Advances in methods and equipment for image acquisition, processing, and analysis are required to effectively deal with this problem. Towards this goal, we discuss here a target detection algorithm based on mathematical morphology. Mathematical morphology is an image processing tool that is used for designing nonlinear operators for image representation, processing, and analysis. In particular, the proposed approach is based on a morphological reconstruction algorithm for detecting targets of interest appearing on a scene. We apply this algorithm to the problem of detecting minelike targets in multispectral images, provided to us by the Coastal Systems Station, Naval Surface Warfare Center, Panama City, Florida. The proposed technique is relatively simple and only requires approximate knowledge of target size. The algorithm also effectively incorporates fusion of data from different bands. The implementation has been done in the KHOROS signal and image processing environment with encouraging results.

Our system fuses information contained in registered images from multiple sensors to reduce the effects of clutter and improve the ability to detect surface and buried land mines. The sensor suite currently consists of a camera that acquires images in six bands (400nm, 500nm, 600nm, 700nm, 800nm, and 900nm). Past research has shown that it is extremely difficult to distinguish land mines from background clutter in images obtained from a single sensor. It is hypothesized, however, that information fused from a suite of various sensors is likely to provide better detection reliability, because the suite of sensors detects a variety of physical properties that are more separable in feature space. The materials surrounding the mines can include natural materials (soil, rocks, foliage, water, etc.) and some artifacts. We use a supervised learning pattern recognition approach to detecting the metal and plastic land mines. The overall process consists of four main parts: preprocessing, feature extraction, feature selection, and classification. These parts are used in a two step process to classify a subimage. The first step, referred to as feature analysis, determines the features of sub-images which result in the greatest separability between the two classes, 'mine' and 'background'. The second step, image labeling, uses the selected features and the decisions from a pattern classifier to label the regions in the image which are likely to correspond to mines. We extract features from the images, and use feature selection algorithms to select only the most important features according to their contribution to correct detections. This allows us to save computational complexity and determine which of the spectral bands add value to the detection system. The most important features from the various sensors are fused using a supervised learning pattern classifier (the probabilistic neural network). We present results of experiments to detect land mines from real data collected from an airborne platform, and evaluate the usefulness of fusing feature information from multiple spectral bands. We show that even with preliminary data and limited testing, the performance (specified in terms of probability of detection and probability of false alarm) is very promising. The novelty of the work lies mostly in the combination of the algorithms and their application to the very important and currently unsolved operational problem of detecting minefields from an airborne standoff platform.

The US Army Night Vision and Electronic Sensors Directorate's Countermine Division has received and evaluated automatic target recognition (ATR) algorithms for a number of years. As part of the Army's Advanced Standoff Mine Detection System (ASTAMIDS) program, Camber Corporation has developed an algorithm simulation and evaluation testbed (ASET) that is capable of testing various ATR algorithms on a variety of images. The ASET tool has been designed to support any image type and contains an extensive database of images from ASTAMIDS and previous mine detection programs including REMIDS and AMIDARS, as well as data from SAR and other sources that may be queried and used to exercise a given algorithm. Algorithms can be easily converted into the ASET environment and executed on selected imagery. ASET accurately simulates the execution of algorithms, tests and grades their performance, and allows manipulation and enhancements of the algorithms during execution. This software runs on a variety of computer platforms and allows ATR algorithms to be evaluated under a variety of circumstances.

By decomposing signals into building blocks that are well localized in space and frequency, the wavelet transform has been shown to be well adapted to detect and characterize singularities. The basis functions employ time compression (or dilation) rather than a variation of frequency of the modulated sinusoid. The wavelets are well founded on rigorous mathematical theory, and the expansions are robust. We have applied a set of orthogonal wavelets to detect magnetic underwater signals.

Infrared imagery scenes change continuously with environmental conditions. Strategic targets embedded in them are often difficult to be identified with the naked eye. An IR sensor-based mine detector must include Automatic Target Recognition (ATR) to detect and extract land mines from IR scenes. In the course of the ATR development process, mine signature data were collected using a commercial 8-12 (mu) spectral range FLIR, model Inframetrics 445L, and a commercial 3-5 (mu) starting focal planar array FLIR, model Infracam. These sensors were customized to the required field-of-view for short range operation. These baseline data were then input into a specialized parallel processor on which the mine detection algorithm is developed and trained. The ATR is feature-based and consists of several subprocesses to progress from raw input IR imagery to a neural network classifier for final nomination of the targets. Initially, image enhancement is used to remove noise and sensor artifact. Three preprocessing techniques, namely model-based segmentation, multi-element prescreener, and geon detector are then applied to extract specific features of the targets and to reject all objects that do not resemble mines. Finally, to further reduce the false alarm rate, the extracted features are presented to the neural network classifier. Depending on the operational circumstances, one of three neural network techniques will be adopted; back propagation, supervised real-time learning, or unsupervised real-time learning. The Close Range IR Mine Detection System is an Army program currently being experimentally developed to be demonstrated in the Army's Advanced Technology Demonstration in FY95. The ATR resulting from this program will be integrated in the 21st Century Land Warrior program in which the mine avoidance capability is its primary interest.

