Historically, unattended ground sensors have played a limited role in Army operations. Applications included perimeter surveillance and monitoring troop movement along established routes. Advancements in communications, sensors, and computer technology are making larger scale applications feasible. Several Army science and technology programs are maturing technologies that will enable sensor networks to be deployed on the battlefield. Such networks will fill in information voids thus providing unprecedented situational understanding of the battlefield.

Imaging sensor systems address a broad range of needs, in both the military and commercial sectors, and must meet a demanding set of requirements. The unattended imaging sensor application places particularly stressing demands on the technology. Unattended imaging sensors collect data from a broad field of regard, process the information, and transmit it to the observer or to other sensors for integration. The unattended sensor usually operates in a semi-autonomous mode, with a man-in-the-loop only as a check. The sensor must operate over an extended time period, with minimum power, and interpret a large amount of data. Changing environmental conditions impose a further set of requirements. High resolution, high contrast images during one period of the day may degrade significantly as conditions change in the evening, during fog and rain, and as the temperature fluctuates. This paper briefly discusses applications of unattended sensors and highlights recent sensor advances, in several spectral regions, contributing to the unattended sensor application.

This paper will describe an algorithm for detecting and classifying seismic and acoustic signals for unattended ground sensors. The algorithm must be computationally efficient and continuously process a data stream in order to establish whether or not a desired signal has changed state (turned-on or off). The paper will focus on describing a Fourier-based technique that compares the running power spectral density estimate of the data to a predetermined signature in order to determine if the desired signal has changed state. How to establish the signature and the detection thresholds will be discussed as well as the theoretical statistics of the algorithm for the Gaussian noise case with results from simulated data. Actual seismic data results will also be discussed along with techniques used to reduce false alarms due to the inherent nonstationary noise environments found with actual data.

A technology demonstrator that detects and classifies different helicopter types automatically, was developed at TNO-FEL. The demonstrator is based on a PC, which receives its acoustic input from an all-weather microphone. The demonstrator uses commercial off-the-shelf hardware to digitize the acoustic signal. The user-interface and the signal processing software are written in MatLab^TM. The demonstrator detects the noise from helicopters; the classification is performed using a database with helicopter-specific features. The demonstrator currently contains information of 11 different helicopter types, but can easily be expanded to include additional types of helicopters. The input signal is analyzed in real time, the result is a classification ranging from `no target' to `helicopter type x', e.g. Lynx Mk2. If the helicopter is classified, its relative speed is estimated as well. The algorithm was developed and tested using a database of different helicopters (hovering and moving) recorded at distances ranging from 90 meter up to 8 kilometer. The sensitivity to noise was investigated using jet, tank, artillery and environmental (wind and turbulence) noise as input.

Ground-based and airborne acoustic systems often target vehicles that are powered by reciprocating internal combustion engines. Typically the far-field acoustic spectra of these vehicles are dominated by a few narrow spectral lines that are harmonically related. The dominant harmonics change with engine speed and also with emission angle. This paper describes a simple model that recreates some of this variability. The model breaks the far-field signature into two components: the generation of a train of pressure pulses at each exhaust outlet, and the radiation of sound pressures from the outlet(s) to the far field. Predictions are compared with field test data for two ground vehicles.

We present performance analysis for source localization when wideband aeroacoustic signals are measured at multiple distributed sensor arrays. The acoustic wavefronts are modeled with perfect spatial coherence over individual arrays and with frequency-selective coherence between distinct arrays, thus allowing for random fluctuations due to the propagation medium when the arrays are widely separated. The signals received by the sensors are modeled as wideband Gaussian random processes, and we study the Cramer-Rao bound on source localization accuracy for varying levels of signal coherence between the arrays and for processing schemes with different levels of complexity. When the wavefronts at distributed arrays exhibit partial coherence, we show that the source localization accuracy is significantly improved if the coherence is exploited in the source localization processing. Further, we show that a distributed processing scheme involving bearing estimation at the individual arrays and time-delay estimation between pairs of sensors performs nearly as well as the optimum scheme that jointly processes the data from all sensors. We discuss tradeoffs between source localization accuracy and the bandwidth required to communicate data from the individual arrays to a central fusion center, and results from measured aeroacoustic data are included to illustrate partial spatial coherence at distributed arrays.

