MLP baggage detector

Abstract

An artificial neural system (ANS) has been applied to the problem of discriminating between suitcases with and without explosives. The input to the ANS was data gathered during the field tests of a prototype explosive detection system. The performance of the ANS is contrasted with the standard statistical technique (discriminant analysis) used, and is shown to exceed the performance of the standard technique over a substantial range. The system that generated the data, the nature of the data, the basics of discriminant analysis, and the technique used in developing the network are described.