Academia.eduAcademia.edu

Figure 39 – uploaded by mark mckenna

See full PDFdownloadDownload figure
Software modifications were made to improve analysis results on the weapons detection system. The results are now displayed in the form of images on the LabView front panel. In the following figures, sample results showing the display of the weapons are shown. We believe that this is a much simpler representation of the results. Figure 35: Experimental set-up for CWD tests on people, with improved software

Figure 35 Software modifications were made to improve analysis results on the weapons detection system. The results are now displayed in the form of images on the LabView front panel. In the following figures, sample results showing the display of the weapons are shown. We believe that this is a much simpler representation of the results. Figure 35: Experimental set-up for CWD tests on people, with improved software

Related Figures (58)

To. ABSTRACT! (Maximum 2UU VWordas) The major findings of this effort clearly demonstrate that Non-linear Acoustics is a low cost alternative to conventione imaging methods for concealed weapons detection. Our approach is to use ultrasonics to create a localized zone where non-linear interactions generate a lower frequency acoustic wave that is steerable and is able to penetrate clothing better than direct ultrasonics. We have been able to detect guns of various sizes, box cutters, knives and distinguish weapons from non-weapons including plastic devices. Sophisticated software generated signals with customized transducer/receiver design have been able to provide characteristic information about a hidden weapon feet away and classify the weapon based on a database approach. Most recently improved software allows user- friendly display of results with images. We are confident of the readiness of our work for prototype development. In addition to being able to detect the weapons mentioned above, we have also been highly successful in imaging improvised weapons (received from the DOJ/AFRL lab) from a near range distance.
To. ABSTRACT! (Maximum 2UU VWordas) The major findings of this effort clearly demonstrate that Non-linear Acoustics is a low cost alternative to conventione imaging methods for concealed weapons detection. Our approach is to use ultrasonics to create a localized zone where non-linear interactions generate a lower frequency acoustic wave that is steerable and is able to penetrate clothing better than direct ultrasonics. We have been able to detect guns of various sizes, box cutters, knives and distinguish weapons from non-weapons including plastic devices. Sophisticated software generated signals with customized transducer/receiver design have been able to provide characteristic information about a hidden weapon feet away and classify the weapon based on a database approach. Most recently improved software allows user- friendly display of results with images. We are confident of the readiness of our work for prototype development. In addition to being able to detect the weapons mentioned above, we have also been highly successful in imaging improvised weapons (received from the DOJ/AFRL lab) from a near range distance.
Figure 1: Long range detection of concealed weapons for street systems from 15 feet. (a) NAC was tested for different weapons (variation in people, their clothing types and weapon orientation were also studied); (b) offline post processing that ensures a clean database; weapon orientation were also studied); (c) Real-time test result for a chirp signal test from a target person with no weapon; (d) Real-time test result for a chirp signal test from a target person with a concealed box cutter, set up shown in (a). 1: Long range detection of weapons.
Figure 1: Long range detection of concealed weapons for street systems from 15 feet. (a) NAC was tested for different weapons (variation in people, their clothing types and weapon orientation were also studied); (b) offline post processing that ensures a clean database; weapon orientation were also studied); (c) Real-time test result for a chirp signal test from a target person with no weapon; (d) Real-time test result for a chirp signal test from a target person with a concealed box cutter, set up shown in (a). 1: Long range detection of weapons.
Figure 2: Handcrafted weapon vs. non-weapon detected from 15 feet.
Figure 2: Handcrafted weapon vs. non-weapon detected from 15 feet.
