WO2008144755A1 - Standoff detection of concealed weapons and explosive devices by ultrasound diffraction radar - Google Patents

Abstract

This invention regards detection of concealed explosive devices; an embodiment comprising detection of explosive devices concealed under clothing from a safe stand-off distance of 30 to 50 meters. The technology employs a phased-array ultrasound radar (SONAR) combined with a proprietary spatial processing of the waveforms to extract the physical dimensions of the reflecting object from the diffraction pattern on the reflected wave front. The determination of the physical cross-section of the reflecting object combined with total reflected power then allows a determination of the scattering coefficient of the reflecting material which is directly related to the composition of the scattering material. The technique works in real-time on multiple moving subjects with sampling rates of 5 to 20 Hz. Detection is fully automatic with detected objects overlaid onto live video of the subjects being screened.

Description

STANDOFF DETECTION OF CONCEALED WEAPONS AND EXPLOSIVE DEVICES BY

ULTRASOUND DIFFRACTION RADAR

CROSS REFERENCES TO RELATED APPLICATIONS.

This application claims the benefit of United States Provisional Application For Patent Ser. No. 60/931,248, filed 21 May 2007, the contents of which are incorporated by reference herein in their entirety.

FIELD OF THE INVENTION.

This invention regards detection of concealed explosive devices.

BACKGROUND OF THE INVENTION.

The idea of using ultrasound to locate and identify objects at a distance, generally for application to robotics, dates back to the early 1960s. However, the basic approach of forming images from the reflected energy and the inability to propagate sufficient ultrasound energy to adequate distances doomed these attempts to failure for any application other than simple ranging of nearby objects.

Success in applying ultrasound techniques to the remote (stand-off) detection and identification of concealed weapons and explosive devices on human subjects hinges on two new ideas: 1) driving the transducers at resonance with controlled modulation to achieve significant wave propagation alongside of phase and amplitude measurements, and 2) abandoning imaging in favor of object size determination though the use of diffraction physics - hence the name Ultrasound Diffraction Radar (UDR). Target material determination follows directly since the returned echo is a factorable function of reflectivity and size. The UDR approach has allowed building a device capable of detecting, locating, and identifying objects obscured by clothing at a safe, stand-off distance of about 30 meters, with the potential of achieving 50 meters. Hard materials (e.g., metal, glass, and ceramics) are necessary and essential to an explosive or kinetic device that is intended to cause fatalities and serious injuries, regardless of chemical explosive composition. Dense and non-porous materials in general are readily detected and discrimination is enhanced by overlaid clothing materials thus allowing for the detection of concealed explosives without a hard component. Ultimately UDR technology may be able to discriminate between materials differing only slightly in acoustic properties. UDR is unique in being able to detect these materials at safe distances.

Prototype UDR devices have demonstrated detection ranges to 35 meters and the basic physical principles for measuring and interpreting echo diffraction patterns. Improvements in range, sensitivity, and composition discrimination are expected with modified transducer and sensor array designs, and with refinements in algorithms.

BRIEF SUMMARY OF THE INVENTION.

Utilizing an embodiment of this invention, Ultrasound Diffraction Radar (UDR) detects and characterizes concealed objects at distances up to 30 to 50 meters by using an acoustic phased array (SONAR) to measure and interpret the spatially extended diffraction pattern that is superimposed on an acoustic reflection. This diffraction pattern originates in the shape and properties of the reflecting surface. The pattern can be sampled by a few large aperture sensors to measure the rough size, shape and reflectivity of the target, in contrast to imaging which uses a large number of tiny sensors to produce information that is overly detailed for computer analysis and yet not easily interpreted by a human operator. UDR vastly extends range and sensitivity for ultrasound sensing while providing information that is optimally relevant and easily interpreted by computer or personnel to assess the likelihood of threat posed by a concealed object.

UDR equipment comprises a phased array of ultrasound transducers with electronics, embedded computer and display, and a video camera mounted coaxially and matched to the field of view of the sensor array. The array generates high power ultrasound pulses with adjustable focus and beam steering and measures the returned echoes to obtain a map of the reflected wave on the surface of the array. The wave map is interpreted using principles from diffraction physics to obtain the rough shape and size of the target. Its location is obtained using time and angle of arrival. The target material is categorized based on its reflectivity. Bomb and weapons materials produce strong reflectivity in contrast to clothing, and concealment produces well resolved echoes for the concealing object and the concealing garment. Threat likelihood is evaluated by a computer algorithm that considers target size, shape, material type and concealment. The complete analysis is done rapidly for each pulse and superimposed in real-time on the display obtained from the video camera. Pulse rates are 5 to 20 Hz.

BRIEF DESCRIPTION OF THE SEVERAL VIEWS OF THE DRAWING.

Figure 1 is a graph depicting a Fraunhofer diffraction profile for a circular reflector.

