The development and investigation of a small associated-particle sealed-tube neutron generator (APSTNG) shows potential to allow the associated-particle diagnostic method to be moved out of the laboratory into field applications. The APSTNG interrogates the inspected object with 14-MeV neutrons generated from the deuterium-tritium reaction and detects the alpha-particle associated with each neutron inside a cone encompassing the region of interest. Gamma-ray spectra of resulting neutron reactions identify many nuclides. Flight-times determined from detection times of the gamma-rays and alpha-particles can yield a separate course tomographic image of each identified nuclide, from a single orientation. Chemical substances are identified by comparing relative spectral line intensities with ratios of elements in reference compounds. The high-energy neutrons and gamma-rays penetrate large objects and dense materials. Generally, no collimators or radiation shielding are needed. Proof-of-concept laboratory experiments have been successfully performed for simulated nuclear, chemical warfare, and conventional munitions. Most recently, inspection applications have been investigated for radioactive waste characterization, presence of cocaine in propane tanks, and uranium and plutonium smuggling. Based on lessons learned with the present APSTNG system, an advanced APSTNG tube (along with improved high voltage supply and control units) is being designed and fabricated that will be transportable and rugged, yield a substantial neutron output increase, and provide sufficiently improved lifetime to allow operation at more than an order of magnitude increase in neutron flux.

A neutron transmission spectrometer has been used to obtain the neutron attenuation signature of a number of substances, including contraband such as drugs and explosives. The neutron attenuations of these substances were determined for the substances alone and when imbedded in suitcases. A pulsed white neutron source was created by bombarding a thick beryllium target with a 5 MeV pulsed deuteron beam. The neutron intensity was measured from about 0.75 MeV to about 4 MeV with the samples in and out of the neutron beam to determine the neutron attenuation. The experimentally determined neutron attenuation curves were used to determine the number of densities of H, C, N, and O throught the samples using measured neutron cross sections. Explosives, drugs, and plastics each fall into different portions of the 3D space formed by certain ratios of H, C, N, and O. The results are presented in 3D representations.

A fast-neutron technique that employs isotopic sources is proposed for the identification of explosive materials. The technique relies on analyzing the spectrum of transmitted neutrons. Identification methods are developed to enable discrimination between explosives and other materials. A threshold-value method and artificial neural networks are used. Monte Carlo simulations, as well as laboratory experiments, are utilized to demonstrate the feasibility of the proposed techniques.

Fast-neutron inspection techniques show considerable promise for explosive and narcotics detection. A key advantage of using fast neutron is their sensitivity to low-Z elements (carbon, nitrogen, and oxygen), which are the primary constituents of these materials. We are currently investigating two interrogation methods in detail: fast-neutron transmission spectroscopy (FNTS) and pulsed fast-neutron analysis (PFNA). FNTS is being studied for explosives and narcotics detection in luggage and small containers for which the transmission ration is greater than about 0.01. The Monte Carlo radiation transport code MCNP is being used to simulate neutron transmission through a series of phantoms for a few (3-5) projections angles and modest (2 cm) reolution. Areal densities along projection rays are unfolded from the transmission data. Elemental abundances are obtained for individual voxels by tomographic reconstruction, and the reconstructed elemental images are combined to provide indications of the presence or absence of explosives or narcotics. PFNA techniques are being investigated for detection of narcotics in cargo containers because of the good penetration of the fast neutrons and the low attenuation of the resulting high-energy gamma-ray signatures. Analytic models and Monte Carlo simulations are being used to explore the range of capabilities of PFNA techniques and to provide insight into systems engineering issues. Results of studies from both FNTS and PFNA technqiues are presented.

Coherent scatter measurements have been shown to be a potentially useful tool in the detection of energetic materials. The relationship between explosive volume, method of concealment, environment, and resulting threat has been considered in the design of our energy dispersive low angle x-ray scatter system. The principle application of the system is for the detection of explosives concealed within passenger luggage in sheet form. The effects of parameters such as scattering angle, beam collimation, explosive material geometry, overlying materials and counting statistics have been measured experimentally and comparison made with theoretical prediction. It has been found that: 1) scatter signature analysis dictates optimum scattering angles for different explosives, and 2) partial volume effects must be combined with scatter angles to give total system optimization.