The specimen was placed on a tissue-simulating gel under two layers of stretched fabric using a 7.5 kHz probing beam. There is no visual information on the presence of the weapons underneath. Near range detection for prison systems (6 inches):
The specimen was placed on a tissue-simulating gel under two layers of stretched fabric using a 7.5 kHz probing beam. There is no visual information on the presence of the weapons underneath. Near range detection for prison systems (6 inches):
Figure 4: Experimental set-up for conducting non-linear acoustic parametric beam mixing. Since the absorption of acoustic energy in air is frequency dependent, as the acoustic wave propagates, the higher frequencies are quickly attenuated and the difference frequency (f; - f2) becomes the dominant frequency present in the acoustic wave. However, because the resulting acoustic beam’s directionality is determined by the transducer size relative to the original high frequencies f; and f., a very narrow sound beam at the low difference frequency (f; - f2) can be achieved. Using this concept, a nonlinear ultrasonic system has been developed to provide an enhanced capability for detecting hidden weapons over traditional ultrasonic examinations. In the set up shown in figure 4, when the modulated signal is radiated from a focusing source, the difference-frequency sound generated due to acoustic non-linear propagation is concentrated at the focus. Thus, a small size low-frequency probe beam can be realized by setting the two frequencies close to each other.
Figure 4: Experimental set-up for conducting non-linear acoustic parametric beam mixing. Since the absorption of acoustic energy in air is frequency dependent, as the acoustic wave propagates, the higher frequencies are quickly attenuated and the difference frequency (f; - f2) becomes the dominant frequency present in the acoustic wave. However, because the resulting acoustic beam’s directionality is determined by the transducer size relative to the original high frequencies f; and f., a very narrow sound beam at the low difference frequency (f; - f2) can be achieved. Using this concept, a nonlinear ultrasonic system has been developed to provide an enhanced capability for detecting hidden weapons over traditional ultrasonic examinations. In the set up shown in figure 4, when the modulated signal is radiated from a focusing source, the difference-frequency sound generated due to acoustic non-linear propagation is concentrated at the focus. Thus, a small size low-frequency probe beam can be realized by setting the two frequencies close to each other.
Figure 5: Theoretical representation of sound pressure field for a non-linear vs. linear pulse propagating away from the transducer.
Figure 5: Theoretical representation of sound pressure field for a non-linear vs. linear pulse propagating away from the transducer.
a.Lab Simulated weapons testing: One of the first major successes on thi: project was the demonstration of the contras between weapon-simulating metal: and tissue-simulating gelatin media through non-linear beam mixing. A ballistic gelatin media was cast into a human torso shape to simulate a human body. Ar image of a simulated weapon on a simulated human body was achieved through « mechanical scanner. These tests use high power focused transducers operable a 140 kHz to generate a spot size of 5 mm at a distance of 14cm.
a.Lab Simulated weapons testing: One of the first major successes on thi: project was the demonstration of the contras between weapon-simulating metal: and tissue-simulating gelatin media through non-linear beam mixing. A ballistic gelatin media was cast into a human torso shape to simulate a human body. Ar image of a simulated weapon on a simulated human body was achieved through « mechanical scanner. These tests use high power focused transducers operable a 140 kHz to generate a spot size of 5 mm at a distance of 14cm.
State Department of Corrections. For all subsequent scanning tests using the non- linear acoustic approach, the experiments were set up such that the presence o1 absence of a weapon could not be detected visually, a test that most closely simulated an actual real life target. Figure 7: Improvised weapons made by inmates (Courtesy National Institute of Justice).
State Department of Corrections. For all subsequent scanning tests using the non- linear acoustic approach, the experiments were set up such that the presence o1 absence of a weapon could not be detected visually, a test that most closely simulated an actual real life target. Figure 7: Improvised weapons made by inmates (Courtesy National Institute of Justice).
The cloth is stretched over the objects. The objects were loosely draped with cloth. The specimen was placed on a tissue-mimicking gel under two layers of stretched fabric using a 7.5 kHz probing beam. No visual information on the presence of the weapons underneath. In these tests, we optimized the orientations of both, the transmitting transducer, and the receiver microphone to enhance reception of the acoustic echo from the
The cloth is stretched over the objects. The objects were loosely draped with cloth. The specimen was placed on a tissue-mimicking gel under two layers of stretched fabric using a 7.5 kHz probing beam. No visual information on the presence of the weapons underneath. In these tests, we optimized the orientations of both, the transmitting transducer, and the receiver microphone to enhance reception of the acoustic echo from the
Simulated and Experimental Absorption Data Comparison The solid line represents the simulated data and the dotted line represents the experimental data.