Figures 2A and 2B are graphs depicting reflection intensities as a function of angle.

Figure 3A is a graph depicting transmission through leather/wool.

Figure 3B is a graph depicting transmission through nylon/polysester.

Figure 4 is a graph depicting reflectivity ratio for metal and plastic as a function of embedding material.

Figure 5A illustrates a detected "hard" echo from an individual carrying a concealed exploding vest device.

Figure 5B illustrates a detected "soft" echo from same subject without view of the device.

Figure 6 is a graph depicting intensity vs. distance at two carrier frequencies.

Figure 7 is a graph depicting empirical detection limits at 49 kHz.

Figure 8A illustrates a transducer layout for a focus calculation with two adjacent transducers.

Figure 8B illustrates a vertical focus for a focus calculation with two adjacent transducers.

Figure 8C illustrates an horizontal cross-section for a focus calculation with two adjacent transducers.

Figure 9 is a schematic of a UDR hardware overview.

Figure 10 is a schematic of UDR hardware components.

Figure 11 is a schematic of UDR software components.

DETAILED DESCRIPTION OF THE INVENTION.

In one embodiment of this invention, the basic UDR device consists of an array of matched ultrasound transducers driven at the resonance frequency to form a short outgoing pulse of ultrasound energy. The pulse length is matched to the Q of the transducers to form a well shaped envelope. The beam can be broad or focused and steered. The returning energy reflected from targets is recorded with amplitude and time resolution at each point in a regular array of transducers. The transducer signals are processed as a group to extract valid echoes from uncorrelated noise. Echoes are processed as a group to locate and characterize targets.

Target angular coordinates are obtained from angle of arrival (AOA) calculations (relative phases over the array elements obtained by correlation methods). Target range is obtained from arrival time or by triangulation among multiple units. Both angle and range are corrected for atmospheric effects as will be explained. Target size and shape are obtained by interpreting the diffraction pattern measured over the surface of the sensor array. Prototype work has been conducted using a linear sensor array and fitting the sensor intensities to a simple target model (rectangle or ellipse to obtain size parameters). In this proposal, 2D array data will be reduced using Fourier transform methods and this will be compared with an extension of the algebraic method used with the linear array prototypes.

Atmospheric effects that distort the reflected wave will be removed using a modified version of the method of artificial stars developed in astronomy. This method consists of placing a fully characterized small target or source in the field of view and observing the resulting diffraction pattern. Since the expected pattern is known precisely, the observed pattern can then be processed to extract the convolving (distorting) function, and this function then used to correct the current or subsequent frame. This method, generally referred to as flat-fielding, can be concurrent or interleaved with each active pulse if necessary. This may not be required in indoor conditions.

Once target size parameters are obtained, the echo intensity is normalized for position and effective size (reflective area) to obtain target reflectivity. High target reflectivity is characteristic of hard bulk materials as compared to fibrous materials such as clothing. Hard materials under fibrous and non-fibrous clothing, are further distinguishable by comparative hardness, and exhibit a specific signature as described later under theory of reflectivity. Threat recognition is a fuzzy logic function of target size, reflectivity, pulse shape, and concealment. The output is a threat assessment rating.

For ease of use, the transducer array is combined with a co-axial video camera matched to the field of view of the UDR device. Targets are scaled to size in the display and color coded for threat assessment probability.

Algorithms and Equipment R&D:

Concealed objects empirically produce a characteristic pattern of multiple echo signatures that seem to arise in layering of cloth on hard bulk materials. This phenomenon is understood in general terms. A detailed understanding will be important in predicting performance and in advancing UDR technology. The phenomenon will be elucidated by constructing a variety of layer systems ranging from ideal to realistic and using them to collect data sets for offline study and testing. The selected algorithms for this capability will be tested using measurements collected for a range of realistic target materials and shapes concealed under a range of realistic garment materials and thicknesses.

Theory:

UDR is described by theoretical models of Fraunhofer diffraction, ultrasound reflection and transmission, propagation and absorption of ultrasound in air, and the physics of ultrasound transducers and resonance.

Fraunhofer Diffraction:

Fraunhofer diffraction can be thought of as a kind of diffraction where an aperture or single object diffracts with itself. The resulting reflection pattern is not an image, but rather a Fourier transform of the aperture or reflecting surface. As a Fourier transform, large scale parameters such as overall size and shape and intensity can be obtained to good approximation from a just few measurements across the reflected wave. In contrast to imaging with a sensor whose aperture is divided among a large number of small aperture elements, with Fraunhofer diffraction we useful information is obtained from a small number of large aperture detectors. The emphasis in UDR is on obtaining useful information with a benefit over imaging of a factor of order one million in sensitivity, and greatly reduced integration time.

