Face Mask Surveillance Using Mobile Robot Equipped with an Omnidirectional Camera

Detecting humans in images not only provides vital data for a wide array of applications in intelligent systems but also allows for the classification of specific groups of individuals for authorization through various methods based on several examples. This paper presents a novel approach to classify persons wearing a face mask as an example. The system utilizes an omnidirectional camera on the mobile robot. This choice is driven by the camera's ability to capture a complete 360 degrees scene in a single shot, enabling the system to gather a wide range of information within its operational environment. Our system classifies persons using a deep learning model by gathering information from the equirectangular panoramic images, estimating a person's position, and computing robot path planning without using any distance sensors. In the proposed method, the robot can classify two groups of persons: those facing the camera but without face masks and those not facing the camera. In both cases, the robot approaches the persons, inspects their face masks, and issues warnings on its screen. The evaluation experiments are designed to validate our system performance in a static indoor setting. The results indicate that our suggested method can successfully classify persons in both cases while approaching them.