Preliminary Interferometric Images of Moving Targets obtained using a Time- Modulated Ultra-Wide Band Through-Wall Penetration Radar

Transcription

1 Preliminary Interferometric Images of Moving Targets obtained using a Time- Modulated Ultra-Wide Band Through-Wall Penetration Radar Soumya Nag, Herbert Fluhler and Mark Barnes Time Domain Corporation 7057 Old Madison Pike Huntsville, AL USA Abstract-A time modulated ultra-wide band (TM-UWB) wall penetrating impulse radar currently being developed at Time Domain Corporation, Huntsville, AL, is used to generate two-dimensional (2D) Interferometric Images (11) of moving targets. The goal is to determine the location of moving people behind walls or non-metallic visually opaque boundaries in real time using such a hand-held wall penetrating radar operating under FCC Part 15, Class B limits. Interferometric Images of an adult male person walking and crawling on the floor are presented for different types of common walls in front of the radar. I. INTRODUCTION In forced entry and hostage situations, law enforcement officers need to know the location and movement of people and animals within buildings and behind barriers. In emergencies, search-and-rescue workers often need to know the location of victims and co-workers masked by opaque smoke obscurants and debris. For these and other related missions, a surveillance sensor is needed to improve situational awareness, decision-making and responsiveness Furthermore, the sensor must be portable, easy to use, covert and resistant to false alarms. Time Domain Corporation (TDC) has already developed a prototype hand-help impulse radar system (Radarvision 1000) with a single transmit and single receive antenna [l]. This system can locate a moving target located up to 6 meters away on the other side of a non-metallic wall. The Radarvision 1000 operates on the patented time-modulated ultra-wide band (TM-UWB) technology developed at TDC. This radar system provides response to a moving target only along the down-range direction. It does not provide any azimuth bearing information other than that which can be gleaned from its rather wide 90 degrees beam-width. An improved hand-held radar is currently being developed based on similar TM-UWB technology, but with a new transmitreceive antenna array which will provide real-time information of motion on the other side of a wall in both range and azimuth. The organization of this paper is as follows. Section I1 briefly describes the TM-UWB radar unit developed at TDC. Section I11 describes the signal processing algorithms leading to the two-dimensional (2D) radar image of a moving person. Images of a moving person obtained from the measured data are presented in Section IV. Finally, concluding remarks are presented. 11. DESCRIPTION OF THE TM-UWB RADAR UNlT A UWB radar has large relative bandwidth compared to conventional narrow-band radars. For carrier-free nonsinusoidal signals, the relative bandwidth is defined as [2]: q=- fh - fl (1) fh + fl For a UWB radar, I 1. The operational frequency band of the TDC TM-UWB radar is approximately 1.0 GHz to 3.0 GHz, which gives 77 = 0.5. UWB systems performance can be enhanced by time modulating the pulses. The short pulses collect a quasiimpulse response from the antennas and environment. Pulse position modulation varies the inter-pulse period of the transmitted short time duration pulses. If a uniformly spaced UWB pulse train were transmitted, the energy would be concentrated at discrete harmonic frequencies within the total bandwidth. These harmonics correspond to integer multiples of the pulse repetition frequency. By changing the timing of each pulse on a pseudo-random basis as illustrated in Fig. 1, the harmonics are eliminated. This produces a noise-like signal that appears to have little structure. In other words, the time modulation technique smoothes the energy uniformly under the envelope defined by the spectrum of the wave shape of the pulse, thereby eliminating harmonics. This provides covertness and improved electromagnetic compatibility with other narrow band and TM-UWB receivers. Pseudo-random time coding also forces the UWB receiver to sample any ambient radio frequency interference or jamming signal in a random manner. However the desired UWB signal is sampled coherently. When combined with an appropriate integrator this helps the TM-UWB radar to reject narrowband interference. An additional benefit of using this pseudo-random modulation scheme is that range ambiguities and clutter foldover can be significantly reduced in a manner analogous to the use of staggered pulse repetition intervals in Doppler radars. Radar returns from previous pulses are de-correlated when the down-sampling technique is employed /01/$10~ IEEE 64

