Translator Disclaimer
29 September 1999 Feature development, selection, and minimization for classification of HRR targets
Author Affiliations +
In automatic/aided target recognition (ATR) applications using high resolution radar (HRR) range profiles (RP), storing and processing large data sets to make a target decision is faster and less costly when compression is used. In this research we focus on feature development, selection, and minimization that reduce data dimensionality and improve target identifiability using a Bayesian quadratic classifier. We explore the development and application of features based primarily on scattering centers represented in HRR range profiles for each of 5 military aircraft targets. Performance of the ATR system improved as optimum features were added producing a probability of correct classification of greater than 99% for 2 training sets of RP using no more than 22 features. Ninety-six percent or more RP were still correctly classified when the system was trained on one set of data and then classified non-training data. When an unknown target was added to either data set, 90% or greater of the range profiles were correctly declared 27.6% unknown RPs were correctly classified as unknown. Compression reduced data dimensionality by a factor of at least 36 while preserving or improving target detection and classification potential.
© (1999) COPYRIGHT Society of Photo-Optical Instrumentation Engineers (SPIE). Downloading of the abstract is permitted for personal use only.
Edmund A. Quincy, Robert B. Stafford, and Tom Chau "Feature development, selection, and minimization for classification of HRR targets", Proc. SPIE 3810, Radar Processing, Technology, and Applications IV, (29 September 1999);

Back to Top