A biometric-based system for unsupervised anomaly behaviour detection at the pawn shop

Abstract

ABSTRACT This article shows a system performing re-identification and description of people entering different stores of the same franchise by means of Face Recognition, Gait Analysis, and Soft Biometrics techniques. Additionally, an anomaly detection analysis is conducted to identify suspicious behavioral patterns.It has been tested on an ad-hoc dataset of a set of pawn shops of a local franchise.The registered users paths have been human labelled as ‘normal’ or ‘abnormal’ achieving a precision of 100%, recall of 72.72%, and an average accuracy of 96.39%.The system is able to report anomalies to support decisions in a context of a security monitoring system..