Smart Prison - Video Analysis for Human Action Detection

Abstract

Prisons or correctional facilities require a high amount of manpower for maintaining safety whilst ensuring service quality. However, high turn-over rate (5% per year) commonly seen in these institutions often brings operational challenges, which may entail risk to safety. Prisoner’s abnormal behaviors, such as self-harming and fighting, are the major concerns posing the highest safety hazards to prisoners and front-end staff. Traditional human-dependent method used for monitoring prisoners’ abnormal behaviors is labor-intensive, and may give rise to problem such as misdetection. A Smart Prison system has been developed to assist front-end staff in detecting prisoners’ abnormal behaviors. Results shows that over 95% of the abnormal behaviors can be detected by the system. This supporting system can lower the operational pressure resulted from shortage of manpower, and reduce the rate of abnormal behavior misdetection. This will improve the safety of both front-end staff and prisoners.