Target Classifier Strategies

doi:10.1117/3.2276891.ch3

Book: Automatic Target Recognition, Second Edition

Author(s): Bruce J. Schachter

Published: 2017

https://doi.org/10.1117/3.2276891.ch3

Author Affiliations +

Abstract

A target classifier receives image or signal data about a detection point. It infers the category of the object portrayed by the data. The classification decision can benefit from a host of other available information; the more information the better. ATR often involves a client–contractor relationship. The contractor is committed to providing a quality product to the customer. Yet, target classification is sometimes viewed in a naïve fashion. The customer throws data “over the fence.” The contractor is asked to classify the “targets.” Little thought is given to the breadth and scope of the problem. The usual “solution” involves showing that the contractor’s favorite classifier outperforms several alternatives. However, the true nature of the target classification problem is more complex. Ironically, choice of a classification paradigm may be the least important aspect of target classification. We will outline the issues involved in target classification. This will be followed by a review of a number of different types of classifiers.