Technical Papers

» Buried fiber intrusion detection sensor with minimal false alarm rates

Jeff Bush, Carol Davis, Pepe Davis, Allen Cekorich, Fred McNair
Optiphase, Inc.
7652 Haskell Ave., Van Nuys, CA 91406

Abstract

A novel design approach for a highly reliable buried intrusion detection sensor is described. The design involves the use of a low cost depolarized Sagnac fiber interferometer with a "sensing loop" consisting of a delay line and buried fiber segment. The intrusion sensor is configured for an "all fiber" remote deployment where active components (source, receiver, demodulator) are located separately and connected to the sensor through an insensitive fiber tether. A robust and cost effective buried sensor "mat" design was developed. This design enabled high sensitivity as well as ease of deployment. Sensors were built and evaluated. Test results indicate an effective design.

Keywords: Intrusion Detection, Fiber Sensor, Sagnac, Interferometric Demodulator

1. MOTIVATION

The intrusion detection systems of interest are those which reliably detect humans with a high probability of detection and a low probability of false (or nuisance) alarms. Most types of current-day intrusion detection systems are based on designs implementing some type of non-linear measurement process. Examples of the most common types of these systems include:

• Beam-Break Sensors                   "electric eye" and microwave sensors;

• Triboelectric Sensors                    voltage produced from physical disturbances of coax cable;

• Motion Sensors                            Infrared type sensors

• Electric Field Sensors                   Intruder changes electric field

• Fiber microbend Sensors               Intruder causes bend loss change in fiber

• Speckle Sensors                          Intruder causes mode interference changes in multi mode fiber

The above types of sensing approaches do not actually measure physical or environmental changes caused by the intruder, but rather cause some type of signal change resulting from the intruder. Such signal changes are typically highly non-linear and do not offer a high degree of discrimination between different types of evens, such as rolling tumbleweed, animals, or precipitation versus an intruder in motion. There is a significant trend in the intrusion detection field to be able to reduce the false and nuisance alarm rates. Most practitioners in the field agree that this can be accomplished through enhanced processing of the information provided by the intrusion detection sensors or to implement (linear) sensors which are capable of accurate measurements, or both. The work presented in this paper outlines the design and test results of a buried fiber optic Sagnac sensor which offers the following features:

• Provides a large linear dynamic measurement range of the load variations caused by an intrusions directly over the buried fiber;

• Provides accurate measurements independent of optical intensity variations;

• Sensitivity may be varied by changing the length of fiber used in the delay line (without having to replace the buried sensor).

• Is amenable to post signal processing to reduce false and nuisance alarm rates

• Is manufactured with low cost components.

2. TARGET APPLICATIONS, HIGH SECURITY FACILITIES

The work performed was aimed at high security facilities, where redundant intrusion detection sensors are utilized, as shown in Figure 1.

Such facilities are typically double fenced with a gravel bed between both fences. Intrusion sensors and cameras are placed on the fences. Microwave sensors are placed above the gravel and buried sensors are placed under the gravel.

A typical "zone" is 100 yards long, and a typical facility may employ 20 or so zones.

With the sensor count being high for such facilities, it is imperative that the false and nuisance alarm rates be kept to a minimum. The objective for current day intrusion sensors being deployed in high security facilities is to keep false and nuisance alarm rates to 1 per month while maintaining a 100% probability of detection of human intruders.

The authors feel that although the sensor concept embodied in this paper is well adapted for the high security facility, its ability to be produced for a low cost also gives it merit for general purpose buried use including any outdoor area or sub-floor indoor area.

3. SAGNAC SENSOR CONCEPT

In Figure 2. (left side) we show the traditional block diagram for the fiber optic gyro, which is sensitive to "nonreciprocal optical phase variations, such as rotation rates. If the gyro design configuration is altered somewhat, such as that on the right side of figure 2, we show the ability to detect perturbations to the exposed fiber (in the sensor loop). If the length of the sensor loop is short (less than 25%) compared to the delay coil, the sensitivity to perturbations at any point on the sensor loop is relatively constant.