While Raleigh scatter radiography can result in image contrast far higher than conventional transmission methods for the detection of low-Z inclusions in light element matrices a means of separating the scattering from each of the individual components is desirable. This would then produce images of each material alone, if effect stripping away the scatter due to the other elements. This has been done by collecting image data at a series of scattering angles each corresponding to a maximum in the diffraction pattern of each constituent in the sample. The data at such angles is then due to the scatter from the material of interest plus a fraction of the scattering from each of the other components. This fraction is measurable from standard scatter patterns. By measuring at several angles a series of simultaneous equations can be obtained and solved to determine the scatter from each component alone. This technique is demonstrated using Rayleigh scatter image data from a typical industrial phantom. Both normal scatter images and individual component images are presented and the very large contrasts in the latter demonstrated. The statistical noise in the individual component images is greater than that for the total scatter images due to accumulating errors arising when solving the simultaneous equations. However the greatly increased contrast enables components to be distinguished more sensitively resulting in shorter run times.

Coherent x-ray scatter is a very powerful analytical tool for the unambiguous identification of explosives having a polycrystalline structure. Its use in a realistic security screening environment depends on optimum matching of the measurement arrangment to the detection problem in order to minimize erroneous recognition while fulfilling stringent scan time requirements. The influence of such parameters as x-ray tube energy, scatter angle, primary beam geometry, and detector arrangement on system performance are considered and ways in which they may be optimized are discussed.

For plastic explosives detection with coherent x-ray scattering, x-ray tubes are required having high brightness and a 'white' spectral shape. Since the photon spectrum of reflection-type x- ray tubes drops rapidly above the K edge, the properties of transmission anode x-ray tubes have been investigated. In order to maximize the brightness, the use of a geometry in which the intensity of bremsstrahlung radiation has a maximum is proposed. Experimental results obtained from such an x-ray tube are presented.

Potential security screening applications of a novel fluorescent x-ray source are discussed. The instrument incorporates a secondary tantalum target within the tube head and its output energy spectrum shows only the Ta fluorescence lines superimposed on a smooth, low, background. Output radiance for optimum operation is 5.9 X 10⁹ photons s^-1sr^-1mm^-2. Densitometry measurements were made on the sample volume formed by the overlap of the highly collimated primary and scattered beams and the ratio of the elastically to inelasticity (Compton) scattered signal was found. This ratio varies approximately as the square of the atomic number and its use reduces errors due to attenuation and geometry. The two main limitations of the ratio method are statistical noise and systematic effects such as multiple scattering and self-attenuation of the primary and scattered beams by the sample. These can be minimized by employing a forward scattering geometry and using a K edge filter to separate the small angle elastically and Compton scattered signals. The feasibility of the use of cheap scintillation detectors in conjunction with filters as opposed to more expensive energy dispersive detectors is demonstrated for low density materials and the implications for contraband detection discussed.

A novel wideband millimeter-wave imaging system concept has been developed at Pacific Northwest Laboratory for the detection of concelaed weapons. Millimeter-waves are ideal for personnel surveillance applications since they will readily penetrate common clothing materials, and have relatively short wavelengths allowing for high resolution imaging. This system concept is based on a circular synthetic aperture imaging technique, in which a circular aperture is scanned and an image is formed of the target located near the scanned aperture. A laboratory imaging system has been developed and results have been obtained using both mannequins and humans with concealed weapons. The technique is readily adaptable to a real- time imaging system using a relatively small number of transceivers and a relatively slow scanner speed.

A new wideband millimeter wave holographic imaging technqiue is under developement for use in concealed weapons detection system. This new wideband technique provides far superior images than single frequency holographic techniques on thick objects such as the human body. The wideband technique obtains fully focused images over a designated volume and provides excellent lateral and depth resolution. Using this method, a 3D volumetric hologram is gathered with a millimeter wave linear array, a mechanical scanner, and a sweep frequency tranceiver. The 3D volumetric hologram is then processed by high-speed computational processors to reconstruct the fully focused image. Two prototype wide band millimeter wave holographic arrays have been developed at the Pacific Northwest Laboratory. The two arrays consist of sequentially switched 2 by 37 Ku band (12.5-18 GHz) and 2 by 64 Ka band (26.5-40 GHz) systems which are coupled to high-speed sweep frequency heterodyne transceivers. The arrays are used to obtain volumetric imaging data at high speeds by electronically sequencing and frequency sweeping the array antennas along 1D while performing a mechanical scan along the other dimension. The current prototype system scans an aperture the size of a large human body in about one second. Extensive laboratory testing has been performed with people carrying various concealed weapons and innocuous items with both imaging arrays during the first quarter of 1995.

