Publications

This project demonstrates the feasibility of a novel imager with a thickness measured in microns rather than inches. Traditional imaging systems, i.e. cameras, cannot provide both the necessary resolution and innocuous form factor required in many data acquisition applications. Designing an imaging system with an extremely thin form factor (less than 1 mm) immediately presents several technical challenges. For instance, the thickness of the optical lens must be reduced drastically from currently available lenses. Additionally, the image circle is reduced by a factor equal to the reduction in focal length. This translates to fewer detector pixels across the image. To reduce the optical total track requires the use of specialized micro-optics and the required resolution necessitates the use of a new imaging modality. While a single thin imager will not produce the desired output, several thin imagers can be multiplexed and their low resolution (LR) outputs used together in post-processing to produce a high resolution (HR) image. The utility of an Iterative Back Projection (IBP) algorithm has been successfully demonstrated for performing the required post-processing. Advanced fabrication of a thin lens was also demonstrated and experimental results using this lens as well as commercially available lenses are presented.

Research into the use of multiframe superresolution has led to the development of algorithms for providing images with enhanced resolution using several lower resolution copies. An integral component of these algorithms is the determination of the registration of each of the low resolution images to a reference image. Without this information, no resolution enhancement can be attained. We have endeavored to find a suitable method for registering severely undersampled images by comparing several approaches. To test the algorithms, an ideal image is input to a simulated image formation program, creating several undersampled images with known geometric transformations. The registration algorithms are then applied to the set of low resolution images and the estimated registration parameters compared to the actual values. This investigation is limited to monochromatic images (extension to color images is not difficult) and only considers global geometric transformations. Each registration approach will be reviewed and evaluated with respect to the accuracy of the estimated registration parameters as well as the computational complexity required. In addition, the effects of image content, specifically spatial frequency content, as well as the immunity of the registration algorithms to noise will be discussed.

The use of combined imagery from different imaging sensors has the potential to provide significant performance improvements over the use of a single image sensor for beyond-the-fence detection and assessment of intruders. Sensing beyond the fence is very challenging for imagers due to uncertain dynamic and harsh environmental conditions. The use of imagery from varying spectral bands can alleviate some of this difficulty by providing stronger truth data that can be combined with truth data from other spectral bands to increase detection capabilities. Imagery fusion of collocated, aligned sensors covering varying spectral bands [1,2,3] has already been shown to improve the probability of detection and the reduction of nuisance alarms. The development of new multi-spectral sensing algorithms that incorporate sensors that are not collocated will enable automated sensor-based detection, assessment, localization, and tracking in harsh dynamic environments. This level of image fusion will provide the capability of creating spatial information about the intruders. In turn, the fidelity of sensed activities is increased resulting in opportunities for greater system intelligence for inferring and interpreting these activities and formulating automated responses. The goal of this work is to develop algorithms that will enable the fusion of multi-spectral data for improved detection of intruders and the creation of spatial information that can be further used in assessment decisions.

There is a need in security systems to rapidly and accurately grant access of authorized personnel to a secure facility while denying access to unauthorized personnel. In many cases this role is filled by security personnel, which can be very costly. Systems that can perform this role autonomously without sacrificing accuracy or speed of throughput are very appealing. To address the issue of autonomous facility access through the use of technology, the idea of a ''secure portal'' is introduced. A secure portal is a defined zone where state-of-the-art technology can be implemented to grant secure area access or to allow special privileges for an individual. Biometric technologies are of interest because they are generally more difficult to defeat than technologies such as badge swipe and keypad entry. The biometric technologies selected for this concept were facial and gait recognition. They were chosen since they require less user cooperation than other biometrics such as fingerprint, iris, and hand geometry and because they have the most potential for flexibility in deployment. The secure portal concept could be implemented within the boundaries of an entry area to a facility. As a person is approaching a badge and/or PIN portal, face and gait information can be gathered and processed. The biometric information could be fused for verification against the information that is gathered from the badge. This paper discusses a facial recognition technology that was developed for the purposes of providing high verification probabilities with low false alarm rates, which would be required of an autonomous entry control system. In particular, a 3-D facial recognition approach using Fisher Linear Discriminant Analysis is described. Gait recognition technology, based on Hidden Markov Models has been explored, but those results are not included in this paper. Fusion approaches for combining the results of the biometrics would be the next step in realizing the secure portal concept.

