Advanced Automatic Target Recognition
Abstract
The grant supported research on techniques for recognizing targets in visible, infrared, laser radar, synthetic aperture radar and high-range aperture radar data. Empirical probability density functions of 'probe values' gives rise to a method of estimating probability that a target of known shape is present. Two-stage Constant False Alarm rate detectors were introduced to reduce the PA rate in the presence of SAR speckled images. A complete algorithm for wide-area site models using polarimetric SAR was developed. Multiscale methods were introduced into HRR classification with an improvement in performance. Operator-theoretic methods were successfully employed in the important area of detection and location of roads and nearby targets.
Document Details
- Document Type
- Technical Report
- Publication Date
- Sep 05, 1997
- Accession Number
- ADA335690
Entities
People
- Azriel Rosenfeld
Organizations
- University of Maryland