28 July 1997 Residual-based adaptive modeling for data trajectory realignment to improve model-based ATR performance
Author Affiliations +
Abstract
The Wright Lab Combat Information Division's system-oriented HRR automatic recognition program plans to develop and mature advanced air-to-ground high range resolution (HRR) automatic target recognition (ATR) capabilities for transition into suitable operational Air Force airborne platforms. ATR technologies have historically been template and feature based and generally provide good performance but at the expense of robustness. However, because of the need for robust ATR performance in an operational scenario, model-based ATR is being considered for possible future transition. This paper presents preliminary results on research dealing with what is commonly referred to as the 'training on synthetic - testing on measured' problem (TOS- TOM) in which algorithm trained on synthetic signatures often do not perform well when tested on measured data. The data trajectory realignment approach proposed in this paper makes the claim that poor TOS-TOM performance is the results of sufficient differences between the trajectories of synthetic data nd their measured counterpart. An adaptive model was developed which captures the trajectory relationship between synthetic and measured data and adaptively estimates the model parameters needed for realignment. Surrogate data sets were used for the preliminary proof of concept experiments and initial 2 class results showed more than 20 percent increase in probability of correct classification.
© (1997) COPYRIGHT Society of Photo-Optical Instrumentation Engineers (SPIE). Downloading of the abstract is permitted for personal use only.
Robert L. Williams, "Residual-based adaptive modeling for data trajectory realignment to improve model-based ATR performance", Proc. SPIE 3070, Algorithms for Synthetic Aperture Radar Imagery IV, (28 July 1997); doi: 10.1117/12.281571; https://doi.org/10.1117/12.281571
PROCEEDINGS
9 PAGES


SHARE
Back to Top