Development of Scale and Rotation Invariant Neural Network based Technique for Detection of Dielectric Contrast Concealed Targets with Millimeter Wave System

Abstract

The detection of concealed targets beneath a person’s clothing from standoff distance is an important task for protection and the security of a person in a crowded place like shopping malls, airports and playground stadium, etc. The detection capability of the concealed weapon depends on a lot of factors likes, a collection of back scattered data, dielectric property and a thickness of covering cloths, the hidden object, standoff distance and the probability of false alarm owing to objectionable substances. Though active millimeter wave systems have used to detect weapons under cloths, but still more attention is required to detect the target likes a gun, knife, and matchbox. To observe such problems, active V-band (59 GHz- 61 GHz) MMW radar with the help of artificial neural network (ANN) has been demonstrated for non-metallic as well as metallic concealed target detection. To validate ANN, the signature of predefined targets is matched with the signature of validated data with the help of the correlation coefficient. The proposed technique has good capability to distinguish concealed targets under various cloths.