Laser-induced acoustic imaging of buried land mines: experiment and modeling

Transcription

1 Laser-induced acoustic imaging of buried land mines: experiment and modeling S. W. McKnighta, J. Stotta, C. A. DiMarzioa, R. Clevelandb, and R. Royb aece Department, Northeastern University, Boston, MA bame Department, Boston University, Boston, MA ABSTRACT The use for subsurface buried object detection of high-frequency (15-30 khz) acoustic waves generated by CO 2 laser pulses incident on the surface of dry sand has been demonstrated previously. In this work, field tests of the technique have demonstrated imaging of landmine simulants buried 2.5 cm below the surface in an outdoor test track. Acoustic finite-difference time-domain calculations have given insight into the observed acoustic lineshapes and verified that the over-estimate of the target dimensions in the outdoor field trials may be related to the lower frequency detector used in these measurements. The models also suggest that a large increase in detected signal may potentially be gained by the use of a Laser Doppler Vibrometer interfacial velocity detector in the place of the present airborne microphone. Keywords: Laser-acoustic, underground imaging, land mines INTRODUCTION Acoustic sensing is a promising technique for the detection of low-metal-content buried land mines due to the large impedance mismatch between the porous soil and the solid non-metallic mine case. As in most mine detection strategies, however, the key is reducing false alarms by distinguishing the buried mines from clutter objects (rocks, roots, etc.). Most of the work in acoustic detection has focused on relatively low acoustic frequencies (<1kHz) where resonant oscillations of the soil above the compliant mine cover plate have been observed in both frequency-domain detection of ground motion above the insonified soil 1 and in the time-domain response to surface wave pulses. 2 In addition, acoustic non-linearities in the mine casing have been exploited to distinguish mines from other underground objects. 3 In contrast to these low-frequency techniques, we have reported previously on a technique to probe buried objects with highfrequency (f>20 khz) acoustic waves generated by directing a pulsed CO 2 laser at the ground surface. 4-6 At these frequencies acoustic waves are attenuated within a fraction of a meter in soil, but with the laser source it is possible to create an acoustic wave in close proximity to the buried mine, even at a safe stand-off distance. We have demonstrated that it is possible to detect the echoes from a buried object with a non-contact microphone in the air above the laser source. Since the acoustic wavelength in soil at these frequencies is about a centimeter, it is possible to probe the shape of the buried object by scanning the laser spot. In experiments in a laboratory sand box, we have demonstrated shape resolution in the horizontal plane of 0.5cm, with depth resolution of a few millimeters. 5 Shape and depth resolution of this order could provide a powerful discriminator for clutter objects, drastically reducing the false-alarm rate. In this work we report on the application of our technique to image mine simulants in an outdoor field trial. We also report on computational simulations of the laser-induced acoustic effect which have enabled us to identify the source of image features and make predictions of the sensitivity limits of the technique. EXPERIMENTAL RESULTS CO 2 radiation is efficiently absorbed in the first few microns of soil. With a sufficiently energetic pulsed source, the rapid heating and thermal expansion caused by the absorption creates a broad-band acoustic pulse. Our laser source is a pulsed laser manufactured by LSI, Inc. with a 100mJ pulse energy, 100ns pulse length, and 20 Hz repetition rate for a modest 5W average Detection and Remediation Technologies for Mines and Minelike Targets VI, Abinash C. Dubey, James F. Harvey, J. Thomas Broach, Vivian George, Editors, Proceedings of SPIE Vol (2001) 2001 SPIE X/01/$

