Anomaly Detection at ATM Center using Machine Learning Algorithm

Abstract

This paper proposes a new supervised algorithm for detecting abnormal events in confined areas like ATM room, server room etc. The aim of proposed work is to establish a technical base that will support a more secure and convenient social infrastructure, and one of the technologies that make up the technical base in the abnormal behavior detection using image processing. Generally, abnormal behavior detection is a method in which a model is created using normal behavior data and any behavior deviating from the model is deemed abnormal. In many cases, it is difficult to comprehensively collect abnormal behavior data in advance, thus being able to detect abnormalities with a model created using only normal behavior data is extremely useful for actual implementation. This article first shows application examples of abnormal behavior detection using image processing, which is followed by typical examples of abnormal behavior detection through motion image processing.