Patent Application
20050099331
This description relates to detecting concealed objects, such as objects concealed on a person who is entering a secured area.
One common approach to detecting concealed objects on a person is to have the person pass through a metal detector.
Techniques are described for using advanced algorithms developed for buried mine detection to yield a very high probability of detection against concealed objects that can be dangerous or undesirable, coupled with a low false alarm rate. In particular, unique, shallow buried plastic target detection technology that employs radar signals may be applied to aviation security to counteract potential terrorist acts involving smuggling of explosives and other objects on the body.
In one implementation, a concealed object detection system for detecting objects concealed on a person includes radar transponders that are each configured and positioned to direct a radar signal at a person and to detect a portion of the radar signal reflected by the person. A processor connected to the radar transponders processes the portions of the radar signals detected by the radar transponders to determine whether the person is carrying a concealed object. The system may produce a real-time alert, such as an audible alert, when a concealed object is detected.
Other features and advantages will be apparent from the following description, including the drawings, and the claims.
Referring to
The current, advanced algorithms for the detection of buried land mines yield excellent detection (˜100%) and very low false alarm rates (˜0.1%). This result is achieved over widely varying material types and conditions, and may even be used to detect small, plastic-encased, buried antipersonnel mines. For example, the algorithms may be used to detect small plastic mines that are six centimeters in diameter and nearly one centimeter thick, which typically are buried from flush to the surface to five centimeters below the ground. Larger plastic mines (greater than ten centimeters in diameter) often are buried up to fifteen centimeters deep.
Concealed object detection is technically similar to mine detection. The radar frequency (RF) signal is transmitted from the antenna and the reflected signal energy is detected and processed. The processing methodology determines when the reflected signal appears to be coming from a particular material to be detected (e.g., explosives) and rejects all other reflections. In the case of mines, the algorithms have been very successful in the detection of explosives contained within a plastic housing. In the case of concealed objects on a person, such as explosives that have a form factor of one centimeter of thickness over an extended area of several centimeters in length and width, the detection methodology is similar. The algorithms ‘see’ the signal reflected from the person in the same way that the mine detection algorithms ‘see’ the signal reflected from the ground. When the explosives are present, the reflected signal will change. It is this change that is recognized by the algorithms and declared as a detection of the explosive material.
One implementation uses only three detectors 110, each of which is similar to a mine detection device. Measurements are made over a signal bandwidth from 1.5 GHz to 10 GHz. This band is wider and of slightly higher frequency than is normally used for buried mine detection. Because clothing transmits higher frequencies than several inches of soil, investigation of this wider band is desirable to support initial optimization of the parameters for the advanced algorithms. RF energy density is less than one hundredth of standard limits. The key factor determining performance is the application of the advanced algorithms to the measured data.
The reflection from explosives differs from that of the body based on the material dielectric properties. The signal passes through the explosives, reflects off the body, and passes back through the explosives. As shown, the portion of the signal that encounters the explosive is attenuated and time delayed relative to the reflection from the body alone.
RF energy interacts with the body depending on frequency: the body absorbs from 30 to 300 MHz, partially absorbs from 300 MHz to 6 GHz, partially or diffusely reflects from 6 GHz to 15 GHz, and highly/specularly reflects at >15 GHz. Previous attempts using frequencies in the highly-reflective region Ka-band (27 to 33 GHz) produced specular reflection images, which made processing to detect the concealed objects difficult. Ku-band (12 to 18 GHz) studies produced slightly better results but straddled the transition from diffuse to specular reflection at 15 GHz.
The techniques may employ the absorption/reflection transition frequency region from 1.5 to 10 GHz. In this region, specific frequencies that maintain high dielectric differences between explosives and the body surface coexist with diffuse reflection from the skin and shallow body depths. This maintains radar cross section or reflection of the target without spoofing the detection process through specular reflection above 12 GHz and without driving the radar cross section to unacceptably low levels below 1 GHz.
