Neural net classifier for millimeter wave radar
Abstract
This paper describes the development of a neural net classifier for use in an automatic target recognition (ATR) system using millimeter wave (MMW) radar data. Two distinctive neural net classifiers were developed using mapping models (back-propagation and counterpropagation) and compared to a quadratic (Bayesian-like) classifier. A statistical feature set and a radar data set was used for both training and testing all three classifier systems. This statistical feature set is often used to test IMATRs prior to using actual data. Results are presented and indicate that the backpropagation net performed at near 100 percent accuracy for the statistical feature set and slightly outperformed the counterpropagation model in this application. Both networks hold promising results using real radar data.
- Publication:
-
Real-Time Signal Processing XII
- Pub Date:
- December 1989
- DOI:
- 10.1117/12.962373
- Bibcode:
- 1989SPIE.1154...71B
- Keywords:
-
- Millimeter Waves;
- Neural Nets;
- Pattern Recognition;
- Radar Data;
- Radar Targets;
- Target Acquisition;
- Algorithms;
- Clutter;
- Communications and Radar