2 Fus mncn I I "RLSRPICD nm Fig. 1. Time Encoding of the Pulses Vanes the Delay from the Periodic Position of a Uniform Pulse Repetition Rate. This scheme allows the radar to operate at the desired short ranges with an average pulse repetition frequency (PRF) of either 5 MHz or 10 MHz. The duration between pulses is pseudo-randomly varied between some set time limits using a pseudorandom code. All these features of the UWB radar conspire to provide the following advantages: (1) inherit covertness because of its ultra-low power noise-like signal, (2) high resolution at low radio frequencies for penetrating building materials, (3) reduced range ambiguities and clutter fold-over because of pseudo-random coding, (4) resistance to jamming, (5) resistance to false alarms, and (6) clutter rejection due to the short pulse duration [l]. A block diagram of the TDC TM-UWB Radar Array Testbed (RAT) is shown in Fig. 2. The system consists of a PulsON Application Demonstrator (PAD), multiple pairs of antennas with horizontal polarization and broad beam-width, an antenna switching circuit, and a desktop PC with Radar Testbed Application Software. The current RAT has the same hardware as that of a Radar Testbed (RT) [3], except that an antenna array assembly has been substituted for the single transmit-receive antenna pair used in the RT. Each antenna pair consists of one antenna element used for transmit and the other one used for receive. The antennas are designed to provide beam-width greater than 120 degrees and less than 180 degrees at -6 db power points of the twoway pattem in the azimuth plane. In the elevation plane these antennas are designed to provide a beam-width greater than 60 degrees and less than 90 degrees at -6 db power points of the two-way pattem. The PAD is a hardware platform developed at TDC to support demonstrations, technical investigations and fimctional application assessments based upon TM-UWB technology. A block diagram of the PAD is shown in Fig. 3. Engineer Ethernet / Multiple Tx-Rx Antcnna Pair Assembly with Antenna Switching Circuit Fig. 2. A block diagram of the Radar Array Testbed. Fig. 3. Detailed block diagram of the PulsON Application Demonstrator (PAD) Board The key function of the PAD is to precisely place the transmitted pulse in time and to precisely sample the received waveform relative to the transmitted one. For commercial applications, the TDC UWB radar must comply with the RF emission limits and characteristics specified under FCC Part 15, Class B limits. Under this FCC limit, the Effective Isotropic Radiated Power Spectral Density (EIR-PSD) of the system must be less than 0.4 nw/mhz for MHz, less than 2.4 nw/mhz for MHz and less than 30 nw/mhz for frequencies greater than 1000 MHz. Our goal in this current effort is to generate 2D Interferometric Images (11) of moving targets in the real time at a 3 Hz update rate using a hand-held TM-UWEi radar. The results presented in this paper are obtained from a simplified early prototype of the final system. This prototype at present does not generate images in real-time. Therefore the data are recorded and transferred to a desktop computer where Interferometric Images are generated using MATLAB SIGNAL PROCESSING The raw waveforms that are collected using each transmitreceive antenna pair contain stationary responses as a function of time history such as antenna coupling, cable ringing, noise, and response due to stationary ambient clutter. Before obtaining an Interferometric Image the stationary responses are reduced from the raw waveforms via a Motion Filter. This filter is an Infinite Impulse Response (IIR) bandpass filter and operates on the raw time domain waveforms as a function of time history for a specific time sample (i.e. range cell). The filter is easy to implement and, unlike ensemble average subtraction, does not rely on several time history waveforms. After filtering, the filtered waveforms contain only the response due to a moving target. In order to improve the signal-to-noise ratio (SNR) of the Interferometric Image of a moving target, the outputs of the Motion Filter are applied to a Range Filter before the 65

