Video surveillance: Anomaly action detection at front yard

Lee, Yong Jin (2024) Video surveillance: Anomaly action detection at front yard. Final Year Project, UTAR.

Preview

PDF Download (2855Kb) | Preview

Abstract

Surveillance system has become increasingly common as a safety measure to enhance the security of the houses and properties. Anomaly detection plays a vital role in such surveillance system because relying on human supervision will be a waste of time and labor force. Thus, a lot of efforts have been put into this field of study. This project proposed a novel fire detection strategy and implemented it into a workable system. The approach of this project differs from many of the general strategy of anomaly detection which is to use deep learning model to learn the structure and pattern of normal events. However, anomalies do not have a clear definition which is what makes anomaly detection a challenging task. In the context of front yard surveillance, anomalies could be loitering, fighting, explosion, arson, and other suspicious activities. Hence, in order to detect such anomalies more accurately, the focus of this study has been narrowed down to tackle the arson. The system has shown its capability to detect fire within 1 second and with high accuracy by only utilizing the motion information and brightness.

Actions (login required)