Plastic mines are cheap, small, and difficult to detect using current methods. IR polarization discrimination where the mines present themselves as a flashing signal may be of potential benefit in finding surface laid or scattered types in a cluttered background. Results from a laboratory study show that plastic has significant 8-14 micrometers IR polarization in emission and reflection. We have constructed a 10.6 micrometers ellipsometer which has been used to measure the complex refractive indices of mine-like plastics. This apparatus was then modified to determine the degree of emission polarization. The measured degree of emission polarization has been compared to that predicted using the complex refractive index and good agreement found.

We cast the training of minimal artificial neural network architectures as a problem of global optimization, and study the simulated annealing (SA) global optimization heuristic under a 'best-so-far' model. Our testbed consists of separated-aperature radar data for subsoil mine detection. In previous analyses, we have found that the traditional SA 'cooling' paradigm can be suboptimal for small instances of combinatorial global optimizations. Here, we demonstrate that traditional cooling is also suboptimal for training minimal neural networks for mine detection. Related issues include (i) how to find minimal network architectures; (ii) considering tradeoffs between minimality and trainability; (iii) the question of whether multistart/parallel implementations of SA can be superior to a single long SA run; and (iv) adaptive annealing strategies based on the best-so-far objective.

A numerical time-domain technique known as the transmission line matrix (TLM) method is used to analyze a ground penetrating radar (GPR) concept known historically as balanced bridge. This GPR concept is a dielectric anomaly (mine) detection sensor which operates in the UHF frequency band. This mine sensor consists of two receive broadband antennas separated by a single center transmit antenna. Traditionally, if care is taken in the construction of the antennas, the direct coupling and ground reflection energies are combined and nulled out by a hardware coupler when the sensor configuration is over homogeneous soil. When one of the two receiving antennas is over a dielectric anomaly (mine), the differenced energies from the two receiving antennas no longer produces a null and a peaked response is observed. This mine sensing technique has performed well under experimental tests at Fort Belvoir and Fort A.P. Hill, Virginia. Testing results, at different sites using different mine types, have indicated the sensor's performance in terms of probability of detection and false-alarm rates. The TLM method is used to describe the balanced bridge mine detector's response to targets and clutter as well as its unique capabilities in an attempt to shed light into occurring fundamental wave interactions.

Mine and minelike target detection using optical correlators is presented. Minelike targets in experimentally collected data are used. The targets are situated in different environments. Optical correlators employing a variety of spatial filters are examined and their target recognition performance evaluated. Very good results were obtained in detecting the targets in most of the backgrounds used.

A numerical time-domain technique known as the transmission line matrix (TLM) method was used to analyze a ground penetrating radar (GPR) concept historically known as the separated aperture technique. This GPR concept is basically a dielectric anomaly (mine) detection sensor which operates near the L band frequency. This mine sensor consists of transmit and receive broadband dipole antenna. Each antenna is contained within a metallic cavity and the cavities are connected by a metallic septum. Normally, when the mine sensor is scanned over homogeneous earth, very little transmitted power is received by the receiving antenna. The power received by the receiving antenna however, is significantly increased when the detector is scanned over a buried dielectric anomaly (mine). This technique has performed in terms of probability of detection and false-alarm rates at different sites with different mine types. The TLM method was used to analyze the separated aperture mine detector's response to targets, clutter, and to provide insight into the fundamental wave interactions.

Presented herein is a method for the interpretation of electromagnetic (EM) response in sea- water. The method is based upon similarities between the EM wave equation in lossless media, and the EM diffusion equation in conductive media. The technique allows for the transformation between the solution in the two propagation modes. The advantages of the technique are its simple implementation and its generality to a wide variety of cases. Application to mine hunting including conducting and nonconducting mines are simulated. Advantages and limitations of the method are discussed. The technique is also applicable to interpretation of time-dependent heat flow.

A statistical parametric multispectral sensor performance model was developed by ERIM to support mine field detection studies, multispectral sensor design/performance trade-off studies, and target detection algorithm development. A key element in this performance model is the influence of the background on the target's multispectral statistics due to the size and shape of the target under the sensor's point spread function and pixel sampling function. The multispectral statistics of interest include the first-order (mean) and second-order moments (covariance) of the target's radiance signature. This paper presents a formulation which not only considers the effects of a multispectral sensor with a single point spread function, but also considers the joint effects of multiple, potentially misregistered, point spread functions on the target's covariance statistics. The model and an example of sensor point spread function and pixel sampling function effects on the target's spectral statistics are presented.

