#

MINIATURE SENSORS FOR DETECTION OF HUMANS IN CARGO CONTAINERS

MINIATURE SENSORS FOR DETECTION OF HUMANS IN CARGO CONTAINERS

The objective of this project is to develop a miniature, inexpensive, low power, wireless, false alarm-free human occupancy detectors based on chip-integrated multi-functional optoelectronic sensors combined with intelligent control, acquisition, analysis, and warning systems. This project should demonstrate the feasibility of a portable laboratory prototype device based on discrete solid-state optoelectronic components, such as Light Emitting Diodes (LED) and Photo Detectors (PD) and a computer operated electronic control, acquisition and analysis system that can be characterized in field-simulated conditions. A pre-commercial chip-integrated wireless device is planned to be tested in field conditions.

The detection of humans hiding in cargo containers is based on real-time simultaneous Non-Dispersive Infra-Red  (NDIR) absorption measurements of carbon dioxide (CO2) resulting from human breathing combined with intelligent comparison of its levels and oscillation frequency. The unique features and benefits provided by the proposed technology are shown in the table below:

The sheer size and number of containers arriving at airports and seaports around the world make them a likely vector of terrorist attacks. Possible terrorist attack threats can be humans hiding inside the containers. Stowaways often bring battery-operated drills to make air holes in the containers. Even tiny holes can be detected, however, because the containers are stacked and restacked before being placed on the ship, the holes can be anywhere on the container, which result in problems finding them. Looking inside each of the 6 million containers, coming into the U.S., from abroad would disrupt the flow of goods. Instead of inspecting every container upon arrival, sophisticated computer and intelligence systems are being established to identify suspicious containers before they leave foreign ports. Drive-through mobile X-ray units that can scan containers as they are driven past a checkpoint, much like luggage through an airport screening station, are sometimes detecting people standing inside a container. In most cases they are illegal immigrants, but there is a high possibility of terrorists transfer by similar means. The X-ray machines used for random screening are costly, expensive, and not infallible.

Better surveillance at the container’s point of origin is required in most cases. Early (within few hours from loading time) detection of hidden humans in cargo containers is required in order to prevent possible terrorist actions. Other screening devices are being tested and deployed. Radiation detectors installed on cranes in some ports screen each container as it is offloaded. As of now, Customs agents use pager-sized radiation monitors that warn of excessive radiation as they walk by rows of containers. Some estimates put the cost of equipping all major ports with large scanners at $5 billion.

Customs, together with Immigration and Naturalization Services and several intelligence agencies have to carefully sort out huge amounts of information about containers before they arrive in a centralized database on the movement of cargo ships. Inspection of the cargo containers before loading onto the ship proved to be impractical, because the security staff had a limited time frame in which to carry out the inspection. Because the containers are lifted onto the ship by cranes, there was a limited amount of time that a security officer could safely stand around or underneath the container to inspect it. The practice is also time-consuming. Visual inspection of every container would make a ship never leave the port.

A portable Human Occupancy Detector CD-2, based on carbon dioxide sensing, manufactured by Armstrong Monitoring in Ontario, Canada, has recently been developed but is yet to be in wide use by the shipping industry. It costs $2700, has a digital display of concentration in ppm, internal sampling pump, audible and visual alarm, 12-hour rechargeable battery, an optional 0-1 VDC recorder output, and wide assortment of accessories and probes for different applications. The product was being used extensively in Canadian prisons to detect whether prisoners were attempting to escape in delivery trucks. The hand-held unit is battery operated and is attached to a small probe that is inserted between the rubber seal and the container.

The unit measures the carbon dioxide in the container and provides a digital reading on the face of the unit. Security officers use the device to measure the ambient carbon dioxide level outside the container followed by that inside the container. When the detectors were put into practice, security learned that human beings are not the only producers of high carbon dioxide levels. Other items, such as tobacco and scrap-metal, also exude carbon dioxide. While these items release carbon dioxide at different levels and can be possibly distinguished from levels released by human beings, there are still needs for security personnel involvement (to compare the levels and make decisions on the inspection) and possibilities of false alarms. In addition, presence of small animals, such as rats and mice can cause false alarms as well.

While the idea of carbon dioxide level measurements looks attractive and feasible, there are still several issues associated with distinguishing the carbon dioxide origin, high cost, and high possibility of false alarms. The proposed here technology can solve the problems associated with such issues by combining an inexpensive and reliable technique for carbon dioxide sensing with features based on intelligent signal analysis.  

