Metricam: Fast and Reliable Social Distancing Analysis in Online Security Cameras

Distance measurements taken from 2D camera images are subject to the correct estimation of the camera's perspective, that is, the spatial mapping from 2D points imaged by a camera to the correspondent 3D ones in the real world. Current solutions to solve this 3D reconstruction are either depend...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:2021 International Conference on Computational Science and Computational Intelligence (CSCI) S. 1567 - 1573
Hauptverfasser: Padilha, Bruno, Takecian, Pedro Losco, Lauro Soares, Decio, Schwerz, Andre Luis, Tadashi Kamaura, Leonardo, Ferreira, Joao Eduardo
Format: Tagungsbericht
Sprache:Englisch
Veröffentlicht: IEEE 01.12.2021
Schlagworte:
Cameras
Crowd Analysis
Deep Learning
Estimation
Human factors
Object Identification
Pocket Crowds
Social Distancing
Software algorithms
Software systems
Three-dimensional displays
Training
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:Distance measurements taken from 2D camera images are subject to the correct estimation of the camera's perspective, that is, the spatial mapping from 2D points imaged by a camera to the correspondent 3D ones in the real world. Current solutions to solve this 3D reconstruction are either dependent on the estimation of vanishing points through the detection of straight lines on targeted images or by employing sophisticated sensors and deep learning algorithms, which require expensive training on huge annotated datasets. Nevertheless, none of those approaches provide the required level of precision and accuracy for social distancing evaluation. In this paper we present Metricam, a real-time lightweight software system for security cameras that computes a 2D to 3D mapping using computational geometry and uses the DBSCAN clustering algorithm to evaluate social distancing evaluation. With Metricam, we have been able to identify several places prone to agglomeration inside the Butantã campus of the University of São Paulo, and provide the local authorities with valuable information to fight off the pandemic.
DOI:10.1109/CSCI54926.2021.00305