This paper describes the Ordnance Detection Expert Support Application (ODESA). ODESA is an intelligent unexploded ordance (UXO) detection application that fuses data from a variety of single sensor detection systems to detect buried objects. The ODESA application implements two intelligent data fusion techniques, trained using reports of targets from detection systems and ground truth describing the exact location and type of emplaced UXO. One fusion method uses a genetic algortihm to generate rules and weights that predict the location and identity of buried objects. The other fusion method uses heuristics and conditional probabilities to predict the location of buried objects. Initial findings prove that both methods produce target reports that more accurately detect buried objects than any single detection system.

By imbedding or placing antennas in or on the surface of plastic mines, or the containers of UXO or hazardous wastes, such objects can be detected relatively easy by GPR. The detection of metal mines is greatly enhanced by antennas presented. Experimental results are presented.

Image-based detection of submerged objects is frequently confounded by optical distortions in the aqueous medium. For example, scattering can severly degrade contrast and resolution in underwater (UW) images when illumination systems and cameras are not range-gated. Prior to the development of range-gated imaging, much research emphasis was placed upon the analysis of greyscale imagery acquired under incoherent illumination. Primarily as a result of current emphasis on coherent optical technologies, the progress of image processing (IP) research that pertains to UW imagery has lagged IP hardware and software development. In this paper, we summarize methods for the digital clarification of images that portray actively illuminated UW scenes, i.e., images of floodlit objects. We model the primary UW image components as: a) contrast degradation resulting from illuminant backscattering from the water column, b) a return signal that results from backscattering of the illuminant from the object of regard, and c) resolution loss, due to forward scattering of the return signal. Letting items a) and c) consititute error sources, one can locally apply the appropriate filters to reduce the contribution of such errors. Our technique emphasized local enhancement, as opposed to the global methods used in previous imaging practice. Our enhancement filters are based upon image-algebraic templates that are designed to compensate for the effects of single and multiple scattering as well as absorption within the water column. Discussion is based upon image clarity, algorithmic complexity, and computational efficiency.

A variety of experimental results indicate that Dolphins possess a unique and sophisticated sonar system. In addition, this sonar system is highly adaptive in detecting, discriminating and recognizing sonar targets in highly reverberating and noisy environments. In this paper a new approach using Resonance Scattering Theory in target detection and recognition is presented. The results seems to imply that this approach may be useful in shallow water target detection and identification.

The detection of land mines has two fundamental goals: the first is a high detection rate (low probability of missing a mine) and the second is a low false alarm rate. Detection of mines and mine-like objects is generally not difficult; the problem is the high false-alarm rate caused by detection of innocuous objects such as shrapnel or metal junk, or even rocks or voids in the soil. The problem is one of discrimination, not one of detection. In order to maximize the success of achieving this goal, a mine detector needs to incorporate many complementary sensor technologies and to utilize the concept of sensor data fusion. Two subsystems employ new signal processing techniques which extract certain features from the data that are unique identifiers on the mines. These features are the natural magentic and electromagnetic resonances, which form the impulse response function, or equivalently, the natural frequencies represented by poles in the complex frequency plane. For different objects these are sufficiently distinct that pattern recognition processes can be used to arrive at a probability of a match to a particular mine.

An automatic, multistage algorithm for detecting minefields is introduced. This algorithm was tested, with encouraging results, against buried metal mines. The measurement data for this test were obtained using a low-frequency airborne SAR, collected during an extensive ground penetration experiment in Yuma, Arizona, in June 1993. Although verified using SAR data, the automatic minefield detection technique may prove applicable to other remote sensors as well.

Texture detection using a multispectral approach is naturally superior to a unispectral one because the multispectral process takes more information into account. Details not obvious in one image may be more prominent in others, hence improving the chances of recognition and detection. In this paper we present a new method for preprocessing and eventually fusing a set of multispectral images. Images are preprocessed using histogram equalization, which is found to be ideally suited for this exercise. A wavelet transform technique is used to fuse data from the different multispectral images.

The Electro-Optic Identification (EOID) Sensors project is developing a Laser Visual Iidentification Sensor (LVIS) for identification of proud, partially buried, and moored mines in shallow water/very shallow water. LVIS will be deployed in small diameter underwater vehicles, including unmanned underwater vehicles (UUVs). Since the mission is mine identification, LVIS must: a) deliver high quality images in turbid coastal waters, while b) being compatible with the size and power constraints imposed by the intended deployment platforms. This project is sponsored by the Office of Naval Research, as a part of the AOA Mine Reconnaissance/Hunter program. High quality images which retain target detail and contrast are required for mine identification. LVIS will be designed to produce images of minelike contacts (MLC) of sufficient quality to allow identification while operating in turbid coastal waters from a small diameter UUV. Technology goals for the first generation LVIS are a) identification range up to 40 feet for proud, partially buried, and moored MLCs under coastal water conditions; b) day/night operation from a UUV operating at speeds up to 4 knots; c) power consumption less than 500 watts, with 275 watts being typical; and d) packaged within a 32-inch long portion of a 21-inch diameter vehicle section.