The perturbations to the fiber, if buried will be those experienced by an intruder causing the fiber to be "loaded" by the intruders weight, which in turn causes a length variation of the fiber. Since we are using a Sagnac interferometer, the light counter propagates through the same sensor fiber (and delay coil). An interferometric signal (or non-reciprocal optical phase shift) is produced because the light traveling clockwise experiences the perturbation at a different time than does the light traveling counterclockwise. Given the buried fiber sensor is section is placed on one side of the sensing loop, and the delay coil is at least twice the fiber length as the buried fiber section, an approximation of the sensitivity of the Sagnac sensor is given as.

Where K = lumped constant, taking into account the wavelength and fiber index
t = the delay time associated with the delay coil
Dx(t) = the optical path change caused by the "intrusion" event

From equation 1, it is obvious that the Sagnac interferometer is only sensitive to dynamic variations caused by intrusion events. The sensor is essentially oblivious to any static (or very low frequency) variations, such as temperature changes or increased weight on sensing region due to snow, rain, erosion, or other such changes.

One question which may be asked concerning the Sagnac sensor relates to its inherent ability to sense disturbances from an intruder. Especially if the intruder is smart and plans to move very slowly over the target detection area.

It turns out, that even if intruder motion is very slow, the intrinsic high sensitivity of the interferometer is highly sufficient to detect the stealthiest intruders. In figure 3, we show the predicted measurement resolution for a Sagnac sensor with a 100 meter sensing length and a 1 Km delay coil. Here we are using a 1.3 um source and a measurement detection bandwidth of 50 Hz. At frequencies as low as 0.1 Hz, the measurement threshold is 1 um.

If it was determined that the measurement threshold needed to be reduced, the sensitivity could be increased by simply replacing the delay coil with one with a longer fiber length.

4. INTRUSION SENSOR DETAIL DESIGN

4.1 TOP LEVEL DESIGN

The details of the optical design for the intrusion sensor are shown in figure 4. This design is based on a depolarized Sagnac configuration and is thus made primarily of low-cost single-mode fiber. The design is also arranged so that the majority of the active components can be located separately from the sensor section.

The "indoor" section includes the broad band source (1.3 um ELED), a SMF 2X2 (–3 dB) coupler, an optical receiver and trans-impedance amplifier, and an Optiphase manufactured Digital Demodulator. The "outdoor"" section includes the sensor coupler, the delay coil, a fiber wrapped PZT modulator a Lyot type fiber depolarizer and the intrusion sensor. The configuration in figure 4 runs a single fiber for the sensor and a wire pair or coax cable for the modulator drive.

In order to address concerns about the piezoelectric modulator drive regarding EMI, Optiphase devised a novel optical drive for the modulator, which is shown in figure 5. Here we drove a surface emitting diode with the modulation drive signal. This signal was sent via multi-mode fiber to the remote sensor, where the light was converted to an electronic signal and then used to drive the piezoelectric modulator. This enabled an "all-optical" remote sensor.

Also shown in figure 5 are some more detail as to how the sensor assembly was actually optically interfaced and packaged. A dual fiber cable was used to connect the Electro Optical Assembly to the remote sensor.

4.2 REMOTE ASSEMBLY AND TEST MAT DETAILS

The box marked Prairie Dog (in figure 5) is a water tight housing which contains all of the sensor support components. This assembly actually constructed is shown in figure 6.

The bottom assembly shown in figure 6 fits inside the housing. Once inside, a strap is placed around the housing (at the input end), which assures a water tight fit. The fiber cable feed through shown in the lower right portion of figure 6 is potted.

The deployment scenario for the remote assembly is shown in figure 7. Here we used an external housing for connectors which terminated the cable from the Electro Optics Module. This external housing is also water tight and buried in close proximity to the test mat.

The design for the test mat is shown in figure 8. This design shows the fiber spacing varied between 4 inches to 18 inches. This was done so that we could determine (through field tests) what the optimum spacing should be for true deployment applications. We found that the 18 inch spacing was fairly effective but the sensitivity to intrusions did vary depending on where the intruder was located on the mat (on or off the fiber). We concluded that a 6 to 8 inch spacing eliminates such effects.