Simulated and Experimental Absorption Data Comparison The solid line represents the simulated data and the dotted line represents the experimental data.
Ultrasound — 53.6kHz Audio — 8kHz
Ultrasound — 53.6kHz Audio — 8kHz
8000 Hz Component of Acoustic Beams for the Flat and Focused Transducer The transducer is located at the left of the figure. The color represents the pressure intensity in decibels.
8000 Hz Component of Acoustic Beams for the Flat and Focused Transducer The transducer is located at the left of the figure. The color represents the pressure intensity in decibels.
Figure 13: A plot of calibration data for ultrasonic sound pressure amplitudes with the xy origir at 15 ft. In figure 13, we show the ultrasonic intensity (at 66.4 kHz) and the beam spread at 15feet from the face of the transducer.
Figure 13: A plot of calibration data for ultrasonic sound pressure amplitudes with the xy origir at 15 ft. In figure 13, we show the ultrasonic intensity (at 66.4 kHz) and the beam spread at 15feet from the face of the transducer.
We have conducted many preliminary studies using an unfocused transducer while building the 15 foot full-scale concealed weapons detection system. In addition to being able to reduce the beam width from 3.5 feet to 1 foot, the important advantage of designing a focused transducer was the increase in sound pressure impinging on the target. As can be seen from the above figure, focusing the beam could increase the soun pressure level on the target at which the tests were conducted.
We have conducted many preliminary studies using an unfocused transducer while building the 15 foot full-scale concealed weapons detection system. In addition to being able to reduce the beam width from 3.5 feet to 1 foot, the important advantage of designing a focused transducer was the increase in sound pressure impinging on the target. As can be seen from the above figure, focusing the beam could increase the soun pressure level on the target at which the tests were conducted.
Figure 15: A picture of the elliptical receiver dish mold under fabrication is shown here.
Figure 15: A picture of the elliptical receiver dish mold under fabrication is shown here.
Figure 16: A picture of the elliptical receiver dish with a sensitive microphone at its focus
Figure 16: A picture of the elliptical receiver dish with a sensitive microphone at its focus
Figure 17: NAC signal received from a specimen covered with clothing and no weapon underneath. Time-frequency based correlation algorithms and FFT based algorithms provide the user with options to analyze data in multiple ways. Particularly, a chirp signal was found to be best fit into the needs of CWD. A sample data from unfocused transducers is shown in the following figures.
Figure 17: NAC signal received from a specimen covered with clothing and no weapon underneath. Time-frequency based correlation algorithms and FFT based algorithms provide the user with options to analyze data in multiple ways. Particularly, a chirp signal was found to be best fit into the needs of CWD. A sample data from unfocused transducers is shown in the following figures.
Figure 18: Experimental set-up for concealed weapons detection using the parabolic receiver dis} and the Hypersonic Sound transmitter.
Figure 18: Experimental set-up for concealed weapons detection using the parabolic receiver dis} and the Hypersonic Sound transmitter.
Figure 19: NAC signal received from a plastic gun covered with clothing.
Figure 19: NAC signal received from a plastic gun covered with clothing.
Figure 20: A 32ms chirp signal created such that the source frequency chirp is starting from 0.1 kHz to 10 kHz, increasing linearly though the chirp.
Figure 20: A 32ms chirp signal created such that the source frequency chirp is starting from 0.1 kHz to 10 kHz, increasing linearly though the chirp.
