Study of masked face detection approach in video analytics

Abstract

Security being of utmost importance, video surveillance has become an active research topic. Video analytics enhance video surveillance systems by performing tasks of real time event detection and post-event analysis. This can save human resources, cost and increase the effectiveness of the surveillance system operation. One of the common requirements of Video Analytics for security is to detect presence of a masked person automatically. In this paper, we propose a technique for masked face detection using four different steps of estimating distance from camera, eye line detection, facial part detection and eye detection. The paper outlines the principles used in each of these steps and the use of commonly available algorithms of people detection and face detection. This unique approach for the problem has created a method simpler in complexity thereby making real time implementation feasible. Analysis of the algorithm's performance on test video sequences gives useful insights to further improvements in the masked face detection performance.