2 Figure 1. Laser-generated acoustic pulse measured by wideband hydrophone located a few millimeters under the surface of dry sand. Figure 2. Amplitude of Fourier transform of laser-generated acoustic pulse in Figure 1, indicated bandwidth of lasergenerated acoustic source. power. It creates an acoustic snap which is audible at in air a distance of several meters from the approximately 1cm-diameter laser spot. The acoustic pulse as detected with a wide-band B&K, Inc.100kHz hydrophone 8 mm under the surface of dry sand is shown in Figure 1. The magnitude of the Fourier transform of this pulse, shown in Figure 2, indicates a peak energy intensity between 15 and 25 khz with significant amplitude from DC to 60 khz. Note that the data in Figures 1 and 2 have not been corrected for the frequency-dependent attenuation in the propagation path. The wide-band nature of this source contrasts sharply with the low-frequency sources employed in other acoustic detection schemes. In our previously reported imaging results, we used a narrow-band 30kHz PZT detector suspended in the air over the laser spot to detect the sound reflected from buried objects. This detector had the advantage of removing the omnipresent low-frequency acoustic noise from our signal, but required a Fourier-domain filtering technique to deconvolve the echo pulse from the detector ringing. It also exploited the bandwidth of our acoustic source and permitted high-resolution shape- and depth-imaging of buried non-metallic objects. To test this technique under more realistic field conditions, we took our laser source to the Northeastern University Dedham test track facility, where nine mine simulants are buried along with several clutter objects in a 2.3m x 21m track filled with screened loam and spanned by a concrete and metal track that carries an instrument cart. A photograph of the laser source on the instrument cart is shown in Figure 3. The acoustic detector for this experiment was an Radio Shack A unidirectional acoustic microphone with a nearly flat audio response below 15kHz. This relatively low frequency detector proved to be overly sensitive to lowfrequency noise and degraded the resolution of the detector as shown below. The target we concentrated on was an M-14 antipersonnel land mine simulant with the charge removed and replaced with silicone filler, buried with its top plate 2.5cm below ground surface. This mine is small in size (56mm diameter x Figure 3. Field test of laser acoustic detection at the Northeastern University Dedham test track. The laser, power supply, and data acquisition computer are mounted on the cart, and the microphone detector is suspended just above the ground over the focused laser spot. 628 Proc. SPIE Vol. 4394

3 Time (sec) Fourier-Filtered Output (arb. Units) Position (cm) Figure 4. Processed acoustic signal as a function of position as the laser source is moved across a buried M-14 land mine simulant. The detector was an acoustic microphone with bandwidth 0-15kHz and the mine simulant was located near the 25 cm point. 40mm high) and has almost no metallic content. It is considered very difficult to detect by convention means. We moved the laser cart over the centerline of the mine, stopping to take a series of eight shots every 1.2 cm. The laser was focused to a spot about 0.5cm in diameter and the microphone was suspended about 4cm over the laser spot. We processed the microphone data first by normalizing the peak amplitude of the signal to correct for different acoustic intensity created when the laser pulse struck different parts of the soil surface (embedded pebbles, etc.). Then we applied the Weiner Fourier filter as described in Reference 6. The data is shown in Figure 4. The Weiner filter was not as successful in removing detector characteristics as in our experiments in the lab with the PZT detector, and the uncanceled detector ringing is the dominant feature of the processed data. Nevertheless, there is a clear anomaly at the position of the M-14. The size of the anomalous region exceeds the 5.6cm diameter of the mine by a factor of two, but is still considerably less than the region of soil disturbed during mine emplacement (~20 cm diameter). We will demonstrate in the next section that the most likely cause of this is the reduced resolution caused by the lower frequency detector. ACOUSTIC MODELING To understand the line shapes and acoustic effects in the laser-induced acoustic detection technique, we have applied a twodimensional acoustic Finite-Difference-Time-Domain (FDTD) computation to model the problem. The soil was modeled as a linear and lossless effective media no attempt was made to simulate the porous nature of the medium. The laser source was assumed to be a point source excitation at the surface of the soil and we did not model the surface roughness or the noise environment. The calculation was second-order in space and time on a two-dimensional Cartesian grid with a march in time. The density and sound speed were taken from empirical data, and an absorbing boundary was placed at the edge of the domain. The parameters for the calculation are indicated in Figure 5, which also indicates the alternative positions of the acoustic point source and three modeled detectors: a microphone in the air above the source, a hydrophone in the soil under the pulse, and a surface motion sensor such as a laser Doppler vibrometer. The goals of this calculation were: 1) to understand the time-domain acoustic Proc. SPIE Vol

