Deep RL-based Abnormal Behavior Detection and Prevention in Network Video Surveillance.

An important source of information for ensuring public safety is control room video surveillance. A DecisionSupport System (DSS) designed for the security task force is vital and necessary to take decisions rapidly using a sea of information. In case of mission-critical operation, Situational Awareness (SA) which consists of being aware of what is going on around you in real time plays a crucial role across a variety of industries and should be placed at the center of our system. In this paper, in order to satisfy the understanding and projection levels of SA in real time, we propose a method based on Reinforcement Learning (RL) using an Actor-Critic algorithm. This algorithm permits our DSS to first, keep SA knowledge up to date by performing online adaptive learning and second, develop a long-term vision and strategy by exploiting a non-myopic agent for real-time sequential decision-making. In contrast to other approaches, our results permit demonstrating the practical potential of our method in real-world scenarios by preventing abnormal situations in an ever-changing environment.