The general requirements for obtaining Fraunhofer diffraction are coherent monochromatic energy and small Fresnel number, i.e. F « 1. In terms of distance , wavelength λ and target size , Fraunhofer diffraction is obtained when or . For targets 5 cm wide at 50 kHz { λ ~ 0.7 cm),

Fraunhofer diffraction occurs at distances larger than 0.1 m. For targets of 30 cm in size, at 50 kHz, Fraunhofer diffraction occurs at distances larger than 3.4 m.

The equations for Fraunhofer diffraction are obtained from scalar diffraction theory in the regime where the distance to the projecting plane is large compared to the size of the reflecting surface or aperture. The starting point is the Fresnel-Kirchoff diffraction integral. The wave function at the projection plane at distance R, is , where x', y' are coordinates in the aperture or reflector, is the aperture facture, k is the wave number , and C is a constant proportional to the amplitude, .At large distances, x << R and y << R, the Fresnel-Kirchoff integral can be approximated by a two dimensional Fourier transform, /_n . = x/R and v = y/R are angular

coordinates in the projection screen. An inverse 2-d Fourier transform of the diffraction pattern wave function then gives the reflecting surface to a level of detail that depends on the number of measurements across the wave front. For acoustics, the wave function represents the sound pressure and intensity is given by the square of the wave function, I = \ψ\ . Transducers produce a voltage proportional to the sound pressure level.

The Fraunhofer diffraction integral is obtained with the approximations that the inclination angle for the surface or aperture is small, that the incident energy is a plane wave, and that the parameter R is a constant, i.e. that the surface is smooth compared to a wavelength. These are reasonable assumptions for practical UDR measurements. The probe energy originates at the observation point, so the inclination angle is inherently close to 0, and the targets are comparatively small objects at large distances. Winds at 5 mph (2.5 m/s) introduce an obliquity of 8 mrads (discussed later). Larger deviations from low obliquity occur if two units operate together (to be explored as option in this effort.) Effective surface roughness is reduced with focus to a particular area and the effective reflector surface area is reduced by geometry to that part of the surface that does not curve or angle away from the detector.

Parameters for simple model shapes can be obtained, or approximated algebraically, from a small number of measurements across the reflection pattern. For example, for a _^circular surface or aperture with radius the

... ₂ J^kaθj diffraction pattern is ψ(θ) = 2πa C where is a Bessel function ,Qf₂ the first kaθ kind. The angular part of the intensity functio 2 4 Jy[kaθ) n is I(θ) = 4π C a At kaθ zero angle the intensity is . Scanning across the reflection produces the following intensity pattern as a function of the angle coordinate

The width of the reflector can be obtained algebraically in two ways. The first is that the first zero occurs at . The second is that the intensity function can be expanded in a polynomial about the center or any of the zeros or higher order maxima. The angular part of the intensity function near the center is b,

e

xpanded as above, witrf kβ for . For a rectangular reflector, the intensity function is . The intensity at zero angle is again . The angular part of the intensity function can be approximated near the center as . The expansions for the elliptical and rectangular intensity patterns differ in the quadratic coefficient. To determine whether a shape is square-like or round-like, both expressions can be extended to the next term and the ratio of the coefficients will indicate which

reappear, and this implies that one must sample the wave front at 5 or more positions to determine the coefficients.

The theoretical model for intensity was checked by measuring intensity as a function of angle, as shown in Figures 2A and 2B. The data was collected indoors using a single channel. The solid lines are theoretical curves for the echo intensity and the hatch marks are experimental measurements. Figure 2A shows intensity as a function of vertical angle, the second as a function of horizontal angle. Positioning was crude, hand manipulated using a laser to measure the angle, and fluctuations are probably due to air motion and temperature gradients as discussed later. Even so, the fit to the theoretical curves is good.

Acoustic Reflectivity Coefficients:

Acoustic reflectivity is a function of both contrast and structure. For bulk materials in a medium such as air or embedded in another material, the important quality is the contrast between the acoustic impedance on either side of the interface. For a layer of material, for example plastic or a thin sheet of metal, reflectivity can vary with thickness from a maximum that depends on the materials to total transmission. Detection of terrorist devices concealed under clothing is then a function of the total structure with the device and clothing compared to the absence of the device. This includes considerations for reflectivity at the air interface, at the device clothing interface, at the clothing flesh interface, and the thickness and nature of the clothing. In practice, hard objects under clothing produce strong signatures compared to the clothing itself, and there is a distinct signature for concealed devices. Consequently, the echo signals of persons carrying such devices are readily distinguished from those of persons who are not. The following describes the theoretical models for bulk and layer reflectivity.

