Image-based Fraud Detection in Automatic Teller Machine

Summary Some criminals access ATM(Automatic Teller Machine) using other’s bank card by disguising themselves with masks, hats, and sunglasses. This paper proposes a novel and efficient method to detect such activities by analyzing the video from the camera mounted inside ATM. We first extract moving objects based on MOG(Mixture of Gaussians) background models and detect face region in HSV color space. We then obtain facial features by detecting face boundary, suppressing the outlier regions(nonfacial features) inside the boundary, and analyzing projection peaks of survived regions. The presence of fraud is determined based on the relative location of features detected..