An important application of seismic and acoustic unattended ground sensors (UGS) is the estimation of the 3D position of an emitting target. Seismic and acoustic data derived from UGS systems provide the raw information to determine these locations, but can be processed and analyzed in a number of ways using varying amounts of auxiliary information. Processing methods to improve arrival time picking for continuous wave sources and methods for determining and defining the seismic velocity model are the primary variables affecting the localization accuracy. Results using field data collected from an underground facility have shown that using an iterative time picking technique significantly improves the accuracy of the resulting derived target location. Other processing techniques show little advantage over simple crosscorrelation alone in terms of accuracy, but may improve the ease with which time picks can be made. An average velocity model found through passive listening or a velocity model determined from a calibration source near the target source both result in similar location accuracies. Surprisingly, the use of average station corrections severely increases the location error.

Under funding from the Air Force Research Laboratory at Hanscom Air Force Base, Quantum Magnetics has been developing a room temperature, multi-axis magnetic gradiometer for the detection and characterization of underground structures. The gradiometer uses small, inexpensive, but highly sensitive magnetoresistive sensors operating with a flat frequency response from DC to several MHz and employs an innovative sensor configuration; called the three-sensor gradiometer (TSG) invented at IBM. The TSG affords unprecedented dynamic range that enables detection of signals near the sensor noise floor even when the system is in motion in the earth's field. The wideband response and high sensitivity of these sensors make them ideal for both passive and active detection techniques. A single sensor can detect perturbations in the earth's magnetic field from ferrous materials used in the construction of the structure, emissions at power and mechanical frequencies from equipment within the structure, and eddy currents in metallic materials within the structure induced by externally applied probe signals. These complimentary data sets can be combined in a sensor fusion scheme to minimize sensor clutter and discriminate against false targets. The sensing technology under development supports deployment by way of unattended ground sensors as well low-flying unmanned aerial vehicles.

Missiles represent strategic weapons in modern warfare. They are stored prior to deployment, and they need to be kept in flight-ready condition during storage. It is therefore necessary to promptly detect any missile fuel leakage. Intelligent Optical Systems initiated an effort to develop a breakthrough device that combines a long period grating- based chemical sensor with neural network software. These reversible sensors perform real-time measurements of liquid fuel components.

Porous silicon chips have been used to detect vapors of explosives and a simulant for the nerve agents Sarin, Soman, and GF using two different transduction modes: reflectivity and photoluminescence. Detection of nitroaromatic compounds is achieved by monitoring the photoluminescence of a nanocrystalline porous Si film on exposure to the analyte of interest in a flowing air stream. Photoluminescence is quenched on exposure to the nitroaromatic. Detection limits of 2 ppb and 1 ppb were observed for 2,4-dinitrotoluene, and 2,4,6-trinitrotoluene, respectively (exposure times of 5 min for each, in air). Specificity for detection is achieved in a two-channel system using catalytic oxidation of the nitroaromatic.

Indium gallium arsenide (InGaAs) focal plane arrays and cameras have demonstrated significant potential in battlefield applications. Room temperature detectivities, D*, in excess of 10¹⁴ cm-(root)Hz/W have enabled night vision imaging under low light level conditions. The 0.9 micrometers to 1.7 micrometers wavelength band allows the use of eye- safe lasers for target designation and covert active illumination. We report here a miniature InGaAs camera designed for unattended ground sensor and robot-mounted applications. The camera is approximately the size of a D- cell battery, weighs less than 200 g. has a 320 X 240 pixel spatial resolution and maintains D* > 10¹⁴ cm- (root)Hz/W. The miniature camera is fully self contained. The only input is DC power (3.6 V). The camera has both analog (RS170) and 12-bit digital (LVDS) video outputs. It is intended as a demonstration vehicle for battlefield distributed robotic vision but will find use in other applications as an unattended sensor or rifle site.