The detection of humans hiding in cargo containers can be based on NDIR absorbance measurements of two human breathing products: carbon dioxide and water vapor and time-resolved comparison and analysis of their levels.

The feasibility of sensing carbon dioxide with a compact inexpensive and highly reliable MID-IR system has already been proven
[1]. The measurements are based on the Lambert-Beer law and current stabilization of the IR source. Employment of a high-quality IR detector, and data compensation for changes in the ambient temperature and pressure, are used in the sensor to measure carbon dioxide with high precision and efficiency.  Diode laser-based detection of water vapor provides a combination of high sensitivity, rapid time response, and wide dynamic range that is superior to most other methods of humidity measurement. Water vapors have been successfully measured by using miniature and reliable LEDs as well
[2].  Carbon dioxide has a strong absorption at 4.26 mm and water has very strong absorption band in the range of 2.55-2.75 mm and second, strong enough absorption band that can be used for measuring in the range of 1.83-1.9 mm4.  Detection of these compounds by optical means is possible with sensitivities of 0.13 ppb for CO2, and 2-60 ppb (depending on the wavelength) for H2O
[3]. The concentration of carbon dioxide in an exhaled breath varies with time: the first portion contains no carbon dioxide and comes from the upper respiratory tract where no gas exchange takes place (the anatomical dead space – 2mls/kg). The concentration of carbon dioxide then rises rapidly to a plateau of about 5% as alveolar gas is breathed out.  The term re-breathing implies that expired alveolar gas containing 5% carbon dioxide (and less oxygen than normal) is inspired as part of the next tidal volume. Table 1 shows concentrations of oxygen, carbon dioxide, and water in the inhaled and exhaled portions.

[1] Guangjun Zhang, Junfang Lui, and Mei Yuan. Novel carbon dioxide gas sensor based on infrared absorption. Optical Engineering, 39(8), 2235-2240 (2000).

[2] Hodgkinson, J., Johnson, M. and Dakin, J. P.. Photothermal detection of trace optical absorption in water using visible light emitting diodes.  Applied Optics 37(31), 7320-6 (1998).

[3] Southwest Sciences. Laser Gas Sensing Tutorial. http://www.swsciences.com/technology/sensors.html

 

While the oxygen is being removed at a rate of about 5 %, both the carbon dioxide and water concentration increase at a rate of about 4% at each following breath. The vital capacity of the lungs is always 1 to 1.5 liters less than the total capacity of about 5 liters because the lungs cannot be completely deflated without serious damage.

Lets consider a human breathing in a quasi-sealed 27000 L (3x3x3m) cargo container for 30 minutes with a breathing rate of 12 times per minute inhaling and exhaling 1L of gas at each breath.  Even though CO2 is 1.57 times heavier than nitrogen and 1.38 times heavier than O2, it will have a tendency to disperse in an isolated volume of air, due to molecular diffusion. The Gaussian probability of a gas molecule to travel due to the Brownian motion on a distance of x is:

where Dt  is the diffusion coefficient of the gas in the given media at a given temperature t.  

The diffusion coefficient of carbon dioxide in air at normal atmospheric conditions is rated at 10 m2/s
[1]. From (1) it follows that about 7% of CO2 molecules can travel from one side to another side of the container (~3 m) per second.  Accounting for this and Table 1 each following breath will increase the CO2 content on the opposite side of the container by approximately 0.0028 L, which corresponds to a concentration of ~104 ppb. This number is much higher than the detection limit of 0.13 ppb indicated above and it will double, triple, etc. with each following breath. This gives some room for measurements feasibility when the diffusion rate is lowered due the container contents.

The diffusion coefficient of water vapor into air at normal atmospheric conditions is much smaller than that of carbon dioxide and is rated at 0.226 x l0-4 m2/ s [2], so it will take a much longer time for the vapor molecules to travel the same distance (~3m) due to Brownian motion. By using (1) the amount of vapor molecules that travel for ~3 m during an hour will be 6.8×10-11 % of total molecules exhaled. By taking into the account the data given for water vapor in Table 1 the amount of vapor traveling ~3 m over an hour will be 4 x 10 –14L, which corresponds to a concentration of 0.00014 ppb, which is way below the detectable limits.