3 Interferometric Image is generated. The Range Filter is a band-pass filter applied to the Motion Filtered time domain waveforms as a function of down-range (range) for a fixed time history. This filter helps improve the SNR by reducing the clutter residue, noise and any radio frequency interference (RFI) when the data are collected using the TDC pseudorandom codes described earlier. The motion and range filtered waveforms are then used to generate an Interferometric Image via the time domain Backprojection algorithm [4]. If t denotes time (ns) and (U, 0) denotes an antenna location coordinate, then the backprojected signal amplitude at the image pixel (xi, &) is given by: f (xi, Yj 1 = C ~(tij (u),u) (2) U where, tij(u) = 2Rij(U)/Vx =q- (3) VX v, is the velocity of the electromagnetic fields in the ambient medium (air). t&) denotes the round-trip delay of the echoed signal for the target at the image scene coordinate (xi, yi) corresponding to the antenna location at (U, 0). Thus, to form the target function at a given grid point (xi, 3) in the spatial domain, the data at the down-range bins that corresponds to the location of that point for all antenna locations are coherently added. In other words, the coherent sum is performed via time-shifting the signal obtained at each antenna location U and then adding across all aperture locations to integrate the value at that image space pixel (xb yi). This time shift (align) and sum sequence-is repeated for all the image space pixels. Since the time axis is discretized in equation (2), a given distance may not coincide with the corresponding point on the time axis. Interpolation is typically required. In our case, the interpolation is approximated via the over-sampling of the waveforms digitized at the hardware level in the radar. Before displaying the resulting Interferometric Image it is envelope detected via the square-law amplitude demodulation technique in order to avoid grating-like ridges that would appear otherwise. These ridges are associated with the lobes in the target response and tend to make the target less discemable if they are not melded together [5]. IV. RESULTS Interferometric Images of a moving person are presented in this section. The data was collected using the RAT with and without a wall jn front of the antenna array. A few common wall samples were used: (1) a 0.09 meter thick red brick wall; (2) an interior building wall made of meter thick gypsum sheathing (sheet-rock) with metal studs (0.051 meter x meter) separated by 0.41 meter (center to center) and; (3) a meter thick sheet-rock wall with wood studs (0.051 meter x meter) separated by 0.41 meter, (center to center).., A top view of the trajectory of the moving person is shown in Fig. 4. The person was walking towards the antenna array and then walking backwards from the array at a speed of approximately 0.3 &sec (1Wsec). During the entire motion the person was facing the antenna array. All the radar images to be presented are generated from the data collected for the target trajectory shown in Fig. 4. The plot of time domain waveforms obtained after applying the Range Filter and the Motion Filter for several time histories ( # meas ) is shown in Fig. 5 for a specific transmitreceive antenna pair and with no wall present in front of the array. The waveforms are plotted in integrated A/D counts. Ih Fig. 5, the response of the moving target is shown for the # meas range and for the Down range -0.5 m to -4.5 m. The strong response corresponding to the range 0 to m corresponds to the antenna coupling clutter residue, which in this case is as strong as the target response. Relatively weaker response at -0.7 m and at -1.2 m correspond to the residue resulting from the responses due to the impedance mismatch between the antennas and the antenna switching circuit. Interferometric Images of a person walking and as well as crawling on the floor at a speed of 0.3 dsec are presented in Fig. 6 - Fig. 12. All these images are obtained operating the radar under the FCC Part 15, Class B limits. They are plotted as the square of the magnitude of the response with a fixed normalization factor. Also, in these images the response of the antenna coupling clutter residue are masked by zeros. DfrecUon 01 hpagatlon of waves Y (dowu-range) T Ddon of Motion I of the Tqct t \ X (cross-renpe) Fig. 4: A schematic diagram showing the top view of the moving target trajectory. 66