Acoustic reflection and transmission at normal incidence for a bulk solid in a medium such as air, is depicted in the following diagram. The acoustic impedances are given in terms of the density and speed of sound on each side of the interface.

and

reflection

coefficient cc_R = R and the transmitted power fraction is then a_τ = l-a_R = AZ₂Zj[Z₂ +Z₁)²

For a thin layer the impedance is replaced by a quantity called the input impedance. The configuration is shown in the following diagram:

The reflectivity at the first surface is then

In the general case, reflection and transmission depend on the thickness of the layer and the bulk material behind the layer. There are thicknesses where the layer becomes transparent for almost any material. For materials with air on both sides, the reflection coefficient simplifies to

reflectivity follows the relation a_R ~ ξ k_od . At small thickness the reflectivity drops, and the layer transmits the acoustic energy. For non-fibrous materials such as plastic or rubber, wavelengths are measured in centimeters, substantially larger than the thickness of typical clothing layers.

Figures 3A and 3B show transmission through (a) a heavy wool lined leather jacket to obtain a reflection from a metal surface covered by the garment, and similarly for (b) a polyester filled nylon winter jacket. The green line is obtained from the metal object by itself.

In practice, hard objects worn under concealing clothing produce a stronger reflection compared to that of a non-threat clothed person, and the multiple echoes seen in Figures 3A and 3B are characteristic of concealed objects. For porous low impedance clothing materials the concealed object intensity appears large behind the clothing layer. Similar line shapes as above have been obtained at distances of 25 meters.

Table 1 lists data for some representative materials of interest. While differences in impedance are large, contrast with air is generally large for all the listed bulk materials. In clothing made of higher impedance materials, the clothing layer enhances the differences in reflectivity among the hard materials. In a sense, the greater the effort to conceal, the more easily recognized the target becomes. Material Speed of Wavelength Wavelength lmpedanc

Steel 4.0 8 cm 13 cm 45

Glass 5.6 11 cm 19 cm 14

Nvlon 2.6 5 cm 9 cm 3

Rubber 1.5 3 cm 5 cm 2.6

Table 1. Acoustic Properties. Of Representative Non-porous Materials Figure 4 shows the ratio of reflectivity of metal to plastic as a function of the bulk impedance of an adjacent material. This is a simplification from a detailed finite layer calculation, to illustrate the enhanced differentiation between the two materials that comes with increasing impedance in the adjacent material.

Figures 5A and 5B show echo indicators color coded for intensity only and superimposed on video for a subject wearing an exploding vest under a light mid-eastern style outer garment. The detector is a laboratory prototype with a four element horizontal linear array, and a vertical focusing transmitter pair in the center flanked by two pairs of out of plane detectors. The out of plane detectors are used for alignment for idealized target studies. Here they are used to extract crude vertical angle values.

Sound Propagation:

Acoustic energy propagates in air as dilatational waves, also known as longitudinal or pressure waves. The velocity of sound in air is given by c = yJγRT where ^is the adiabatic index (1.40 for air), R is the gas constant (287.053072 J -kg^"1 -K^"1 for air) and T is the temperature in kelvin degrees. This can be approximated by c ~ 331.5 + 0.6T° (C) with the speed of sound c obtained in meters per second. For a point source, neglecting absorption and other loss

product P ^c formed by the density and the speed of sound is the acoustic impedance, about 400 to 430 Rayles for air. Common usage refers to a logarithmic sound pressure level (SPL), with 0 dB at 20J^^a , with IPa = lN/m² . In the absence of losses, sound pressure levels at different distances are related by

with molar humidity h from H% relative humidity ⁿ — (H%) P 10 I ^τ J J and, ^₀ = 293.15° K . The net atmospheric attenuation in dBvjs given by j r = ^αTθθ " ^TalD'^{e 2 l ists} absorption coefficients (dB/rm) calculated according the method described above, for combinations of temperature and humidity with air pressure 1 atm, at 30 kHz.

Attenuation in dB/m at 30 kHz. 1 Atm -i n c 0 C 1 0 C 70 C 3O C 40 C 50 C

10% 0.141 0.156 0.203 0.343 0.678 1.167 1.399 30% 0.159 0.233 0.455 0.859 1.064 0.917 0.716 50% 0.188 0.339 0.684 0.936 0.826 0.639 0.512 70% 0.223 0.448 0.794 0.833 0.656 0.510 0.432 90% 0.262 0.543 0.807 0.722 0.551 0.441 0.395

Table 2. Attenuation Coefficient in Air at 30 kHz.

Figure 6 shows laboratory measurements of reflection intensity as a function of distance for two prototype UDR sensors operating at two resonant frequencies: 49 kHz (blue circles), 31 kHz (red squares). The target was a specular metal target. The solid lines are the theoretical curves calculated for the weather conditions at the time of the measurement. The data follow the predicted intensity distance relationship well within experimental errors (positioning and alignment).

Figure 7 shows echo intensity normalized for size and reflectivity for soft targets (a person) and a flat (specular) metal target using the 49 kHz sensor at current power level. Each curve corresponds to a different value of the absorption coefficient. The dotted curves are the alpha curves for the prevailing weather conditions at the time of the measurements. Horizontal dotted lines are each drawn to intercept the corresponding dotted alpha curve at the empirically determined detection limit (S/N ~ 5).

