Mark L. Yee

Multisensor probabilistic fusion for mine detection

In this paper, probabilistic fusion of multi-sensor data is applied to mine detection. Probabilistic fusion combines information in the form of scores from automatic target recognition (ATR) algorithms for each sensor. This fusion method has previously demonstrated improved mine detection performance when used with multi-sensor data from the Mine Hunter/Killer system. The sensor suite includes a ground- penetrating radar, metal detectors, and an IR camera; data were collected at a prepared test site. Results of applying the probabilistic fusion method to recent MH/K multi-sensor data using various new ATR algorithms are presented and analyzed in detail. Changes in detection performance are quantified for different combinations of the various ATR algorithms and sensors. It is shown that fusion improves mine detection performance even when the individual sensor and ATR algorithms have very different performance levels. This implies that multi-sensor approaches to mien detection should continue to be pursued.

Distributed algorithms for small vehicle detection, classification and velocity estimation using unattended ground sensors

This study developed a distributed vehicle target detection and estimation capability using two algorithmic approaches designed to take advantage of the capabilities of networked sensor systems. The primary interest was on small, quiet vehicles, such as personally owned SUVs and light trucks. The first algorithm approach utilized arrayed sensor beamforming techniques. In addition, it demonstrated a capability to find locations of unknown roads by extending code developed by the Army Acoustic Center for Excellence at Picatinny Arsenal. The second approach utilized single (non-array) sensors and employed generalized correlation techniques. Modifications to both techniques were suggested that, if implemented, could yield robust methods for target classification and tracking using two different types of networked sensor systems.

Algorithm for Automatic Mine Detection using a Vehicle- Mounted Metal Detector

A new algorithm for automatic detection of mines using a vehicle-mounted metal detector is detailed and demonstrated in this paper. The sensor is the Vallon VMV 16 vehicle- mounted array on the Mine Hunter/Killer (MH/K) system; data were collected at a prepared test site. The resulting data have two notable characteristics. First, the sensor outputs a linear combination of the integrated time decay response over separate time windows, and not the time decay response itself. Second, the data are sampled in two dimensions, along the 16 coil array and along the direction of travel of the vehicle. Thus 2-dimensional spatial processing techniques are used, treating the sensor data as a 2-D array of pixels. Our algorithm also uses adaptive methods to enable detection of both high and low metal content mines. Mine detection results using this algorithm are presented and analyzed for the first time. In particular, the ability to detect mines with low metal content in the presence of noise will be quantified and discussed.

Multitarget data association using an optical neural network

A neural network solution to the data association problem in multitarget tracking is presented. It uses position and velocity measurements of the targets over two consecutive time frames. A quadratic neural energy function, which is suitable for an optical processing implementation, results. Simulation resultsusing realistic target trajectories with target measurement noise including platform movement or jitter are presented. The results show that the network performs well when track data are corrupted by significant noise. Several possible optical neural network architectures to implement this algorithm are discussed, including a new all-optical matrix-vector multiplication approach. The matrix structure is employed to allow binary-ternary spatial light modulators to be used.

Burn-depth estimation using thermal excitation and imaging

Accurate estimation of the depth of partial-thickness burns and the early prediction of a need for surgical intervention are difficult. A non-invasive technique utilizing the difference in thermal relaxation time between burned and normal skin may be useful in this regard. In practice, a thermal camera would record the skins response to heating or cooling by a small amount--roughly 5 degree(s) Celsius for a short duration. The thermal stimulus would be provided by a heat lamp, hot or cold air, or other means. Processing of the thermal transients would reveal areas that returned to equilibrium at different rates, which should correspond to different burn depths. In deeper thickness burns, the outside layer of skin is further removed from the constant- temperature region maintained through blood flow. Deeper thickness areas should thus return to equilibrium more slowly than other areas. Since the technique only records changes in the skins temperature, it is not sensitive to room temperature, the burns location, or the state of the patient. Preliminary results are presented for analysis of a simulated burn, formed by applying a patch of biosynthetic wound dressing on top of normal skin tissue.

Multiple target-to-track association and track estimation system using a neural network

A three-component system for tracking multiple moving objects is presented. A neural network is used to perform frame-to-frame data association. A Hough Transform system is used to perform multiple-frame data association and track correction. An estimation filter system is used to provide updated track estimates. The tracking ability of this integrated system is tested with realistic simulated flight trajectories. The system response to simulated measurement noise and estimation errors is detailed, and the interaction of the three system components to correct errors is illustrated. Optical processing is used in the neural net and Hough Transform systems.

Optimization neural net for multiple-target data association: real-time optical lab results

Abstract The Hopfield neural network was first used for optimization in solving the famous Traveling Salesman Problem.We have applied a similar approach to the solution of another problem, namely data association for multiple targets.Simulation data are presented which demonstrate the networks ability to successfully determine the optimum dataassociation solutions, with target noise present. Simulations also indicate the ability to solve the problem on a low accuracy (analog optical) processor. Optical implementation issues are discussed, and an bptical architecture is presented with laboratory results. Introduction \t=n+1 t=n xx x association matrixtarget estimates target measurements Figure 1 : ifiustration of the Data Association Problem.The data association (DA) problem for multitarget tracking (MU) is illustrated in Fig. 1 . Existing targets in agiven time frame must be associated with target measurements obtained in the next time frame. The existing targetdata are target state estimates which have been computed using previous measurement data. The properly associated

Application of normalized gray-scale correlation

Real-time gray-scale correlation in the spatial domain has been demonstrated previously using an acousto-optical (AO) correlator. This work demonstrates normalized gray-scale correlation as implemented on an AO correlator system capable of operating at real-time video rates. Motivation for using normalized gray-scale correlation is presented. The normalized correlation algorithm as implemented on the AO correlator is detailed. The entire real-time AO correlator system is described, including the electronic support hardware and the user interface. Since normalization requires a division operation, system numerical precision issues are addressed. Test results obtained in non-real time experiments are presented.

Multitarget Tracking With An Optical Neural Net Using A Quadratic Energy Function

Multitarget tracking over consecutive pairs of time frames is accomplished with a neural net. This involves position and velocity measurements of the targets and a quadratic neural energy function. Simulation data are presented, and an optical implementation is discussed.

Automatic recognition of Malicious intent indicators

A major goal of next-generation physical protection systems is to extend defenses far beyond the usual outer-perimeter-fence boundaries surrounding protected facilities. Mitigation of nuisance alarms is among the highest priorities. A solution to this problem is to create a robust capability to Automatically Recognize Malicious Indicators of intruders. In extended defense applications, it is not enough to distinguish humans from all other potential alarm sources as human activity can be a common occurrence outside perimeter boundaries. Our approach is unique in that it employs a stimulus to determine a malicious intent indicator for the intruder. The intruders response to the stimulus can be used in an automatic reasoning system to decide the intruders intent.