In the region from 1.5 to 10 GHz, the explosives dielectric is near 3, while the body covers ranges of 5 for fat, 10 for bone, 40 for skin, and 50 for muscle. This compares well with the mine detection problem, where the explosives dielectric near 3 was embedded in dielectrics between 4 for sand and 20 for wet loam. The clothing dielectric is near 1.1 (with air at 1.0) and therefore provides no significant reflection or effect on the RF energy. The higher dielectric is behind the target, and there is typically much greater difference between the explosive dielectric and the body (for example, explosive to sand is a difference of 1; explosive to skin is a difference of 37).
Mine detection equipment transmits 10 milliwatts and yet enough of a signal penetrates soils that have high attenuation characteristics to achieve excellent performance. Measurements will be made to determine the minimum possible transmitted signal strength to achieve scanner performance through clothing.
Allowable general population/uncontrolled exposure sets the toughest limit on exposure to RF radiation. At the frequencies of interests, the allowable density is 1 mw/cm2 (FCC) for 30 minutes averaging or about ⅕ to {fraction (1/10)} of controlled exposure limits (IEEE and OSHA) for 6 minutes averaging time. At the surface of the transmitter, a density of 0.1 mW/cm2 is anticipated. Exposure through the portal is for less than 10 seconds or about {fraction (1/200)} the allowed time period; this combined with the {fraction (1/10)} lower maximum energy density yields a very safe {fraction (1/2000)} exposure margin. (Note: for cell phones 600 mW is allowed; one could place his head against a transmitter at maximum power of 10 mW and still be {fraction (1/60)} of this criteria as well.) The proposed portal system is also well below the specific absorption rate (SAR) limit of 0.4 W/kg. Even if somehow an individual absorbed the RF energy from all 40 transmitters employed in one implementation of a portal-based detection system simultaneously, which is physically impossible, the total would be 0.008 W/kg for a typical 50 Kg individual. This is 50 times less than required; typical exposure to the two corner set of 20 transmitters would be 100 times less.
The walk-through portal scanner of
The vertical array of beams scans horizontally and vertically. The array of beams scan only a small vertical angle. Together with the horizontal scanning of the array, the surface reflectivity of the person for the left and right sides and the front and back is collected. A flat two-dimensional display is then created of the reflectivity measurements.
The RF models that have been developed, updated, and correlated with mine radar background and target characteristics, are extended to the aviation security scenarios through use of the described radar system. The models include target cross-section, SNR, and algorithm receiver operating characteristics (ROC) curves.
Radar data is collected against targets (recent examples being the buried mines) and segregated into algorithm training and testing groups. Typically, enough data is collected to separate the groups wholly (to demonstrate the required Pd and FAR at 90% confidence, for example). Alternatively, so-called ‘leave-one-out’ testing may be performed to achieve nearly the same statistical significance with half the data. Training and testing on the same data (Train A/Test A) only provides a lower bound on the Bayes error for the population distribution. Train A/Test B (or equivalently, leave-one-out training/testing) provides the upper bound on the Bayes error that is practically a more useful predictor of system performance.
System performance may be defined in view of Pd versus FAR. Both Pd and FAR are evaluated together, preferably using ROC curve methodology including confidence intervals. There are two measures of FAR, per individual and per group. For the former case, using 1.75 m2 for surface area, if the system performance is similar to the mine detection performance, then scaling this performance along the ROC curve yields a 2% FAR per individual at the required 90% Pd. Only 1 in 50 individuals would have to be rescanned using a handheld device, assuming individual and group statistics are similar. An example of the tradeoff between Pd and individual FAR is shown in
For the aviation security problem, significantly better performance (near 100% Pd and near 0 FAR) is expected because of the relatively controlled/benign environmental factors involved, versus the unconstrained, outdoor, small mine detection problem. Specifically, the mine detection technology was designed and successfully tested against three types of backgrounds: gravel, grass, and bare soil. These terrains included, for example, buried roots and rocks. The targets included a 6 cm circular plastic encased explosive that is significantly smaller than a baseline 10 cm by 7.6 cm ‘wallet’ target.
Use of radar technology permits the system to provide real-time alerts. Typically, a real-time alert is in the form of an audio signal. In some implementations, a two-dimensional offline image is provided to indicate the on-body position of detected objects. For example, icons representing detected objects may be presented on a notional/universal wire-frame body to indicate the location and size of the detected objects. Typically, no actual imagery of either the object itself or the subject body will be created or displayed. The two-dimensional plot includes icon and text overlays that indicate the object position and size with absolutely no operator training or interpretation required.