Figure 21: Weapon simulants used for testing and developing analysis algorithms
Figure 21: Weapon simulants used for testing and developing analysis algorithms
Figure 22: Pictures of experimental set-up for detecting signatures through thick clothing
Figure 22: Pictures of experimental set-up for detecting signatures through thick clothing
Tube behind the coat To make the results appear pictographically, for easy interpretation, we decided to display the chirp signal in the form of spectrum (joint time-frequency plots). In the spectrogram plots shown below, even though the input chirp was a linear continuous chirp, the return signal from the tube behind the thick coat had resonance features, shown clearly in the spectrogram. These bands in the spectrogram are separated by a frequency that coincides with the resonance frequency of the tube.
Tube behind the coat To make the results appear pictographically, for easy interpretation, we decided to display the chirp signal in the form of spectrum (joint time-frequency plots). In the spectrogram plots shown below, even though the input chirp was a linear continuous chirp, the return signal from the tube behind the thick coat had resonance features, shown clearly in the spectrogram. These bands in the spectrogram are separated by a frequency that coincides with the resonance frequency of the tube.
Figure 27: Pictures showing the experimental set-up and the use of the real guns as concealed weapons.
Figure 27: Pictures showing the experimental set-up and the use of the real guns as concealed weapons.
Case 1: Person standing in the focal zone with |
Case 1: Person standing in the focal zone with |
Case3: Unknown case — Correctly identified as Case 4: Unknown case: Gun was present but the person turned away such that the hidden
Case3: Unknown case — Correctly identified as Case 4: Unknown case: Gun was present but the person turned away such that the hidden
Figure 28: Offline correlation coefficient result for a Big gun. The resulting value of the coefficient was significantly higher compared to other objects.
Figure 28: Offline correlation coefficient result for a Big gun. The resulting value of the coefficient was significantly higher compared to other objects.
Figure 29: Offline correlation coefficient for a Box cutter (actual weapon). The resulting coefficient was significantly higher compared to other objects in the database.
Figure 29: Offline correlation coefficient for a Box cutter (actual weapon). The resulting coefficient was significantly higher compared to other objects in the database.
In the left plot, the gun (middle color bar) was classified correctly indicated by a high correlation coefficient when its position was such that the barrel was facing the transmitter. When the orientation of the weapon changed, the classification result was less obvious. The test results shown in figure 28 and figure 29 were helpful in defining the mathematical correlation functions suitable for CWD data. The results were extended for tests on people; however the weapon had to be positioned in the exact same orientation under the person’s clothing. As an example, in the following picture, the weapon was hidden such that the barrel of the gun was pointing the transmitter.
In the left plot, the gun (middle color bar) was classified correctly indicated by a high correlation coefficient when its position was such that the barrel was facing the transmitter. When the orientation of the weapon changed, the classification result was less obvious. The test results shown in figure 28 and figure 29 were helpful in defining the mathematical correlation functions suitable for CWD data. The results were extended for tests on people; however the weapon had to be positioned in the exact same orientation under the person’s clothing. As an example, in the following picture, the weapon was hidden such that the barrel of the gun was pointing the transmitter.
In this figure, the effect of both time-domain and FFT domain correlation put together correctly detected and classified the gun hidden in a random position under a person’s clothing.
In this figure, the effect of both time-domain and FFT domain correlation put together correctly detected and classified the gun hidden in a random position under a person’s clothing.
This procedure was repeated for all the objects that were tested.
This procedure was repeated for all the objects that were tested.
Isolation of specular reflection from resonance features to address size of the weapon: Figure 33: Input chirp signal, with a varying frequency structure over 32 ms time, designed to resonate a wide range of objects.
Isolation of specular reflection from resonance features to address size of the weapon: Figure 33: Input chirp signal, with a varying frequency structure over 32 ms time, designed to resonate a wide range of objects.