4 Figure 5. Parameters for acoustic modeling of laboratory experiment in dry sand. The position of the laser source, airborne microphone, buried hydrophone, and an interfacial velocity sensor (LDV) are indicated. waveform we measure in the experiment, 2) to compare the effects of different detection bandwidths, and 3) to predict relative detection performance from different detector types, airborne or buried pressure sensors and interfacial velocity sensors. The sensitivities of the three receivers were selected to mimic the following commercial sensors: a Radio Shack Model Electret Condenser Microphone, with a nominal sensitivity = 2.5 nv/ Pa (airborne microphone), a B&K 8103 Hydrophone with a nominal sensitivity = 12.6 pv/ Pa (buried sensor), and a Polytec Model OVD-02 Laser Doppler Vibrometer with a nominal sensitivity = 0.2 V/(mm/sec) (interfacial velocity sensor). The frequency dependencies of the detectors are not considered in the calculation, and they were all taken as point receivers. With the target modeled as a 25mm x 60mm hard rubber disk to simulate the laboratory hockey puck targets, we calculated the response at the airborne sensor as the acoustic source was moved from a horizontal offset of -20mm from the target edge to +5mm from the target edge (over the target). The airborne receiver is maintained at a constant offset of +15mm from the source. (The target edge is taken as the zero point of the distance scale.) The data for several positions of the source and receiver as they approach and overlap the target position are shown in Figure 6. The direct reception from the source to the receiver is the first feature in each trace. This is reduced from the signal in the ground by a factor of over 100 due to the impedance mismatch at the air-ground interface. The small feature following the direct pulse that appears in every trace is an artifact due to incompletely absorbing boundary conditions. When the source is -10mm away from the target edge (and the receiver is at +5mm over the target), a second feature appears in the traces at a later time. This feature grows in size as the source approaches closer to the target. By observing the time evolution of the pressure waves in twodimensional movies, the source of this second feature becomes apparent: it is a second bounce from the top surface of the target by the acoustic pulse which reflects from the top of the target, then off the air-ground interface, reflects off the target again, and transmits across the interface to the receiver. While this signal will certainly be reduced from the model calculation by the actual acoustic attenuation of the soil (which is not modeled), we have observed this second signal in our data. Figure 7 shows previously 630 Proc. SPIE Vol. 4394

5 Figure 6. Predicted signal from the acoustic model as measured at an airborne microphone when the laser source is moved closer to the edge of the mine simulant. Ds-t is the horizontal distance from the laser source to the edge of the target. Dr-t is the horizontal distance from the microphone receiver to the edge of the target. published data showing the time signal as a function of position as the laser is moved over a rubber hockey puck buried 4mm below the surface. The direct signal between 0 and 20s, the return from the top of the target near 30s, and the second bounce at 60s (previously misidentified as a reflection from the bottom of the puck) are all clearly visible. A second question that can be addressed by the model calculations is to what extent the apparent increased size of the target in Figure 4 can be attributed to the lower resolution caused by the lower frequency of the microphone detector (<15kHz as opposed to ~30kHz for the PZT detector used in Figure 7). We modeled this by observing the response at an airborne detector as the laser point-source is moved across the top of a 56mm dia x 40 mm target buried 25mm below the surface, as in the acoustic field tests. The positions of the source and receiver with respect to the target are as in Figure 6. The acoustic source is taken to be a single cycle at 30kHz, a signal which has significant intensity across the frequency spectrum from DC to above 50 khz. The signal received at the detector position is Fourier transformed and the magnitude of the Fourier component at three frequencies, 7, 14, and 28kHz are compared with the actual target extent in Figure 8. The decrease in target resolution is apparent in the figure, and the signal extent at 7kHz, in the midband of the microphone used in the experiment, is approximately twice the true target extent. This suggests that bandwidth considerations may well explain the larger extent of the image in the experiment of Figure 4, and emphasizes the resolution advantage of utilizing the entire bandwidth of the laser-acoustic source. Figure 7. Laboratory data image of target (hockey puck) buried under 4mm of dry sand, showing the second return predicted by the acoustic model in Figure 6. Proc. SPIE Vol