4 discontinuous. This is due to the fact that in addition to the direct waves illuminating and received from the target, there are also waves that are transmitted and received after being reflected by the metal studs. Note the cross-range ambiguity response is also discontinuous, unlike the responses observed in Fig. 6 and Fig. 7. In the case of the sheet-rock wall with wood studs, the image of the walking person (Fig. 9) is stronger and not distorted as in the case of the wall with the metal studs. Fig Fig. 12 show the Interferometric Images of a crawling person in the presence of the brick wall, the sheetrock wall with the metal studs, and the sheet-rock wall with the wood studs respectively. The person was crawling at the rate of approximately 0.3 dsec. The true location of the target is indicated in the caption of each figure. The Interferometric Images of the crawling person are weaker by 2-5 db compared to the corresponding cases when the person was walking. This is because of the fact that the radar cross section (RCS) of a person crawling with his head facing towards the antennas is smaller than that of the same person walking with his chest facing the antennas. V. CONCLUSION Fig. 5: The intensity plot of time domain waveforms as functions of range and time history for a specific transmit-receive antenna pair after applying the Range Filter and the Motion Filter. Fig. 6 - Fig. 9 shows Interferometric Images of a person walking at a velocity of approximately 0.3 dsec behind different types of walls. The true location of the target is indicated in the caption of each figure. In Fig. 6 (with no wall present), the response extending from (-3.2 m, 0 m) to (-3.2 m, 2.4 m) is due to the cross-range ambiguity ( ghost ) of the true target produced in an Interferometric Image for a given antenna element separation and at a specific down-range location. These ghosts are usually weaker in magnitude and more defocused than that of the true target response. Currently, investigations are being pursued to reduce such ghosts in the Interferometric Image. As shown in Fig. 7, the response of the moving person in the case of a brick wall is approximately 5 db weaker than that in Fig. 6 because of the attenuation of the electromagnetic fields through the brick wall. Otherwise, the image is barely defocused. A refocusing back-projection technique described by Cai [6] can be applied to Interferometric Images for the cases when a wall is much thicker or has a higher dielectric constant. The refocusing technique did not make a significant change to the Interferometric Image shown in Fig. 7. On the other hand, the response of the target in the sheetrock wall with metal studs shown in Fig. 8 appears to be Interferometric Images of a moving person through some common wall samples were presented using a TM-UWB radar operating under the FCC Part 15, Class B limits. The images of a person walking while facing the antenna array are stronger by approximately 2-5 db than the images of a person crawling with his head facing the array. This is because the RCS of a walking person is larger than that corresponding to the crawling person. The images of the moving person behind the sheet-rock wall with the metal studs were observed to be discontinuous due to the interference of the direct signals with the signals reflected from the studs. This effect is less pronounced in the case of the sheet-rock wall with wood studs. In the case of the brick wall, no distortion or defocusing of the radar images was observed other than the attenuation of the waves by approximately 5 db. We are currently investigating different techniques to help improve the SNR of the targets in the images, especially in the case of multiple moving persons with different RCS. [I] REFERENCES M. A. Bames, Covert Range Gated Wall Penetrating Motion Sensor Provides Benefits for Surveillance and Forced Entries, ONDCP Conference Paper, March [2] H. F. Harmouth, Nonsinusoidal Waves for Radar and Radio Communications, Acadamic Press, New York, [3] T. Payment, A Low Power, Ultra-Wideband Radar Testbed, EuroEM 2000, UWB Radar Session, Edinburgh, Scotland, May- June

5 [4] M. Souinekh, Svnrheiic Aperture Radar Signal Processing ny/h MATLAB Algorithms, John Wiley and Sons, Inc., H. Taub and D. Schilling, Principles of Communica/ion Systems, McCraw-Hill lntemational Series, L. Cai, Uti-a-Wide-Band Model Based Synthetic Aperture Radar Imaging through Complex Media, Ph.D. Dissertalion, The Ohio State University, Columbus, OH, Cross Range Imelcrl Cross Range In?uerl Fig. 8. Interferometric Image of a person at the coordinate (2.0 m, 2.4 in) walking with a velocity of approximately 0.3 sec (I Wsec). A sheet-rock wall with metal studs was present at m in front of the RAT antenna. I Fig. 6. Interferometric Image of a person at the coordinate (2.3 m, 2.7 m) walking with a velocity ofapproximately 0.3 sec (1 ft/sec). No wall was present in front of the antenna array Crass Range Imetw] Cross Range [meter] Fig. 9. Interferometric Image of a person at the coordinate (1.8 in, 2.4 m) walking with a velocity of approximately 0.3 ndsec (I ft/sec). A sheet-rock wall with wood studs was present at in in front ofthe antenna array I ? Fig. 7. Interferometric Image of a person at the coordinate ( I.8 m, 2.5 in) walking with a velocity of approximately 0.3 m/sec (1 ftlsec). A brick wall was present at in in front of the RAT antenna elements. 68

6 " Cross Range [meter[ Crosr Range [meter! Fig. 10. Interferometric Image of a person at the coordinate (1.8 m, 2.4 m) crawling with a speed of approximately 0.3 m/sec (1 Wsec). A brick wall was present at m in front of the RAT antenna elements. Fig. 12. Interferometric Image of a person at the coordinate (1.7 m, 2.3 m) nawling with a velocity ofapproximately 0.3 m/sec (1 ftisec). A sheet-rock wall with wood studs was present at m in front of the RAT antenna elements Cross Range ImHerl Fig Interferometric Image of a person at the coordinate (I.5 in, 2.0 in) crawling with a velocity ofapproxiiiiately 0.3 idsec (I ft/sec). A sheet-rock wall with metal studs was present at 0.0 I3 m in front f the antenna array 69