The characterization of the type and quanity of explosives residue left behind as fingerprints is critical for the problem of trace explosives detection as well as forensic investigation. A nondestructive analytical technique has to be used to identify the energetic component of the explosive from the plasticizers, dyes, and fingerprint oils that make up the background. Raman microspectroscopy has been demonstrated in the past to separate explosive particulate from other residue in a microscopic image by filtering out other spectra except the region of the strong bands displayed by PETN and RDX using He-Ne excitation. In addition, gray level/measurements have been done on features of the sample, captured under white light onto a CCD, to obtain quantitative data about size and volume distribution. The objective of this paper will be to show how integrated line images of the sample, captured with high spectral resolution using a scanning Raman spectrometer, can be used to separate out components in the image scene captured by the CCD. This paper will also show how confocal scanning through the depth of the sample, while taking an image, can be used to come up with a quantitative measure of the concentration of chosen components in the entire image. The special instrumentation used for the work will be shown as well as any modifications done to it to obtain a protocol for analysis. The image analysis results will be presented of actual fingerprint samples containing plastic explosives. The variance between the Raman imaging method and other more traditional destructive methods for doing quantitative analysis will be presented. And the probability of doing direct Raman microspectroscopy in the UV region without any background subtraction will be determined for its potential for doing in-situ analysis for explosives detection.

In explosives detection with coherent x-ray scattering, radiation scattered in the forward direction at a small fixed angle is accumulated from volume voxels of interest with an energy- resolving multichannel detector. For most military and industrial explosives, the resulting scatter spectra exhibit characteristic diffraction peaks which correspond to their polycrystalline structure. The performance of a spectrum processing and classification scheme can be analyzed by simulation. The simulation is based on spectra measured with good counting statistics. It allows the influence of the following parameters or effects to be studied: 1) total photon flux, responsible for the quantum noise; 2) arbitrary mixing of the measured substances and partial volume effects; and 3) attenuation of the primary and scatter radiation.

This paper investigates the use of tranform coding and the neural tree network on data obtained from two security systems; face recognition and explosive detection. The use of discrete cosine transform components as features for classification are demonstrated on face recognition data. The use of cepstral components as features for classification are demonstrated for explosive detection on coherent x-ray scattering data, where surrounding materials nonlinearly affect the spectral data obtained from crystalline explosives. The neural tree network is described and shown to be an effective classifier in both applications.

The theory of stochastic signal processing describes signals, which contain statistical fluctuations, by their statistical properties like mean and variance. Since the fluctuations, or simply the noise, in coherent x-ray scatter signals are known to follow Poisson statistics, this theory can be used to derive detection properties under different system conditions. This paper will show how the Poisson-noise can be transformed into white noise and that a false detection rate for signals containing white noise can be calculated. Furthermore the application to the optimization of collimator design will be studied.

This paper describes a method for determining the range of visibility of a driver or observer with normal vision by means of analyzing image sequences. The scope of application of such a range-finding system can be seen on the one hand as a source of information to assist the human driver in cases of poor visibility, and on the other hand as part of a higher decision- making stage in an image-based automatic vehicle guidance system. For safe vehicle guidance these systems require information concerning visibility conditions in order to produce reliable interpretations of image data. The system may also be of some help in the developement of vehicles where it could serve as a tool concerning headlamp technology.

Image sequence processing has become important in a wide variety of practical applications, such as robot vision, traffic control, etc. Change detection plays a very important role for it allows the tracking of moving objects. Unfortunately, one remaining problem is due to the fact that temporal and spatial changes of illumination, induce false detections. The present approach can be used for the surveillance of a robotic site and analysis of traffic flow under uncontrolled ambient light. The aim of this work is to develop a safety device based on artificial vision which allows to avoid collision between human and moving machines. This paper presents a new approach of surveillance allowing unpredictable robotic tasks and tolerating independent illumination chnages.

We present a driver face recognition system for comfortable access control and individual settings of automobiles. The primary goals are the prevention of car thefts and heavy accidents caused by unauthorized use (joy-riders), as well as the increase of safety through optimal settings, e.g. of the mirrors and the seat position. The person sitting on the driver's seat is observed automatically by a small video camera in the dashboard. All he has to do is to behave cooperatively, i.e. to look into the camera. A classification system validates his access. Only after a positive identification, the car can be used and the driver-specific environment (e.g. seat position, mirrors, etc.) may be set up to ensure the driver's comfort and safety. The driver identification system has been integrated in a Volkswagen research car. Recognition results are presented.