Low power and low cost are primary requirements for an imaging infrared camera used in unattended ground sensor arrays. In this paper, a 120 X 160 amorphous silicon (a- Si) microbolometer-based uncooled infrared camera technology offering a low cost, low power solution to infrared surveillance for UGS applications is presented. A 120 X 160 micron infrared camera (MIRC) has been demonstrated which exhibits an f/1 noise equivalent temperature difference sensitivity approximately 63 mK. This sensitivity has been achieved without the use of a thermoelectric cooler for array temperature stabilization thereby significantly reducing the power requirements. Chopperless camera operation at a 20 Hz frame rate with power consumption of 380 mW has also been demonstrated. The 120 X 160 MIRC operates under digital signal processor (DSP) control. To reduce cost, this DSP-controlled architecture employs commercial off-the-shelf DSP, A/D, memory and voltage regulator chips. The detector chip, employing an integrating amplifier per unit cell ROIC design, is the single custom chip used. The camera also employs low cost f/1 optics, as well as low cost wafer-level vacuum packaging. In this paper, a-Si microbolometer technology for the MIRC will be presented. Also, the key features and performance parameters of the MIRC are presented.

The objectives of the Low Cost Microsensors Program are twofold. The first is to develop and deliver a long-range infrared (IR) sensor built upon an uncooled vanadium oxide (VOx) 640 X 480 format focal plane array (FPA) engine. The second is to develop an expendable microsensor built upon a VOx 160 X 120 format FPA engine. The 640 X 480 sensor is applicable to long-range surveillance and targeting missions and is a reusable asset. The 160 X 120 sensor is designed for applications where miniaturization is required as well as low cost and low power. The 160 X 120 is also intended for expendable military applications. The intent of this DUS&T effort is to further reduce the cost, weight, and power of uncooled IR sensors, and to increase the capability of these sensors, thereby expanding their applicability to military and commercial markets never before addressed by thermal imaging.

The U.S. Army Engineer Research and Development Center Cold Regions Research and Engineering Laboratory is currently developing a human Intruder Thermal Model (ITM) for predicting the average surface temperature of an intruder. ITM provides steady-state predictions of average surface temperature. It accounts for metabolically generated heat and heat exchange with the environment via conduction, convection, perspiration and respiration. It also accounts for long and short-wave radiation exchanges with the environment, the short-wave component being extremely important to daytime surface temperature predictions. Clothing thermal properties and intruder height and weight are factored into model calculations as well.

A sensor array is tomographic if sensor modules share raw information to form a joint target model. Target identification is then implemented on the global model. This paper considers sensor head and sensor array resource budgets for tomographic ground sensors. In this application, mixed local and global analysis schemes are likely to prove optimal. We illustrate potential approaches with images and analysis of 3D visual and infrared targets.

The current trend to develop low cost, miniature unattended ground sensor (UGS) will enable a cost-effective, covert means for surveillance in both urban and remote border areas. Whereas the functionality (e.g., sensing range and life in the field) of these individual UGS (i.e., acoustic, seismic, magnetic, chemical or biological) are limited due to size and cost constraints, a network of these sensors working cooperatively together can provide an effective surveillance capability. A key factor is the ability of these sensors to work cooperatively to achieve a `collective' functionality that can meet the surveillance objective. For example, a realistic mission objective would be to use the minimum number of sensors necessary (i.e., preserve the life of the network) to detect, identify and track vehicles in a desert canyon area that has variable wind and temperature conditions. The network would have to assess the effect of the wind direction and temperature on the sensing range of its acoustic sensors, turn on those sensors that can initially detect the target and dynamically activate other appropriate sensors (e.g., seismic, acoustic or imaging sensors) that can identify and track the vehicle as it moves into and across the canyon area covered by the sensor network. To achieve this type of functionality requires system algorithms that are capable of optimizing the utilization of the sensors. This paper describes results that show improved target tracking accuracy by optimizing the selection of acoustic sensors that measure bearing angles to the target. Also, recent results are described from testing the tracking algorithm with real data.

Inmarsat D+ service is a short measuring service that offers advantages allowing for a small transceiver design. SkyWave Mobile, Inc. was contracted to miniaturize their first generation D+ transceiver, the DMR-200, to produce a 20 cubic inch device. This device, designated the DMR-400, incorporates an Inmarsat D+ transceiver, GPS receiver, antennas, I/O lines, and advanced power management into a 14 cubic inch unit. The DMR-400 has received Inmarsat land- mobile certification and will soon enter an aeronautical approval phase packaged in a radome housing. The first step in the aeronautical certification phase was the design of a radome and mounting kit for external mounting on an aircraft fuselage. This housing also incorporates a heating element to extend the operational temperature range to -55 degree(s)C. The assembled unit will soon begin DO-160 testing to be followed by Doppler correction testing, FAA flight-testing, and Inmarsat Aeronautical testing. The key to the entire process is the successful implementation of Doppler compensation based on the GPS data. The expectation is an FAA and Inmarsat aeronautical certified unit.

