Adaptive Automated Detection for Synthetic Aperture Sonar Images Using Seabed Classification
Abstract
In this work we have shown an initial method for using sensed environmental features to predict automatic target recognition (ATR) performance. Although we have applied this method to a cascade detector, the work can be extended to other detector/classifier methods such as a matched filter. As with all trained methods, generalization is the critical measure of performance. The utility of the trained method is in its ability to effectively detect targets in a previously unseen area. Using the data sets, we have shown the effective generalization performance of the trained cascade detector.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jan 01, 2014
- Accession Number
- AD1003590
Entities
People
- J. A. Fawcett
- W. A. Connors
Organizations
- Defence Research and Development Canada