ATR Performance Modeling and Estimation

Dudgeon, Dan E.

doi:10.1006/dspr.2000.0375

Cited by 30 publications

(37 citation statements)

References 21 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…The analysis tools standardly used to measure ATR performance include classical receiver operating characteristic (ROC) curves, confusion matrices, and probabilities of correct decisions [2,[8][9][10]. The performance of an ATR algorithm is characterized in this paper in terms of the confusion matrix.…”

Section: Resultsmentioning

confidence: 99%

“…MIT Lincoln Labs have developed a complete ATD/R system for SAR data [2]. In [8] the approaches and issues currently in use for modeling the performance of SAR ATR systems are presented. At Wright State University, they describe a standard experiment to use as a benchmark for the SAR ATR algorithm [9].…”

Section: Literature Survey Of Target Classification Recognition Andmentioning

confidence: 99%

See 1 more Smart Citation

Automated Target Recognition Using the Karhunen–Loéve Transform with Invariance

Suvorova¹,

Schroeder²

2002

Digital Signal Processing

View full text Add to dashboard Cite

Section: Resultsmentioning

confidence: 99%

Section: Literature Survey Of Target Classification Recognition Andmentioning

confidence: 99%

Automated Target Recognition Using the Karhunen–Loéve Transform with Invariance

Suvorova¹,

Schroeder²

2002

Digital Signal Processing

View full text Add to dashboard Cite

“…Spatio-temporal filters of this type are non-separable, and therefore require 2-D filter design techniques (e.g., Dudgeon et al [5]). The design can be performed in two steps: (i) specification of the ideal filter, and (ii) approximation by a realizable filter.…”

Section: Filtering Frequencies Vs Plane Wavesmentioning

confidence: 99%

Applications of short space-time fourier analysis in digital acoustics

Pinto

Vetterli

2011

2011 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)

View full text Add to dashboard Cite

This paper presents a signal processing tool for analyzing and manipulating digitized acoustic wave fields, based on a spatio-temporal extension of the time-frequency representation space. The emphasis is on wave fields acquired with a 1-D linear array of equidistant microphones (representing the spatial samples), but the basic formulation is valid for the three dimensions of space. We start by defining a spatio-temporal extension of the short time Fourier transform, in both continuous and discrete space and time, and then focus on applications of the proposed representation. In particular, we show that acoustic scenes with multiple sources can benefit from the use of space-frequency analysis in applications such as source separation (spatial filtering) and spatial audio coding (wave field coding). The experiments suggest that there is a spatial window size for which the performance of filtering and coding is optimal.

show abstract

“…Before we detail the complexity associated with the elements of the human-in-the-loop target acquisition process and how they influence modeling the target acquisition process, it is useful to briefly say why the human is in the target acquisition loop to begin with. Although research into automatic target recognition (ATR) and aided target recognition proceeds at a rapid pace, current ATR systems lack sufficient accuracy and flexibility to allow them to take over the process of target acquisition from humans (e.g., Dudgeon, 1998). The deficiencies of ATR become particularly apparent when they are called upon (a) to perform acquisition tasks when the space of possible targets is large, and (b) when non-visual factors such as situational context, experience, and judgment must be taken into account before an action is taken regarding a potential target.…”

Section: • Identificationmentioning

confidence: 99%

Soldier-in-the-Loop Target Acquisition Performance Prediction Through 2001: Integration of Perceptual and Cognitive Models

Vaughan¹

2006

View full text Add to dashboard Cite

Public reporting burden for this collection of information is estimated to average 1 hour per response, including the time for reviewing instructions, searching existing data sources, gathering and maintaining the data needed, and completing and reviewing the collection information. Send comments regarding this burden estimate or any other aspect of this collection of information, including suggestions for reducing the burden, ARL-TR-3833 SPONSOR/MONITOR'S ACRONYM(S) 9. SPONSORING/MONITORING AGENCY NAME(S) AND ADDRESS(ES) SPONSOR/MONITOR'S REPORT NUMBER(S) DISTRIBUTION/AVAILABILITY STATEMENTApproved for public release; distribution is unlimited. SUPPLEMENTARY NOTES ABSTRACTModeling Soldier-in-the-loop target acquisition performance is necessary for the development of improved sensors, more effective training methods, and better war game simulations. Accurately modeling this performance requires a detailed understanding of the environment, how a sensor responds to the environment, how it displays information to an observer, and how the observer employs that information to acquire a target. The first three requirements have been met; the fourth requirement, however, has not yet been achieved. Attempts to model the observer's visual and decision-making processes have been compromised by the analysis of the scene, based on physical parameters alone rather than how the visual system interprets the scene. Models based on such scene-derived factors have had limited success.This report takes a two-pronged approach to how future models can be improved by the sensible integration of human visual processing. One prong concerns basic research from the perceptual psychology community. Over the last few decades, this research has generated a detailed theoretical understanding of visual processing and decision making, based on visual information. The other prong concerns important models, modeling frameworks, and scene metrics from the military target acquisition community. Particular attention is paid to issues of clutter, the extendibility of the Johnson criteria, classical and neoclassical search frameworks, the selection of methods and performance metrics, and existing Night Vision and Electronic Sensors Directorate models. Issues related to the validation of target acquisition models are also discussed. iii Abstract (continued)Existing target acquisition models tend to base performance on (a) one-dimensional (1-D) metrics defining the amount of information in the target (e.g., resolvable bar cycles, contrast, area, size, perimeter, speed of motion) and how that information correlates to level of performance in a target acquisition task (i.e., detection, classification, recognition, and identification), (b) search processes that are unrealistic (e.g., that assume random eye movements), and (c) 1-D metrics to define the whole scene (clutter) or regions of the scene (e.g., clutter, conspicuity, attractiveness). These tendencies fail to account for known human behavior, although models incorporating them may be insensitive t...

show abstract

ATR Performance Modeling and Estimation

Cited by 30 publications

References 21 publications

Automated Target Recognition Using the Karhunen–Loéve Transform with Invariance

Automated Target Recognition Using the Karhunen–Loéve Transform with Invariance

Applications of short space-time fourier analysis in digital acoustics

Soldier-in-the-Loop Target Acquisition Performance Prediction Through 2001: Integration of Perceptual and Cognitive Models

Contact Info

Product

Resources

About