Classification of metallic targets using a single frequency component of the magnetic polarisability tensor

7th Sensors & their Applications

Makkonen, J.^1 ; Marsh, L.A.^2 ; Vihonen, J.^1 ; Visa, A.^1 ; Järvi, A.^3 ; Peyton, A.J.^2

Tampere University of Technology, Department of Signal Processing, Korkeakoulunkatu 10, FIN-33101 Tampere, Finland^1

School of Electrical and Electronic Engineering, University of Manchester, Manchester M13 9PL, United Kingdom^2

Rapiscan Systems Oy, Klovinpellontie 3, Torni 2, FIN-02180 Espoo, Finland^3

关键词: Classification accuracy;    Classification algorithm;    K nearest neighbours (k-NN);    Library data;    Metal detection systems;    Metallic targets;    Polarisability;    Single frequency;   

Others  :  https://iopscience.iop.org/article/10.1088/1742-6596/450/1/012038/pdf
DOI  :  10.1088/1742-6596/450/1/012038

来源: IOP

【 摘 要 】

A k-nearest neighbour (KNN) classification algorithm has been added to a walk-through metal detection system which is capable of inverting the magnetic polarisability tensor of metallic targets at a frequency of 10 kHz. Pre-computed library data is used to determine the class of the object, e.g. 'knife' or 'mobile phone', and is consequently capable of determining if an object is considered a threat. The results presented show a typical success rate of 95%. An investigation into classification accuracy between different candidates is also presented to determine the significance of the body effect on the success of the classification.