The above estimations indicate that only CO2 measurements should be considered within the proposed approach. A comparison of the absorption and spectral characteristics of the proposed LEDs and phototedectors for CO2 measurements is shown in Figure 1.

A simplified configuration of the detector is shown in Figure 2. The matched LED/photodetector pairs will be mounted at a distance of about 1” across each other with optically opaque separations between the compartments in order to prevent background noise.  The space between LEDs and photodetectors will be directly connected with the internal cargo ambient atmosphere through a gas permeable high-conductivity optically opaque filter that will filter out mechanical particles even of submicron size. The electronic control, acquisition, analysis, and transmission systems will be included within the same housing.

In order to avoid false alarms we can not only monitor the concentration level of CO2 with time but the periodic oscillations of the concentration levels as a result of breathing can be also detected and analyzed. The frequency of normal human breathing is in the range of 10 to 15 times per minute. Lets assume a uniform breathing rate of 12 times per minute, with equal times for inhaling and exhaling intervals. Let us also ignore the gas leak out of the cargo container. Using data from Table 1, the CO2 concentration level measured during a time of 1 minute can be roughly modeled as a monotone periodic function. The sensor output signal will change similarly.  By analyzing both the slope and the frequency of the signal change the CO2 level changes can be identified as resulting from human breathing.

There are possibilities of small animals, such as rats or mice to be present in cargo containers.  However, the normal breathing rate of rats ranges from 91 to 132 breaths/min[3], while a running human may breath as often as only 50 times a minute[4], so their breathing rate ranges do not overlap even in extreme conditions providing for the perfect mean to avoid false alarms.

The considerations above indicate that false alarm-free detection and identification of humans in cargo containers using NDIR CO2 measurements combined with intelligent signal analysis   is feasible and can be realized by employment of miniature, inexpensive, low-power solid-state optoelectronic and electronic components.

[1] http://www.netl.doe.gov/publications/proceedings/01/carbon_seq/7b1.pdf

[2] http://wwwt.emc.ncep.noaa.gov/gmb/wd23ja/doc/web2/chap5.html

[3]Evelyn H. Schlenker. Dextromethorphan affects ventilation differently in male and female rats. Journal of Applied Physiology. Vol. 81, No. 5, pp. 1911-1916 (1996).

[4]Jerry G. Johnson. Biology II.  http://www.sirinet.net/~jgjohnso/respiratory.html

As light sources, for NDIR measurement, miniature (TO-5 packaged) relatively inexpensive low power consumption LEDs based on InAs/InAsSbP heterojunnctions emitting at a maximum wavelength of 4.2 mm are currently available[1] along with a matching photodetector, a PbSe-based photoresistor sensitive in the range 4.3 mm and below. In order to evaluate the capability of the setup lets calculate the concentration of the CO2 in the container resulting from 30 minute breathing. The volume of this portion will be 1.008 L taking into account that a mole of gas in normal conditions has a volume of ~24.5L and assuming the same container volume of 27000 L, the molar concentration of the gas will be 1.524xmol/L. The absorption of light from the LED caused by presence of CO2 can be evaluated by using the Beer-Lambert law:

I1/I0=exp(-alc)     (2)

where I1 and I0 are the intensities of the light passing through the gas and without it, respectively;  a is the molar absorptivity of CO2 at a wavelength ~4.3 mm (or a wave number of 2342 cm-1) that equals 1280 L/mol cm [2]; l is the length of the optical path assumed l=2.5 cm; and  c is the CO2 concentration of 1.524×10-6  mol/L calculated above. By putting the numeric values into (2) we obtain:                                                 I1/I0»0.995132613      (3)

As shown in Figure 3, the proposed LED43 emits light in a parallel 5 mm diameter beam provided by a parabolic reflector of a total optical power of 0.000001 W according to its specifications. That produces an optical power density of 5.1×10-8 W/mm2. The photoresistor’s (PR43) sensitive area is 2 x 1.8 mm2 and a sensitivity of 500 V/W, as given in the specifications. Since the photosensitive area is completely covered by the LED beam, the photoresistor under light with a power density of 5.1×10-8 W/mm2 will produce a signal of ~91.800 mV. The power density of the LED after passing a ~2.5 cm (~1”) thick layer of CO2 will be ~91.35317389 mV. The signal change of about 0.45 mV can easily be measured using standard electronic circuitries.