With increased terrorist threats in the past few years, it is no longer feasible to feel confident that a facility is well protected with a static security system. Potential adversaries often research their targets, examining procedural and system changes, in order to attack at a vulnerable time. Such system changes may include scheduled sensor maintenance, scheduled or unscheduled changes in the guard force, facility alert level changes, sensor failures or degradation, etc. All of these changes impact the system effectiveness and can make a facility more vulnerable. Currently, a standard analysis of system effectiveness is performed approximately every six months using a vulnerability assessment tool called ASSESS (Analytical Systems and Software for Evaluating Safeguards and Systems). New standards for determining a facility's system effectiveness will be defined by tools that are currently under development, such as ATLAS (Adversary Time-line Analysis System) and NextGen (Next Generation Security Simulation). Although these tools are useful to model analyses at different spatial resolutions and can support some sensor dynamics using statistical models, they are limited in that they require a static system state as input. They cannot account for the dynamics of the system through day-to-day operations. The emphasis of this project was to determine the feasibility of dynamically monitoring the facility security system and performing an analysis as changes occur. Hence, the system effectiveness is known at all times, greatly assisting time-critical decisions in response to a threat or a potential threat.

Vulnerability analysis studies show that one of the worst threats against a facility is that of an active insider during an emergency evacuation. When a criticality or other emergency alarm occurs, employees immediately proceed along evacuation routes to designated areas. Procedures are then implemented to account for all material, classified parts, etc. The 3-Dimensional Video Motion Detection (3DVMD) technology could be used to detect and track possible insider activities during alarm situations, as just described, as well as during normal operating conditions. The 3DVMD technology uses multiple cameras to create 3-dimensional detection volumes or zones. Movement throughout detection zones is tracked and high-level information, such as the number of people and their direction of motion, is extracted. In the described alarm scenario, deviances of evacuation procedures taken by an individual could be immediately detected and relayed to a central alarm station. The insider could be tracked and any protected items removed from the area could be flagged. The 3DVMD technology could also be used to monitor such items as machines that are used to build classified parts. During an alarm, detections could be made if items were removed from the machine. Overall, the use of 3DVMD technology during emergency evacuations would help to prevent the loss of classified items and would speed recovery from emergency situations. Further security could also be added by analyzing tracked behavior (motion) as it corresponds to predicted behavior, e.g., behavior corresponding with the execution of required procedures. This information would be valuable for detecting a possible insider not only during emergency situations, but also during times of normal operation.

Sandia National Laboratories has been investigating the use of remotely operated weapon platforms in Department of Energy (DOE) facilities. These platforms offer significant force multiplication and enhancement by enabling near instantaneous response to attackers, increasing targeting accuracy, removing personnel from direct weapon fire, providing immunity to suppressive fire, and reducing security force size needed to effectively respond. Test results of the Telepresent Rapid Aiming Platform (TRAP) from Precision Remotes, Inc. have been exceptional and response from DOE sites and the U.S. Air Force is enthusiastic. Although this platform performs comparably to a trained marksman, the target acquisition speeds are up to three times longer. TRAP is currently enslaved to a remote operator's joystick. Tracking moving targets with a joystick is difficult; it dependent upon target range, movement patterns, and operator skill. Even well-trained operators encounter difficulty tracking moving targets. Adding intelligent targeting capabilities on a weapon platform such as TRAP would significantly improve security force response in terms of effectiveness and numbers of responders. The initial goal of this project was to integrate intelligent targeting with TRAP. However, the unavailability of a TRAP for laboratory purposes drove the development of a new platform that simulates TRAP but has a greater operating range and is significantly faster to reposition.