Note that measurements of a person repeated under different conditions, give nearly the same horizontal line. The technique is reproducible and can be used to predict performance under different weather conditions, power levels, and carrier frequency. At 30 KHz, alpha is reduced by about 1 dB/rm under typical conditions, and the detection limit intercepts the corresponding alpha curve further to the right in the graph, by about 10 meters. This is similar to the result previously obtained experimentally by comparing 49 kHz and 30 kHz side by side. Increasing power, either by increasing the driving amplitude or by driving more transducers simultaneously, moves the detection limit further down the curve.

To predict performance in anticipated weather conditions, the atmospheric absorption coefficient is calculated for average summer and winter conditions and then used to obtain the following detection limits for 30 kHz at current power levels.

Detection limits at 30 kHz Summer Winter

Soft Tarαets 23 meters 38 meters

Hard Tarαet (flat 32 meters > 50 meters Table 3. Estimated detection limits at 30 kHz Transducers:

The transducers currently in use for UDR are of conventional, commercial- off-the-shelf (COTS) design with proprietary improvements on the design to give the transducers improved efficient and controlled pulse shape.

Beam Width:

The beam spread (and detection cone) for a sή^e ψafςiβώj^βf is

is a Bessel function of the first kind. The first zeros occur at ^D where

D is the diameter of the transducer. Current transducers operate at 49 kHz with a diameter of about 5 cm. A new design is being tested that operates at 30 kHz with the same diameter. However, the important size for focusing calculations is the width of the active portion of the array. The proposed configuration will drive one transducer, or as many as 25 transducers. The effective width of the transmitter array then varies from 5 cm to 30 cm, giving a field of view of 1.3 to 7.9 degrees at 50 kHz or 2.2 to 13.2 degrees at 30 kHz. The beam can be fully configured from pulse to pulse.

Transmitter Array:

For an array of N transmitters (transducers), the net outgoing wave is

. The coefficients C₁ contain the amplitude for each element and the distance from that element to the target or observation point. The angles θ, are similarly calculated from the sensor to the target or observation point. Intensity is as usual, the square of the wave. ^ = H ■ Figures 8A, 8B, and 8C show the intensity maps calculated for two adjacent transducers.

These intensity maps have been checked with a two sensor transmitter array. The proposed configuration will be able to drive from 1 to 25 transducers with independent amplitude and phase to focus and steer beam. The phases and intensities at each transmitter element to achieve a given focus are determined by the identical Fraunhofer diffraction problem.

Receiver Array: The size of the transducer array sets the range of target sizes that can be measured. Small targets produce broader reflection patterns (wave fronts). Waves that are the width of two transducers or less are aliased. Waves comparable in size to the array will reduce the precision of the measurement by reducing the signal levels over some of the sensors. At the other extreme, waves that are much broader than the array reduce the precision by reducing the difference in intensity from one position in the array to the next. The optimum arrangement is with the array somewhat narrower than the expected wave front.

The reflected pattern full width at half maximum (FWHM) is approximated from Fraunhofer diffraction as, . To establish a suitable range for the size of the array, we considered the widest and narrowest reflection patterns of interest. At 30 kHz, the wavelength is approximately 1.1 cm. For a 10 cm wide target the reflection FWHM is about 110 mrad, or 3 meters at 30 meters distance. For a 35 cm target the full width at half height is about 30 mrad or 30 cm at 10 meters.

The accuracy of the system will be related to how well the device can measure the intensity angle relation for the reflection within the width of the sensor array. For this purpose we consider a square, with size 2a x 2a. As described above, the intensity near the center varies as For a 30 cm array at 30 meters, is about 5 mrad. For a 10 cm target, with a reflection centered on the array, the intensity at₂the edge of the array compared to the center is .007 . in other words, the

30 cm sensor

array measures about 1% difference in intensity across the subtended width of the reflection for a 10 cm object at 30 meters. The current transducers show comparable or smaller standard deviations when there are no wind or temperature gradients.

Power, Sensitivity, and Theoretical Detection Limits:

Output for the first specimen 50 kHz transducers was measured at 1 meter, giving approximately 130 dB for a 10 V rms sine wave input, or in terms of amplitude, 4.5 Pa/V. Response was measured at 134 dB, giving a 0.5 V amplitude signal or about 5 mV/Pa. Transducers are currently driven at 1 kV amplitude (2kV peak to peak) to provide 4.5 kPa per element, or 167 dB at 1 meter. However, the output is a pulse of about 20 cycles in duration, or 0.1 ms, and five such pulses may be sent per second. The power level normalized to one second is under 10 Pa, or about 113 dB. A 25 element array projects a pulse level of 195 dB at 1 meter, and after a 60 meter round trip at 1 dB/rm, the return echo from a unit reflector is 135 dB. For a 10 cm target at unit reflectivity, the echo level at the detector would be about 2.4 Pa and the detector output would be about 12 mV. This is an easily recorded signal level. For a target at 50 meters, the round trip intensity is 1.1 Pa for a large target, scaling to a small target gives 0.02 Pa, and the sensor response is about 0.1 mV. After amplification at a gain factor of 20,000 the signal is 2.2V, the amplifier noise level is 5 mV, so the signal should be well within detectable limits. The signal approaches the noise level with an additional 50 dB of attenuation, or another 0.5 dB/rm for a 100 meter round trip (50 meter target).