After several of these modifications, the LabView code was tested a few times on different objects to ensure that these changes helped improve the results. Also, the number of real weapons that were analyzed was increased to five weapons and hence we had to record some more signatures for the new test objects. In all tests that followed, the weapons identified were much more obvious and the angle at which the weapon was oriented was less of a problem. Figure 34: A medium size gun (blue has highest correlation coefficient) that was classifiec correctly with the improved analysis technique
After several of these modifications, the LabView code was tested a few times on different objects to ensure that these changes helped improve the results. Also, the number of real weapons that were analyzed was increased to five weapons and hence we had to record some more signatures for the new test objects. In all tests that followed, the weapons identified were much more obvious and the angle at which the weapon was oriented was less of a problem. Figure 34: A medium size gun (blue has highest correlation coefficient) that was classifiec correctly with the improved analysis technique
Figure 36: Data analysis result with an image display for a case tested with “a person with no weapon on him”.
Figure 36: Data analysis result with an image display for a case tested with “a person with no weapon on him”.
hiding a gun as a weapon”.
hiding a gun as a weapon”.
Figure 39: Schematic diagram for the phased array system proposed for future work. This is based on an ongoing project funded by DARPA
Figure 39: Schematic diagram for the phased array system proposed for future work. This is based on an ongoing project funded by DARPA
Figure Al: Experimental set-up for show that non-linear acoustic approach has th capability to see through thick clothing
Figure Al: Experimental set-up for show that non-linear acoustic approach has th capability to see through thick clothing
Figure A2: Audio propagation through thick clothing. Due to presence of thick fabric, even though the ultrasound drops off drastically, the audio beam attenuates by a small amount and a strong signal passing through the thick fabric was recorded by the receiver. This audio signal is used as the probing beam to detect the weapon signature. The audio x axis is from x to y Hz while the y axis is normalized intensity. The ultrasonic x axis goes from xHz to Y Hz.
Figure A2: Audio propagation through thick clothing. Due to presence of thick fabric, even though the ultrasound drops off drastically, the audio beam attenuates by a small amount and a strong signal passing through the thick fabric was recorded by the receiver. This audio signal is used as the probing beam to detect the weapon signature. The audio x axis is from x to y Hz while the y axis is normalized intensity. The ultrasonic x axis goes from xHz to Y Hz.
The quality factor or Q of the tube weapon would have reduced when it was mounted on the flat background. The transducer (HSS) characteristics due to its bandwidth were also expected to play a role when the AM frequency was changed. To get rid of these effects, the no weapon data was subtracted from the weapon data and the result is shown in figure A3(b). The curve in figure A3 (b) peaks at 2.1kHz which is only due to the vibration of the target. Here, we also notice the second peak at the end of the plot characteristic of the mode we are exciting. Figure A3: A3(a) is a plot showing the response of a flat absorbing background in pink and a metal tube in blue to changing frequency of the probing beam. A3(b) is the pink curve subtracted from the blue curve to show the effect of presence of weapon. Since the tube resonance falls within this frequency range we see the bottom curve peak at 2.1 kHz
The quality factor or Q of the tube weapon would have reduced when it was mounted on the flat background. The transducer (HSS) characteristics due to its bandwidth were also expected to play a role when the AM frequency was changed. To get rid of these effects, the no weapon data was subtracted from the weapon data and the result is shown in figure A3(b). The curve in figure A3 (b) peaks at 2.1kHz which is only due to the vibration of the target. Here, we also notice the second peak at the end of the plot characteristic of the mode we are exciting. Figure A3: A3(a) is a plot showing the response of a flat absorbing background in pink and a metal tube in blue to changing frequency of the probing beam. A3(b) is the pink curve subtracted from the blue curve to show the effect of presence of weapon. Since the tube resonance falls within this frequency range we see the bottom curve peak at 2.1 kHz