Detection of concealed metallic or plastic objects poses a significant challenge for the radar hardware. Current short pulse or video penetrating radar systems are sensitivity limited by problems with extraneous signal interference competing with the desired signal. The result is an unrecognizable signal, and even after processing and display, the data is unusable. The signal interference arises in most cases due to mismatches in components internal to the radar system, not from external sources. Some of the specifics are: (1) antenna mismatch due to bandwidth limitations and dispersion; (2) transmitter waveform sidelobes; (3) multiple transmitter/receiver/antenna reflections due to mismatch impedances; and (4) non-uniform bandwidth, no absolute control of transmitter. The described frequency stepped radar system design eliminates the problems of the current ‘short-pulse’ radar systems to allow for an increase in measurement capability. Bandwidth control improves detection of small, non-metallic objects and plastic objects with small cross sections, and a reduction in interference improves sensitivity and signal clarity.
The frequency-stepped radar permits operation at an RF duty factor approaching unity, to remove the short-pulse radar requirement that the RF equipment (transmitter, antenna, and receiver) be instantaneously broadband. It also achieves a fully coherent radar capability while retaining (and expanding) the achievable high-range resolution capability. The significance of the high RF duty factor is that the thermal-noise-limited detection sensitivity of the radar can be achieved using readily available components. In fact, with the high-RF, duty-factor-stepped frequency waveform, the sensitivity limit is dictated by parameters related to the environment in which objects are located. The frequency-stepped radar is fully coherent which allows for compensation of hardware amplitude and phase (dispersion) errors over the operating RF band.
A bi-static system (separate transmit and receive antennas) is used with appropriate calibration and signal processing. The critical technology advantage for the described system proposed lies in the bandwidth control, the resultant sensitivity increase and the antenna match that can be achieved given the frequency bandwidth that must be covered and the various conditions through which the antenna must propagate energy. The overall radar specifications for one implementation are provided below in Table 1.
The system employs a derivative, penetrating radar system based on mine detection hardware and software. To the extent available, off-the-shelf electronic components are used to extend/shift current frequency-stepped RF module performance from 1 to 3 GHz to cover the range from 1.5 to 10 GHz.
The spiral antenna is constructed by etching the pattern on a printed circuit board. A planar, printed circuit, spiral antenna radiates perpendicularly to the plane of the spiral. The spiral itself is located at the end of a cylindrical metal cavity (the cavity back) to provide isolation from neighboring elements and electronics. Typically, an absorber is used on the top-side of the spiral inside the cavity to make sure the element responds only downward. Separate transmit and receive antennas are used to simplify the electronics, provide spatial separation and reduce very shallow reflections.
Suitable processing hardware includes a PowerPC 750 processor running at 350 MHz on a PMC mezzanine form factor. The PowerPC 750 is ideally suited for portable applications due to the small die size and low power consumption of less than 4 Watts and its low cost.
The described systems and techniques promise to provide low cost, ease of use, and higher performance (better detection and lower false alarm rate). A low-cost, user-friendly radar system for difficult concealed object detection employs a comprehensive antenna and RF module design, sophisticated concealed/buried target detection algorithms. Data collected against different targets and in different conditions, together with adaptive/advanced algorithms, will enable prediction and (through testing), validation of performance for a range of potential operating frequencies.
This application is a continuation, and claims the benefit of priority under 35 U.S.C. 120, of U.S. application Ser. No. 10/241,835, filed Sep. 12, 2002 and titled: Concealed Object Detection, which itself claims priority under 35 U.S.C. 119(e) to provisional application No. 60/318,327, filed Sep. 12, 2001, and titled: Concealed Object Detection, both of which are incorporated by reference.
| Number | Date | Country | |
|---|---|---|---|
| 60318327 | Sep 2001 | US |
| Number | Date | Country | |
|---|---|---|---|
| Parent | 10241835 | Sep 2002 | US |
| Child | 11009135 | Dec 2004 | US |
