Showing papers by "Grant R. Gerhart published in 1998"

Detection probability using relative clutter in infrared images

Abstract: Clutter plays a very important role in the area of machine and human-in-the-loop target acquisition. A great deal of interest has recently been shown in assessing several different definitions of clutter. In spite of so many definitions available, no single clutter definition has been agreed on by the target acquisition modeling community as being the best. Here we develop a new clutter metric, called relative clutter, based on factor analysis which is extensively used for statistical analysis. This relative clutter combines many definitions of clutter. Different methods for calculating the relative clutter based on the magnitude of the eigenvalues obtained from the correlation matrix are suggested. The relative clutter of many images is analyzed. The relative clutter is used to calculate probability of detection on Night Vision Lab (NVL) Terrain Board Infrared images.

A relative clutter metric

Abstract: Clutter plays a very important role in the area of machine and human-in-the-loop target acquisition. A great deal of interest has recently been shown in assessing several different definitions of clutter. In spite of so many definitions available, no single clutter definition has been agreed on by the infrared community as being the best. We develop a new clutter metric, called relative clutter, which is based on factor analysis which is extensively used for statistical analysis. This relative clutter combines many definitions of clutter. Different methods for calculating the relative clutter based on the magnitude of the eigenvalues obtained from the correlation matrix are suggested. The relative clutter of many images is analyzed.

Guest Editorial: Special Section on Advances in Target Acquisition Modeling

Abstract: The U.S. Army Tank-Automotive Research, Development and Engineering Center (TARDEC) has had a broad interest in modeling and simulation techniques during the last several decades. Specifically, as army ground vehicle designers, our group is interested in predicting the performance of military observers for detecting and discriminating vehicle targets in complex background scenes.

U.S. Army TARDEC R&D program on intelligent subsystems for enhanced mobility

Abstract: The U.S. Army Tank-automotive and Armaments Command Research Development and Engineering Center (TARDEC) has iniliated a systems engineering program to develop mutually compatible and complementary components and subsystems for enhanced Unmanned ground vehicles (UGV) mobility. UGV have historically lacked the mobility of manned tracked vehicles, especially in obstacle-crossing and off-road maneuver. To achieve comparable mobility, UGV require supervised navigation with enhanced locomotion subsystems and complementary, mutually adapted machine perception, routing and driving control. The TARDEC program is funding technology development, maturation and packaging to produce NonDevelopmental Items that can be scaled and configured for different UGV. The TARDEC is also developing a Systems Integration Laboratory (SIL) to evaluate compatibility and system-level performance of component technologies. This TARDEC program complements and extends the UGV Technology Exchange and Exploitation (UGVTEE) DEMO III program. It fills unfunded gaps in the Army Research Laboratory's Concerted Technology Thrust (CU) projects in machine perception and supervised navigation. It fills technology gaps in the integration of "smart" mobility subsystems with supervised navigation. It specifically addresses semi-autonomous navigation deficiencies in obstacle crossing, other wheeled UGV mobility issues not being addressed in UGVTEE or other UGV programs. As part of the systems engineedng activity to ensure that the ND! components will be scaleable across a wide range of UGV, TARDEC is developing modular concepts for UGV across a range ofweight classes and functions. Keywords: Supervised navigation, obstacle crossing, non-developmental items, systems integration laboratory, unmanned ground vehicles, locomotion

Adaptive wavelet-based contrast enhancement of IR imagery

Abstract: This paper presents a suite of techniques called the Adaptive Wavelet-based Contrast Enhancement Method (AWCEM) for improving the subjective quality of an image for observation by a human. The central idea in these techniques is the space-varying stretching of the contrast of an image by enhancing or attenuating its detail wavelet transform coefficients. The degree of stretch is governed by a multi- resolution region of interest mask which is generated by a low-level feature extraction mechanism. Intelligent image enhancement seeks to preserve and amplify target details while at the same time suppressing the background clutter. This calls for an enhancement mechanism that (1) is adaptive across the image, (2) performs some form of low-level feature extraction and (3) uses these features to control the level of contrast stretching. The proposed algorithms use a combination of the multiscale energies, edge strengths, texture and motion as the features. Local scale- space anomalies in the image seed the regions of interest. We present the suite of techniques within an interactive environment, the AWCEM Tool, which allows the user to choose and optimize the algorithm for the particular type of data encountered in the final application. The algorithm applies well to both single-channel IR and visual imagery.

Average search time for ground-vehicle target detectability

Abstract: This paper summarizes N.P. Travnikova's model as a method to compare average search times by military observers using powered optics such as binoculars. Both discrete and continuous scanning methods are considered for target searches. This empirical model quantifies which type of vision system is best suited for the most efficient target detection for a given field of view. An analysis is also provided of the relative importance of target diameter, background luminance, and contrast upon overall detectability with the subsequent results compared to known field test data. The detectability of specific military ground vehicles over a variety of search and target acquisition tasks with several off-the-shelf binoculars is examined. Some examples on various types of search studies, such as compare looking at the target, both line retrace time effects, etc., for a low contrast target are also considered. This paper consists of two sections. The first explains the derivation of methodology and limits of its applicability. The second section offers a parametric analysis that compares the relative importance of target diameter, background luminance, and contrast upon overall target detectability with the subsequent results compared to field test data.

Multiresolution motion estimation and target detection

Abstract: Detecting and characterizing motion in a scene can play a critical role in target detection algorithms, since many targets can be camouflaged so completely that, if they are not moving, they are nearly undetectable. However, once they begin moving, they `popout' and are immediately detected. Estimating motion is also important in human vision modeling, because motion is generally detected with peripheral vision, which can cover the field of regard much more quickly than foveal vision. In this paper, we present two hierarchical multiresolution methods for computing the optical flow in a scene. We use statistical properties of the resulting flow fields to compute a motion feature vector, which we relate to the conspicuity of the moving target in a scene via a neural network.