The test mat was constructed by coating the Mylar with a pressure sensitive adhesive, then placing the fiber down. The second step was to laminate the Mylar to the foam, and the third step was to punch holes in the finished assembly. The holes serve to provide drainage when the sensor field experiences rain.

The mat fabrication was performed at a commercial lamination facility. Based on the processes and material used, we estimate initial manufacturing pricing to be approximately $1.50 per square foot, which would drop to lower levels when production techniques mature. This type of cost is similar to existing conventional buried sensor technology (coaxial electric field sensing).

4.3. ELECTRO-OPTIC PROCESSOR

The Electro-Optic assembly, as seen in figure 5 contains the following key components:

1. ELED @ 1.3 um Pigtailed to Single Mode Fiber;

2. Fused Coupler, 2X2;

3. SLED and Driver;

4. InGaAs PIN (photodiode) and Trans-Impedance Amplifier;

5. Interferometric Demodulator.

Items 1-4 are rather common low cost devices which are offered by many vendors serving the telecommunications industries.

The last item, the interferometric demodulator, is not a common device. Optiphase, Inc. has developed an all digital interferometric demodulator for the purpose of performing high resolution, high accuracy measurements of interferometric signals.

Optiphase’s COTS demodulator board is shown in figure 9. This is an imbedded DSP based design. Key features of this demodulator are described below.

» Full Featured Demodulator

• Provides Sinusoidal Modulation Signal, up to 120 KHz;

• Incorporates Servos to maintain constant modulation level and proper signal phasing;

• Provides both Analog and Digital Outputs

» Measurement Characteristics

• Signal Measurement Range ± 6 urad to ± 12,000 rads (32 bit digital)

• Analog Output scalable to any 20 bit segment of the 32 bit Signal Measurement Range

• Measurement Linearity, 0.1% typical

• Self Noise < 3 urad/rt-Hz

• Interferometric measurement independent of Optical Intensity Fluctuations.

The important features with the use of this type of demodulator for the intrusion detection application are as follows.

• It provides accurate (linear) measurements of load variations caused by intrusion events;

• It never needs adjustments after set-up (operators do not need to be skilled);

• It is DSP based and has the capability to process intrusion signals after they are measured using pattern recognition algorithms to further minimize false and nuisance alarm rates.

5. TEST FIELD INSTALLATION

Optiphase fabricated four separate sensor systems like that shown in figure 5. These sensors were deployed for testing at Sandia National Laboratories, exterior intrusion detection facility. The four Remote Sensor Assemblies and eight test mats were deployed as follows.

• Remote Sensor Assemblies buried approximately 1 foot down (prairie dog vertical);

• Two Test mats were strung together to make one sensor;

• Three sensors were buried under gravel (two at 3 inches deep and one at 6 inches deep);

• One sensor was buried under dirt, three inches deep.

The deployment of the sensors was also a test of their ability to withstand deployment using typical deployment tools (i.e. a small "Bobcat" tractor). Figures 10 and 11 show the two methods of burial shown. In figure 11, the tractor is driven directly on the sensor mat. In figure 10, the tractor places rocks on the mat before driving over it.

Figure 10. Sensor Burial (right)

All eight test mats were buried using the tractor on the mats without incident. The approximate weight of the tractor with a gravel load is about 5000 pounds.

The four Electro-Optic Processors and data logger were operated from a small shed located close to the sensor field. The fiber tethers to each of the sensor were buried. In figure 12, we show the four Electro-Optic Processors, packaged inside a single rack mount chassis.

The four demodulation cards are standing upright (front middle). The ELEDs and receivers are located mid-left (receivers on top), the couplers and splices are located in splice trays (left) and the SLED and associated drivers are located front right. The ELEDs used were in coaxial packages and driven in constant current mode (no TEC). Over the course of testing performed the external chassis temperature varied from 25 F to over 100F. The optical output power fluctuations of the ELEDs were about 4X, which had no adverse effects on the demodulators ability to perform accurate optical phase measurements.

The interfaces to the four channel Electro Optic system are also shown in figure 12.