The LED43 and PR43 will be pre-mounted in the enclosed measurement chamber across each other by a miniature hardware setup that allows for focusing in order to achieve maximum response from the PR43 at a set current through the LED 43, and then fixing the two components in the selected position. The electronic control of the LED43 and PR43 and signal measurements can be performed through a sealed multi-pin connector mounted on one of the measurement chamber walls.

A hollow fiber composite membrane made of hydrophobic micro-porous polyethylene with a 1 μm thick non-porous polyurethane layer in the middle of the membrane wall that has been developed for atmosphere analysis[3] can be used at two opposite chamber walls. Even in contact with water, the membrane pores do not wet and remain dry and gas-filled. Gas diffuses through both the gas-filled pores and the polyurethane. The hollow fiber membranes provide a high specific surface area and as a result, optical opacity and a high gas transfer rate can be simultaneously achieved.

[1] http://www.roithner-laser.com/

[2] Raymond J. Haines, Rebecca E. Wittrig, and Clifford P. Kubiak. Electrocatalytic Reduction of Carbon Dioxide by the Binuclear Copper Complex [ Cu2(6-(diphenylphosphino)-2,2′-bipyridyl)2(MeCN)2]

Inorg. Chem. 33, 4723-4728 4723 (1994).

[3] Ahmed, T., Semmens, M.J., Voss, M.A., Oxygen transfer characteristics of hollow-fiber, composite membranes

Advances in Environmental Research, (2004).

The sensor system prototype can use a laboratory-based portable electronic system, controlled by a notebook computer, in order to provide for sensor testing and calibration in field-simulated conditions, and to tune our detection algorithm.

The absorption of 4.3 mm wavelength light by the carbon dioxide content present in the container will be continuously measured via a PR43 phototresistor, during one-minute periods that are repeated every half an hour. Timing will be provided by a built-in clock that triggers the measurements. The

schematic of the measurements is shown in Figure 4. The resistance change produced by the transducer will be converted to voltage by a bridge circuit to benefit from its high sensitivity to small changes in resistance. Also, since the resistance change in the transducer will be very small, it will be susceptible to noise contamination once converted to voltage. The bridge circuit will eliminate any common mode noise signal that accompanies the output signal from the bridge.

The next stage is the low-noise instrumentation amplifier that will amplify the incoming signal to a level that is sufficient for digitization by the Analog to Digital (A/D) converter. Since the whole process is performed at a low frequency, choices of low-noise amplification stage and high resolution A/D converters are attainable. The digitized data should be collected by a microcontroller, which can perform regression analysis of comparing the data with the concentration slope. The digitized data collected by the microcontroller should be also supplied to the computer for Fourier analysis. This analysis is important to find the frequency spectrum of the data, so as to distinguish from false alarms created by non human breathing patterns. Both the results from the microcontroller and the Fourier spectrum are analyzed using either a regular decision-tree or fuzzy algorithm against preset thresholds, thus inferring the presence or absence of humans in containers. In the end, a final comprehensive and fine tuned detection algorithm that can be implemented using a system on a chip (SoC) approach will be obtained.

The project is also directed towards the development of a compact remotely operated low-cost control, acquisition, and analysis circuitry that uses a high autonomy low power source. For this purpose the architecture should be modified to accommodate the wireless transmission of data. The schematic of the signal transmission part of the control system is shown in Figure 5. The data collected by the microcontroller are analyzed by the microcontroller itself. The results along with the collected data will be transmitted upon request through a standard secure wireless communication protocol to a secure portable or pocket

computer for display. In the event of an alarm, the results can also be automatically transmitted, via a wireless link, to a preset phone number as a SMS message, or to an remote computer that has an Internet address. In all circumstances, the data/result on the system can be accessed at any time by authorized users using a secured wireless communication protocol. Also, the detection end of the system including its microcontroller and transceiver will be designed using low power technology. Given that the bursts in power consumption will only occur during wireless transmission and that the systems consumes minimal energy otherwise, a battery operated system can be designed that has an autonomy of up to 6 months, depending on the wireless technology used and the frequency of transmission.

Some conservative estimates for cost would be $150 for the microcontroller-ADC subsystem in the server end, and $50-200 for the associated wireless transceiver, depending on which technology is used. At the client end, there is no material cost if WI-FI is chosen as the wireless technology, since its transceiver is already present in most current laptops/pocket PCs; else the cost is between $100 and $200 per transceiver, again depending on the technology.