In Carnegie Mellon University's CyberScout project, we are developing mobile and stationary sentries capable of autonomous reconnaissance and surveillance. In this paper, we describe recent advances in the areas of efficient perception algorithms (detection, classification, and correspondence) and mission planning. In detection, we have achieved improved rejection of camera jitter and environmental variations (e.g., lighting, moving foliage) through multi-modal filtering, and we have implemented panoramic backgrounding through pseudo-real-time mosaicing. In classification, we present methods for discriminating between individual, groups of individuals, and vehicles, and between individuals with and without backpacks. In correspondence, we describe an accurate multi-hypothesis approach based on both motion and appearance. Finally, in mission planning, we describe mapbuilding using multiple sensory cues and a computationally efficient decentralized planner for multiple platforms.

A system has been developed for delivering and attaching a sensor payload to a target using a standard 40-mm grenade launcher. The projectile flight characteristics are similar to existing 40-mm rounds, with a useful range of up to 300 m. The projectile incorporates an attachment mechanism, a shock mitigation system, a power source, and a transmitter that allows sensor data to be transmitted to a receiver at up to 1/4 mile range. Impact g-loads have been limited to less than 10,000 g's, enabling sensor payloads to be assembled using Commercial Off-The-Shelf (COTS) components. The GLIMPS projectile is intended to be a general purpose delivery system for a variety of sensor payloads under the Unattended Ground Sensors program. The Phase I proof-of- concept design was demonstrated using a low-power CMOS camera. Demonstration results and Phase II progress are presented.

Raytheon Electronics Systems, under contract from the DARPA Advanced Technology Office, has designed, fabricated and delivered the Modular Miniature Integrated Sight (M²IS). M²IS is a rifle-mounted system that integrates a high- performance multispectral sensor with an eyesafe laser rangefinder and a digital compass. A cooled 480 X 640 InSb focal plane array and multi-FOV reflective optics provide capability to acquire and identify targets at ranges of several kilometers. The LRF and compass facilitate hand- off to remote fire power. M²IS provides the soldier an integrated surveillance, targeting, and fire control system that consumes less than 6.5 W and weighs less than 7.5 lbs. This paper describes measured performance and capabilities of the system.

This paper describes two camera systems based on the advanced 160 X 128 uncooled micro-bolometer FPAs. The UL3 ALPHA camera is in production and takes advantage of the patented bias equalization FPA performance to produce the world's smallest IR production camera. UL3 ALPHA weighs less than 195 grams, uses 1.5 W of power (nominal) and has a overall dimensions of 4.3 cm X 4.3 cm X 7.5 cm. ULS ALPHA production cameras have demonstrated 62 mK NEdT operation with over 99% operability.

One can detect and track vehicles and personnel using a three-component seismic velocity transducer. Persons or vehicles moving over ground generate a succession of impacts; these soil disturbances propagate away from the source as seismic waves. Because the soil is an elastic medium both vertical and longitudinal waves propagate, diminishing in intensity as R^-2. Furthermore because the surface of the soil is the boundary of an elastic space, a Rayleigh surface wave is also generated, diminishing in intensity as R^-1. This surface wave is a vector wave that can be used to track the source. In addition to the classic model of surface waves on an elastic half space we discuss special features of seismic measurements. Among these are: contamination of the seismic signal by local acoustic waves, the excess non-geometric attenuation of the seismic signal, the influence of reflections from layered soil in tracking personnel, and finally a method of ranging using the periodic impact signature of vehicles.

This paper describes the development of a robotic system that will provide limited mobility capability to ballistically deployed sensors for acquisition of intelligence from building interiors. The future of warfare is expected to shift from large-scale battles to smaller conflicts in increasingly urban environments. Success in such situations can be improved by providing the war fighter detailed information about the location, activities, and capabilities of the opponents. A limited mobility capability adds to the usefulness of grenade launcher deployed sensors in two important ways. First, it relaxes the targeting accuracy requirement allowing a higher probability of successful placement even at greater ranges. Second, it increases the covertness of the sensor by allowing precise placement in an unobtrusive location. Underlying technologies required to implement this approach include a method of attaching the sensor package to the wall and shock tolerant electronics.