6 Figure 8. Predicted acoustic signal vs. laser position compared with actual target extent for different receiver frequencies, showing effect of detector frequency on target resolution. A major loss mechanism for laser-acoustic detection is the impedance mismatch at the air-ground interface which causes a reflection of about 99% of the signal reflected off the target. While the laser source produces acoustic signal directly in the ground, thereby bypassing this loss in transmission, the received signal has to be transmitted to the airborne detector above. The solution suggests itself to use a Laser Doppler Vibrometer (LDV) to detect the motion of the free surface of the ground instead Figure 9. Predicted signal received at detector positions as shown in Figure 5 for airborne microphone sensor (top), buried, hydrophone (bottom), and LDV intefacial velocity sensor (middle). Note that the voltage scale for the LDV is approximately 100 times that for the microphone. (The extra returns in the hydrophone trace result from the detection of both the outgoing and returning wave at the buried sensor location.) 632 Proc. SPIE Vol. 4394

7 of the weak transmitted pressure signal. This configuration could not only result in increased sensitivity, but would permit lookahead, stand-off detection which is very desirable from a counter-mine operations point of view. Using the configuration of Figure 5, and the detector sensitivities as given above, we modeled the received detector voltages as a function of time when the laser excitation was directly over the target. No attempt was made to estimate the different noise floors of the detectors. Nevertheless, as shown in Figure 9, the predicted detector voltage output from the LDV is a factor of 100 greater than that predicted at either the airborne microphone or the buried hydrophone. The potential signal-to-noise gain of this detector configuration would be an enormous step toward making laser-acoustic detection a reality. Tests of LDV detection with laseracoustic excitation are underway to confirm the advantage suggested by the model. CONCLUSION Laser-acoustic imaging of underground objects has been shown in the laboratory to allow detailed imaging of the shapes of buried objects. The shape and depth resolution could contribute to a drastic reduction in the false-alarm rate for underground clutter. We have demonstrated the practicality of the technique in tests using realistic antipersonnel mine simulants at an outdoor test track. The images obtained with laser-acoustic excitation and a low-frequency (<15kHz) microphone detector indicate anomalies associated with the position of the mine simulant, but the size and depth of the simulant are not as clearly defined as in the laboratory measurements in dry sand with a higher frequency (30 khz) detector. Acoustic modeling has been performed using a two-dimensional FDTD code, assuming linear and lossless propagation in a singlephase effective media. The results of the simulation accurately predict lineshape features observed in the data, and indicate that the excessive size of the image in the field trials at the test track may be a result of the lower frequency of the detector, compared with the detector used in the laboratory experiments. The acoustic model predicts a factor of 100 gain in the signal voltage may be obtained by using a LDV interfacial velocity sensor to avoid the acoustic loss at the air-ground interface. ACKNOWLEDGMENTS This work has been supported by the Army Research Office under the Multidisciplinary University Research Initiative Grant DAAG Other support from the Engineering Research Centers Program of the National Science Foundation under award number EEC is gratefully acknowledged. Computational resources were provided by the Scientific Computing and Visualisation group at Boston University. REFERENCES 1. J. M. Sabatier and N. Xiang, Laser-Doppler-based acoustic-to-seismic detection of buried mines, Proc. of the SPIE, Detection and Remediation Technologies for Mines and Minelike Targets IV, Vol. 3710, 215(April 1999). 2. C. T. Schroder and W. R. Scott, Jr., IEEE Trans. on Geoscience and Remote Sensing, Vol. 38, 1505 (2000). 3. D. M. Donskoy, Detection and discrimination of nonmetallic land mines, Proc. of the SPIE, Detection and Remediation Techologies for Mines and Minelike Targets IV, Vol. 3710, 215 (April 1999). 4. S. W. McKnight, W. Li, and C. DiMarzio, Imaging of buried objects by laser-induced acoustic detection, Proc. of the SPIE, Detection and Remediation Techologies for Mines and Minelike Targets IV, Vol. 3710, 231 (April 1999). 5. S. W. McKnight, C. A. DiMarzio, W. Li, R. A. Roy, Laser-induced acoustic generation for buried object detection, Proc. of the SPIE, Detection and Remediation Techologies for Mines and Minelike Targets V, Vol. 4038, 734 (April 2000). 6. S. W. McKnight, C. A. DiMarzio, W. Li, and J. Stott, Laser-induced acoustic imaging of buried objects, Journal of Subsurface Sensing Technologies and Applications, Vol. 2, No. 2, pp , April Proc. SPIE Vol