Operation at 30 kHz keeps the attenuation to around ldB/rm or less for a wide range of conditions, whereas attenuation at 50 kHz can reach 1.5 dB/rm and more during hot dry conditions. The theoretical limits for the device operating at 30 kHz are therefore well above 50 meters for a wide range conditions.

Effects of Wind and Non-uniform Environment:

Atmospheric effects can modify UDR signals in three ways: (i) Absorption and detection limits are modified as discussed above, (ii) Apparent target position and orientation angle can be altered by wind, (iii) Phase and amplitude across the echo wave front can be altered by non-uniform wind or temperature gradients. Humidity and pressure gradients are typically smaller effects. Wind and temperature effects are addressed in the following. Propagation time is given by the distance to the target divided by the speed of sound. Kcho = rl c . The change in propagation time for small changes in the speed

"^ echo _ dC of sound from an average velocity ^co is obtained from the derivative, ^~~ echo

Using c = ,JγRT , the change in arrival time for a small variation in temperature from an average temperature ¹O , is given by ^~ ---— or more conveniently,

(U echo dC -I dl ^lecho ^{Z 1} O

For reference, the period of a 30 kHz wave is 33 microseconds and for a 50 kHz wave 20 microseconds. The change in arrival time per meter per degree of temperature difference is about 5.5 microseconds per meter of transit, per degree of difference in temperature, or in terms of phase, about or per degree of temperature gradient per meter of transit. Temperature gradients can also modify the air absorption coefficient by up to about 0.03 dB/rm per degree. In 30 meters, the change can be 1 dB, or about 25%. The effect of air motion is modeled by assuming that the air currents simply add in a vector fashion to the propagation vector for the acoustic wave. The change in arrival time per meter, for a difference in wind speed of about 1 mile per hour (0.45 m/s) in zero wind conditions, is about 4 microseconds, about or per mph of wind gradient per meter of transit. Wind also effects UDR by displacing the apparent position and orientation of the target. Angular coordinate, path length, and orientation are important for target display and data reduction. The change in apparent angular coordinate for a target in a transverse wind is given by At standard temperature (273.15 K) the shift in apparent angular coordinate is about 1.5 mrad per mph of transverse wind. A shift in apparent angular coordinate produces a corresponding shift in apparent target orientation.

The field of view at 30 kHz is 220 mrad for one transducer and 44 mrad with array operated at maximum focus. At 50 kHz the range is 140 mrad to 28 mrad. The reflection pattern from a 20 cm wide target is about 50 mrad wide at 30 kHz and about 35 mrad at 50 kHz. Wind fluctuations on order of 1 mph produce fluctuations in target position and orientation that are small compared to the FOV and reflection pattern width.

This analysis indicates that the largest effects due to weather conditions are phase shifts from temperature gradients and wind gradients. The next largest are in amplitude from variations in absorption coefficient over long distances. The smallest effect is from effective target coordinate and orientation displacement by wind.

Wind gradients can be present near obstacles, and both temperature and wind gradients can be present near the open windows or doors of heated or air conditioned buildings. Therefore, the ability to remove these effects from the measurements is essential for operations in realistic environments. As previously mentioned, the compensation for these effects is through the process of flat- fielding, an adaptive technique using echoes from a known source or reflector to extract the convolving function then using this function to correct the next pulse. Formally the relationship is expressed in terms of the product Fourier transform pair, where h(x) represents the reflecting surface and g(x) represents the effect of gradients in temperature or wind. The Fourier transform is then the product of the Fourier transform of the true scattered wave front H and the convoluting air scattering or transfer function G. For a point source, the true function is flat, i.e. 1, and the net reflection pattern is that of the transfer function G directly, and that same G is the correction for any other wave front traversing a similar air path.

Threat Assessment:

The data accumulated by methods described in this document generate several critical parameters: a) the size of a detected object in two dimensions (2 parameters); b) the position angle, that is rotation of the long direction about the reflection vector (1 parameter); c) the reflectivity of the detected object (1 parameter); d) concealed/open object (yes/no); and e) the reflection coefficient of the concealing material if present (1 parameter). These factors constitute the basis for a fuzzy rule base which can be utilized for threat assessment. Each rule is a multiple-input single-output (MISO) Mamdani type linguistic rule which takes the form of the following sample: If size is large and orientation angle is near and reflectivity is very high and concealed is yes then the threat is very high.

