A Social Distance Estimation and Crowd Monitoring System for Surveillance Cameras

Abstract

Social distancing is crucial to restrain the spread of diseases such as COVID-19, but complete adherence to safety guidelines is not guaranteed. Monitoring social distancing through mass surveillance is paramount to develop appropriate mitigation plans and exit strategies. Nevertheless, it is a labor-intensive task that is prone to human error and tainted with plausible breaches of privacy. This paper presents a privacy-preserving adaptive social distance estimation and crowd monitoring solution for camera surveillance systems. We develop a novel person localization strategy through pose estimation, build a privacy-preserving adaptive smoothing and tracking model to mitigate occlusions and noisy/missing measurements, compute inter-personal distances in the real-world coordinates, detect social distance infractions, and identify overcrowded regions in a scene. Performance evaluation is carried out by testing the system’s ability in person detection, localization, density estimation, anomaly recognition, and high-risk areas identification. We compare the proposed system to the latest techniques and examine the performance gain delivered by the localization and smoothing/tracking algorithms. Experimental results indicate a considerable improvement, across different metrics, when utilizing the developed system. In addition, they show its potential and functionality for applications other than social distancing.

Deep visual social distancing monitoring to combat COVID-19: A comprehensive survey

“…However, the complexity of VSDM systems (e.g., detecting all mutual distances) increases with the increased density of monitored crowds (i.e., the rise in the number of observed people). For instance, the computational complexity of the VSDM system introduced by Al-Sa’d et al (2022) has been measured by its frame rate (the number of processed video frames per second) and processing rate (i.e., the amount of processing time per frame). This VSDM system includes (i) person detection and localization, (ii) top-view transformation, (iii) smoothing/tracking (smooth noisy top-view positions and compensate for missing data due to occlusion with tracking), (iv) distance measurement and (v) violation detection.…”

“…In this regard, the VSDM task problem has been transformed into a sphere collision problem. In Al-Sa’d et al (2022) , a VSDM and crowd management system is introduced, which is based on (i) detecting pedestrians using a global nearest neighbor tracking (GNN), which is a real-time light-weight MOT approach (based on allocating detection/prediction annotations to tracks, (ii) and preserving their track records), (ii) filtering region of interest (ROI), (iii) transforming video frames into a top-View, (iv) tracking and smoothing, (v) estimating parameters, and (vi) detecting (SD) violations. In Aghaei et al (2021) , a semi-automatic VSDM approach is proposed for approximating the homography matrices between the image plan and scene ground.…”

“…In Al-Sa’d et al (2022) , EPFL-MPV ( Fleuret, Berclaz, Lengagne, & Fua, 2007 ), EPFL-Wildtrack ( Chavdarova et al, 2018 ), and OTC ( Benfold & Reid, 2011 ) datasets are used to evaluate DL models. The EPFL-MPV comprises four video sequences of six individuals freely moving in a room.…”

“…Moreover, multi-object tracking (MOT) algorithms have also been considered to track detected pedestrians. This is the case of Al-Sa’d et al (2022) , where the global nearest neighbor (GNN) tracking technique has been used.…”

“…Fixed surveillance is mainly used in indoor environments, such as shopping areas, airports, sports facilities (e.g., stadiums), etc. ( Al-Sa’d et al, 2022 ).…”