In order to extend our tests to a more practical situation in the presence of a noisy environment, we focused some of our efforts in conducting NAC tests outdoors as shown in figure A4. The non-linear acoustic approach allows us to set-up probing frequencies with precise control. The narrow bandwidth of such a signal could easily be distinguished from noise as shown in figure A6. Figure A4: Luna experimental set-up moved from laboratory to outdoors to study the signal. to-noise ratio and effect of ambient noise that may interfere with weapons detection data in < real test situation. shown in figure A4. The non-linear acoustic approach allows us to set-up probing
In order to extend our tests to a more practical situation in the presence of a noisy environment, we focused some of our efforts in conducting NAC tests outdoors as shown in figure A4. The non-linear acoustic approach allows us to set-up probing frequencies with precise control. The narrow bandwidth of such a signal could easily be distinguished from noise as shown in figure A6. Figure A4: Luna experimental set-up moved from laboratory to outdoors to study the signal. to-noise ratio and effect of ambient noise that may interfere with weapons detection data in < real test situation. shown in figure A4. The non-linear acoustic approach allows us to set-up probing
Performing nonlinear acoustic simulations for large transducers out to 15 meters (requested by the TPOC) requires substantial computer resources because the entire volume is discretized in the EFIT technique. The spatial resolution of the simulations had to be reduced in order to squeeze the simulation space onto the confined memory restraints of a standard desktop PC. This reduction in spatial resolution introduces a small uncertainty into the results shown in Tables B1-B4. The uncertainty in the focal distances is + 10cm and in the focal width is + 1 cm. Table B1: 0.6 Meter Diameter Transducer
Performing nonlinear acoustic simulations for large transducers out to 15 meters (requested by the TPOC) requires substantial computer resources because the entire volume is discretized in the EFIT technique. The spatial resolution of the simulations had to be reduced in order to squeeze the simulation space onto the confined memory restraints of a standard desktop PC. This reduction in spatial resolution introduces a small uncertainty into the results shown in Tables B1-B4. The uncertainty in the focal distances is + 10cm and in the focal width is + 1 cm. Table B1: 0.6 Meter Diameter Transducer
Figure B1: 50 kHz beam profiles for the four transducer diameters focused at 15 meters. For reference, the top plot shows the 0.6m (2ft) diameter transducer focused at 5 meters. The color represents the amplitude of the sound intensity. Effects of absorption are removed for this plot.
Figure B1: 50 kHz beam profiles for the four transducer diameters focused at 15 meters. For reference, the top plot shows the 0.6m (2ft) diameter transducer focused at 5 meters. The color represents the amplitude of the sound intensity. Effects of absorption are removed for this plot.
The technology that you presented at the meeting was a work —in Progress from your company Luna Innovations on an ultrasonic/acoustic approach to finding concealed weapons on a person. Your briefing stimulated discussions among the NTPAC members encouraged me as Chairman of NTPAC to prepare this letter in support of the development of a system such as the NAC for the corrections community. Based on the data we saw, this may be an ideal capability enhancing security in corrections facilities. For example, the bureau of Justice reports 14,165 attacks on prison staff by inmates, with 14 deaths and a 32% increase in five years. One problem that was discussed by NTPAC was the concern over improvised weapons made from plastic or metal and hidden under clothing. The ability of an inmate to hide a cell phone under clothing was also raised.
The technology that you presented at the meeting was a work —in Progress from your company Luna Innovations on an ultrasonic/acoustic approach to finding concealed weapons on a person. Your briefing stimulated discussions among the NTPAC members encouraged me as Chairman of NTPAC to prepare this letter in support of the development of a system such as the NAC for the corrections community. Based on the data we saw, this may be an ideal capability enhancing security in corrections facilities. For example, the bureau of Justice reports 14,165 attacks on prison staff by inmates, with 14 deaths and a 32% increase in five years. One problem that was discussed by NTPAC was the concern over improvised weapons made from plastic or metal and hidden under clothing. The ability of an inmate to hide a cell phone under clothing was also raised.
Related topics:
Signal ProcessingNonlinear AcousticsAcoustic WavesSpecular Reflectionimage resolution
580 California St., Suite 400
San Francisco, CA, 94104
© 2025 Academia. All rights reserved