The fuzzy rule base itself is generated from experimental data under expert supervision. The underlying fuzzy sets are triangular and divide each universe of discourse into seven (not necessarily equal) regions. Defuzzification is parameterized and utilizes BADD (Basic Defuzzification Distribution). Processing is very fast and results (i.e., final threat assessment) are generated in real-time. At the same time, a discretized property set is derived from the obtained signature. This property set forms the input to a multi-category fuzzy classifier whose output serves to identify the signature both as a threat and as a specific type of threat. The techniques used are well-established. Let {p_lr p₂, ..., p_n} be the n discretized properties of the signature (normalized so that all values are in the closed unit interval), and let {p'_lr p'₂, ,,,, p'_n} be the n discretized properties of the i^th reference prototype (similarly normalized). Then the similarity vector {s'_lf Sf₂, ■■■, s'_n} for the input signature with respect to the i^th reference prototype is given by s\ = [(I + W) ( I 1 - (Pj / Pj) I )]^~2D where I/I/ and D are positive constants. A crisp estimate of similarity between the input signature and each reference prototype can be obtained by standard de-fuzzification techniques. The crisp estimates can then simply be sorted to obtain a final best estimate E of the category of the signature (metal, glass, ceramic, ). The values of the reference prototypes and their number are obtained experimentally. Since computation of similarity vectors is very fast, it is feasible to use as many as seem necessary. The estimator, E₁ can be used directly as a threat indicator or can be used as an additional input for the fuzzy rule base described above. System Components:

The UDR system comprises (i) a sensor array including transducers, electronics, and data acquisition and pulse control interfaces, (ii) a video camera with telephoto lens matched to the field of view (FOV) of the array, (iii) a computer with control and display interfaces, (iv) GPS and wireless network for remote control and monitoring and for coordinating with other units, and (v) miniaturized sensors for ambient temperature, pressure and humidity. The device is depicted in Figure 9.

Figure 10 presents a component view of the UDR hardware system.

Sensor Array:

The sensor array comprises a 30 cm x 30 cm square planar array of transceiver elements, arranged in a 5 x 5 grid. Each transceiver element comprises a 5 cm diameter ultrasonic transducer, a signal amplifier, a pulse driver circuit, and a circuit that allows the transducer to be shared by the signal amplifier and pulse driver. The array size and number of transducers are in conformity with theory as discussed above, and at the minimum deemed necessary in order to control weight, size, and costs.

Transducers:

Output for the 50 kHz transducers is 4.5 Pa/V. Response is about 5 mV/Pa. 30 kHz transducers provide similar performance. Transducers are driven in present configuration, with 1 kV amplitude waveforms (2kV peak-peak) and produce a sound pressure level 167 dB (4.5 kPa) at 1 meter. However, a pulse is about 20 cycles at 50 kHz, and five such pulses are sent per second. The sound pressure level normalized to one second is about 113 dB (10 Pa).

Signal Amplifiers and Computer Interface

Signal amplifiers provide digital controlled gain from 20 to 20,0000. The inputs and outputs are differential, the maximum output is +/- 10V and the noise is under 5 mV at maximum gain. The amplified signals are interfaced to the embedded computer by an analog input data acquisition (DAQ) card. The DAQ card simultaneously converts the inputs to 16 bit digital values at a rate of up to 1 megasample per second per channel. The data is transferred into the computer by interrupt driven DMA so that there is close to no load on the computer for the data transfer. DAQ input card is COTS.

Pulse Drivers and Computer Interface

Transducers are each driven by a high voltage pulsing circuit. The circuit board pulses the transducers from a high voltage line in response to a pulse train, and at the end of the pulse train and when disabled it connects the transducer to the signal amplifier. The enable is controlled by a digital line from a FPGA pulse controller. The FPGA pulse controller enables transducers for pulsing during the pulse train, and time shifts the pulse train by a controllable amount for each transducer. The timing is controlled digitally for each transducer. The focus tightens as more transducers are enabled and the time shifts steer the beam. An ideal focus, to a point, corresponds to a flat field at the array and in maximum focus the operator is most interested in a narrow region including a particular subject. So the size and added production expense for individual amplitude controls is unnecessary. The FPGA pulse controller is driven by a digital pulse output from the host computer. The pulse train is generated by the DAQ card used for the analog input. A separate parallel interface controls the timing offsets.

Video Camera:

The video camera is a miniature high resolution low-light COTS model with a telephoto lens matched to the field of view of the sensor array. The interface to the computer is by USB adapter. A near-infrared camera may be used for night operations.

