Database Generation for Non-Cooperative Air Target Identification

Summary form only given. In this paper we will discuss the approach that we have evolved for populating a database with representative training data for the purpose of non-cooperative target identification (NCTI) of air targets by radar. Such NCTI should help to augment procedures like IFF to reduce the risk of fratricide and allow early designation of targets such that threats can be engaged before significant risk to vulnerable assets. This paper will focus on the use of synthetic data generated from limited information, such as photographs, to represent target types for which we don't have access to real radar measurements over any or a sufficiently representative range of aspects and configurations. This can often be the case for hostile aircraft types. Subscale models and computer electromagnetic predictions are both common methods that researchers have explored to simulate the radar returns from real aircraft. Both have advantages and disadvantages but the former is thought to be a feasible approach with current technology. We will discuss the methodology required for subscale model generation and subsequent use of the data to populate classifier databases and how to quantify the fidelity of the data such that confidence in classification result can be quantified and thus classifier performance as a whole estimated