English

Abstract

The rising rate of Auto Teller Machine (ATM) robbery cases in Malaysia has drawn our attention. Despite having the CCTV surveillance system, the ATM robbery has been occurring in an alarming rate recently. One of the reason is the CCTV recorded video in ATM is limited to use as forensic evidence after crime occurrence while the video for serving as a real time alarm for the security guard (usually occupied with number of monitors in the control room) during robbery has been disregarded. Real time analysis of human motion from CCTV video data can be used to notify the security guard about a suspicious individual committing burglary and thus gives chance to immediately cease the criminal red-handed. In the process of human motion analysis, at first human must be detected in an image or video following three stages namely i) environment/background modeling, ii) motion segmentation and iii) object classification. In this paper we address existing human detection methods that can be effectively used in CCTV surveillance of ATM. Specifically; first we investigate the recent records of ATM robbery cases in all over the world including Europe, America and Asia. Next, we discuss the methods used in the three steps of human motion detection explicitly. The main purpose of the study is to provide a general overview of human motion detection methods used in ATM-CCTV surveillance system