Embedded Computer and Display:

The embedded single board computer (SBC) for current efforts is a 1 GHz Intel compatible with a PCI104 connector and interfaces for LCD and LVDS displays, internet, USB, serial line, parallel port, hard drives, and compact flash system disk. The computer will be upgraded to a 2 GHz model if needed to support the processing load. The SBC is about 3.5" x 6" and runs from 5 VDC power.

Ambient Conditions Sensors:

Temperature, humidity, and air pressure are measured using miniaturized sensors that interface to the embedded computer by USB. The sensors are COTS.

Software:

An acceptable host operating system is Linux with customer device drivers to operate the DAQ cards at the hardware limit of performance and efficiency. All transfers are done by interrupt driven DMA. The processing modules and process is depicted in Figure 11.

Data acquisition consists of collecting echo sets, echo sets are extracted from data frames, and a data frame consists of a data input record that starts with a pulse sequence and runs until the end of the user specified time window. Each echo set is processed to extract target coordinates and parameters and the information is encoded onto the current frames from the video input. The pulse output sequence and analog to digital conversion record begin simultaneously. The data is transferred by DMA to a buffer, and at transfer complete, another data frame can start immediately, with the data transferring into a second buffer. The process repeats into one buffer and then into the other.

Echoes are recognized by correlation between channels and with a model waveform. The relative phases are saved for angle of arrival and shape and reflectivity calculations. Demodulation calculates the intensities by taking the power within a frequency band centered on the pulse carrier frequency. Atmospheric corrections for phase and intensity are applied to the echo set data at this stage. Concealment is detected by digital signal processing techniques. Angle of arrival is calculated from correlation lags. Alternatively, AOA is determined as part of the shape and reflectivity calculation if done by Fourier methods. Distance is calculated based on speed of sound and echo arrival time, speed of sound is calculated from weather data or from known targets. Shape and reflectivity are calculated from intensities and phases. Threat assessment processes the output from shape and reflectivity and concealment detection. Target location, shape, reflectivity and threat assessment are superimposed on the video display.

Multiunit Synchronization:

Multiunit operation is considered optional for the proposed effort. If implemented, accurate timing and position are required for each unit relative to the other units. Standard GPS will provide timing reference. Accurate relative position will be provided by timing acoustic pulses exchanged between units in a setup round alongside angular measurements. Two modes are contemplated for multiunit operations. The first mode has the units emitting pulses sequentially with unambiguous round trip timing. The normal pulse rate is then divided among the number of units in operation. The data frames are triggered by the GPS unit. The software calculations of distance, shape and reflectivity are adjusted to consider the pulse origin. The second mode has units firing in a dynamic scheduling scheme that sets the rate as high as possible while avoiding pulse pile up. Target locations have to be established by triangulation methods. Pulse code operation will be explored to resolve ambiguities.

GPS:

GPS timing is provided by a COTS card that provides a programmable trigger sync pulse. The card is controlled by the computer over a serial interface.

WIFI:

UDR devices equipped with WIFI communicate as an ad-hoc (peer-to- peer) network. The GPS cards provide precision timing so that the units can triangulate each other's positions and coordinate pulsing schedules.

Operational Scenario

The intended use of URD in an operational scenario is focused on the concept of an active zone (AZ). An AZ is ideally established by two UDR units with intersecting beams to provide target coverage from both front and rear. Because the FoV of UDR is narrow, the two units should be approximately face to face creating an active elliptical area approximately 5 by 50 meters. A security monitoring station monitoring both TV cameras in split screen mode is established at a tactically suitable point for the intended application. Since UDR units are battery powered and wirelessly linked great flexibility exists in where the AZ is established.

For point access control, to a military camp for instance, an AZ can be constrained by barricades or walls and the monitoring station placed outside of the confined area. Any detection above threshold will be displayed on the TV monitors and an alarm, both visual and audio, will be triggered. All individuals entering the controlled area regardless of affiliation or security clearance will be scanned during all entries and all exits to the controlled area. Locally implemented tactics, techniques, and procedures will establish the proper course of action upon triggering of a threshold. Possible actions include lowering of automatic gates to isolate the detected individual(s) within the AZ or training of weapons on suspect individuals with a command to halt in place. Although an AZ can operate with several individuals within it simultaneously, local procedures may dictate one individual at a time. Since scanning time is rapid a continuous stream of individuals walking at normal pace can be screened.

Although this invention has been described with a certain degree of particularity, it is to be understood that the present disclosure has been made only by way of illustration, and that numerous changes in the details of the composition, construction, and use may be resorted to without departing from the spirit and scope of the invention.

Claims

CLAIMS.I claim:

1. Apparatus for detecting potential explosive devices comprising means to analyze reflected wave energy from the device to determine size, shape, and material properties for the device, utilizing diffraction methods.

2. A method for detecting potential explosive devices comprising the step of analyzing reflected wave energy from the device to determine size, shape, and material properties for the device, utilizing diffraction methods.