Each channel has panel indicators for the ADC and modulation level status, a coax connector to monitor the analog signal for each channel and two fiber connectors for each channel. FC for the sensor and ST for the optical modulation signal.

The 36 pin "D" connector front (lower right) is the four channel interface to the data collection apparatus.

Figure 12. Four Channel Electro-Optic system

6. TEST DATA

Testing of the deployed systems was initiated in August 1997 and is still ongoing. The key issues of interest for the testing are itemized as follows:

• Test mat sensitivity and optimal fiber spacing on the mat;

• Sensor System survivability and operability throughout extreme weather conditions;

• Demonstration of minimum false alarm counts (ability to discern different types of signals);

• The ability to distinguish features of intrusion events or to separate actual alarms from nuisance alarms.

The signal characteristic of a person walking over the test field is shown in figure 13.

The data shown in figure 13 was taken under normal walking conditions. Bear in mind that the sensors are producing signals which vary with the time derivative of the load variations on the mats. For both plots the data bandwidth is about 50 Hz. As expected, the signals on gravel show a higher frequency content during intruder generated load shifting.

Figure 14 shows the signal patterns for a variety of input signals. This data is all derived from sensing under gravel.

A brief description of the data in figure 14 is as follows.

Baseline           This shows the noise floor of the system when there is no activity on the test field.

Walk on            Normal walking directly over sensor (identical to right side of figure 13).

Walk beside       Normal walking directly beside sensor, 2 feet away from sensor surface.

Stealthy Walk    An attempt to sneak over the sensor field. A very slow walk trying to keep load variations to a minimum.  Walk rate is approximately 0.5 to 1 foot per second.

Crawl          Intruder with most of the body in contact with the ground. Crawl rate less than 0.5 feet per second.

Shift Weight  Standing Still on the sensor and slowly shifting weight from one foot to the other at a rate of one shift every 5 seconds.

Heavy Rain     Rainfall rate was approximately 0.5 inches per hour.

Explosion      A bomb squad training exercise was nearby (approximately 2 miles away). We aren’t sure if this data is from air acoustics or ground shock.

From the data displayed, it seems apparent that the sensitivity required to detect almost any type of event is readily achievable. Also the signal features associated with the events are readily discernable. We intentionally show the data reduced to a 50 Hz bandwidth (it was originally sampled at rates approaching 1 KHz). Given the low bandwidth and the large signal diversity for the different events recorded we believe that it will be relatively easy to implement signal processing algorithms in real time which will be able to separate out the types of events shown.

There a few types of events not shown in figure 14, which we believe to also be important. These include actual intruders that would be considered as nuisance types. These include animals which weigh less than humans. We plan to conduct such tests in the near future.

7. CONCLUSIONS

We have demonstrated a buried intrusion detection approach which is capable of high-resolution linear measurements. The device provides analog and digital signal outputs which provide a high degree of discernable characteristics to separate actual intrusion events from those considered nuisance events.

The optical sensing technology utilized in this effort is one which can be manufactured at low cost. Given present pricing for conventional buried intrusion detection equipment, we believe this new optical approach to be cost competitive.

Further evolutions of the design are planned which will invoke signal pattern recognition algorithms such that only human intrusions may be separated out from other (nuisance) intrusion signals. Once done, the capability of very low false and nuisance alarm rates combined with 100% detection should be achievable, and at an affordable cost.

ACKNOWLEDGMENTS

A significant portion of funding for the development of the buried intrusion detection sensor was funded by the U. S. Department of Energy under a SBIR Phase II Grant (DE-FG03-94ER81675).

The authors would also like to acknowledge the following individuals.

Mirko Ivancevic and Bill Klimowych of Optiphase, Inc. for their dedicated efforts in the fabrication of the optoelectronic assemblies utilized in this work.

Don LaCommare, an independent consultant for his efforts in design and fabrication of the sensing mats.

Larry D. Miller, Exterior Sensor Project Leader, Intrusion Detection Technology Department, Sandia National Laboratories, Albuquerque, NM., who arranged all affairs for deployment and monitoring of sensor systems at Sandia’s Exterior Intrusion Detection test field.