Bayesian Techniques in Automatic Target Recognition (ATR)

In the past decade, new techniques on how to achieve robust classification performance have emerged from the machine learning community. These classification methods are based on a kernel similarity metric and achieve a balance in what is known as the bias variance trade-off. Examples of these techniques are the kernel principal component analysis, Support Vector Machine (SVM), and the Relevance Vector Machine (RVM). Signal Innovations Group, Inc. has built a reputation for successfully applying kernel based methods to meet our customer’s performance criteria. We are able to build classifiers based on supervised and semi-supervised learning to provide robust classification performance, but do not suffer the over training and memorization problems association with other types of classification techniques.