The concept of a lidar is the optical synonym of the well known radar. It describes an optical remote sensing technique for the detection of the atmospheric parameters like the optical density, etc. The name lidar originates, analogous to radar, from the first application of this technology, the determination of the distance to large solid bodies (laser range finder). The evolution of the lidar technique reaches from the measurement of the distance to diffuse targets, for example clouds (cloud ceilometer), to the determination of the range resolved turbidity of the atmosphere (visibility sensor). The modern lidar systems have been reduced so far, that they can be built into a van or even a car. The problem of the eye-safety of the laser radiation has been solved by using low power laser diodes, but with a high repetition rate. The first results with a modified laser range finder, manufactured by Jenoptik GmbH, will be reported. This device, at the size of binoculars, promises to be a first approach for a handheld visibility sensor, measuring local turbidities, like fog banks.

Laser doppler velocimetry can provide many advantages to traffic monitoring, as compared to tradiitional microwave systems. Narrower beamwidth, lower transmitted power and higher doppler frequency shift are some of them. A low-power laboratory prototype is presented. It is based on the Michelson interferometer. Measurements of working range, SNR and resolution are presented.

In this paper, infrared distance sensors are compared regarding technology, environmental, and practical aspects. Different methods for obtaining lateral resolution and covering the required detection range are presented for both sensor technologies. Possible positions for sensor installation at the test vehicle have been tested. Experimental results regarding cleaning devices and other environmental problems are presented. Finally, future aspects, e.g. speed over ground measurements or technological steps are discussed.

Most explosive detection technologies have been focused on nitro-based military explosives becuase they have figured in international terrorist incidents. Not only are they readily available through purchase or theft or from sponsoring states, but methods for home synthesis are widely available. Many of explosive detection technolgies now under development target a specific characteristic of military or commercial explosives (e.g. mass density, nitrogen density). However, as counterterrorist measures make traditional explosives more difficult to obtain or more risky to use, we should anticipate terrorists may turn to nontraditional explosives. There are hundreds of energetic compounds and many common explosives which, while they do not meet exacting military demands, might be effective terrorist tools. Although explosive handbooks list hundreds of explosives, this talk focuses on only a handful. These have been chosen because they do not follow the classic patterns of military explosives or because they are easily obtainable. This paper will also point out energetic systems that can produce violently exothermic reactions without the aid of traditional initiating systems, such as batteries or detonators.

We present in this paper, an electronic copilot embedded in a real car. The system objective is to help the driver by sending alarms or warnings in order to avoid dangerous situtations. An onboard perception system based on CCD cameras and proprioceptive sensors is used ot provide information concerning the environment and the internal state of the vehicle. From this set of information, the copilot is able to analyze the situation and to generate adequate warnings to the driver according to the circumstances. The definition and the development of such a system deal with multisensor data fusion and supervision strategies. The framework of this work was the European Prometheus Pro-Art program. The electronic copilot has been integrated in a prototype vehicle called Prolab2. This French demonstrator integrates the works of nine research laboratories and two car companies: PSA and RENAULT. After a brief presentation of the global demonstrator, we present the two principal parts developed in our laboratory corresponding to the high level modules of the system: the dynamic data manager and the situation supervision.

The MALDI (matrix assisted laser desorption/ionization) technique, widely used to desorb and ionize large biomolecules, is applied here to small molecules having low vapor pressure, such as drugs and explosives. Furthermore, we report the coupling of the MALDI technique with a small, highly portable tiny-TOF (fime-of-flight) mass spectrometer developed in our laboratories. This mass spectrometer is a low voltage coaxial reflectron design with a short flight tube that is specifically designed for low molecular weight substances. The reflectron is designed to operate in two different modes that provide an expremely powerful pseudo-tandem mass spectrometry capability that is crucial for field applications. Using this system we have measured mass spectral signatures for cocaine, heroine, and the explosive RDX in the sub- nanogram range. Also reported here are continued developements on advanced MALDI sampling technologies, sensitivity, and mass resolution enhancements of the tiny-TOF, further decreases in system size and weight, and concepts for field operational systems.