Pattern-Theoretic Foundations of Automatic Target Recognition in Clutter
Abstract
This effort advanced the art of applying Grenander's pattern theory to automatic target recognition (ATR) problems. We extended jump-diffusion ATR algorithms to accommodate unknown infrared camera calibration effects and include more numerically stable diffusion procedures for pose refinement, and developed flexible shape models to accommodate clutter. We also developed performance bounds on estimation and recognition performance for low-frequency radar data, single-image laser radar data, and 3-D "point cloud" data assembled from multiple sources. Further work explored data fusion using the "probability hypothesis density" approach.
Document Details
- Document Type
- Technical Report
- Publication Date
- Nov 30, 2006
- Accession Number
- ADA469413
Entities
People
- Aaron Lanterman
Organizations
- Georgia Tech