Related Experiment Video
Updated: Aug 16, 2025
16:14
Trajectory Data Analyses for Pedestrian Space-time Activity Study
Published on: February 25, 2013
13.6K
Real-time pedestrian pose estimation, tracking and localization for social distancing.
Bilal Abdulrahman1, Zhigang Zhu2
1The Graduate Center, The City University of New York, New York, NY, 10016 USA.
Machine Vision and Applications
|December 19, 2022
Summary
This study introduces a new method using CCTV footage for real-time social distancing analysis. The system accurately detects and tracks pedestrians, improving public health monitoring and pandemic prevention.
Area of Science:
- Computer Vision
- Artificial Intelligence
- Public Health Technology
Background:
- The COVID-19 pandemic necessitated new public health measures, including social distancing.
- Existing visual surveillance systems offer potential for repurposing to monitor public health compliance.
- Challenges include real-time processing, accuracy in low-resolution footage, and handling occlusions.
Purpose of the Study:
- To propose a novel integration technique for real-time pose estimation and multiple human tracking for social distancing.
- To enhance processing speed and detection accuracy in low-resolution CCTV footage.
- To develop a system for measuring social distancing and detecting violations using existing surveillance infrastructure.
Main Methods:
- Integration of state-of-the-art object and pose estimation algorithms.
- Utilizing a bespoke version of the FASTMOT algorithm for pedestrian tracking.
- Employing temporal/analogous estimation for occlusions and projective geometry for localization.
- Implementing a "smart violations detector" to reduce false positives in group settings.
- Real-time implementation in Python with intuitive distance visualization.
Main Results:
- The proposed Pedestrian Pose Estimation, Tracking, and Localization for Social Distancing (PETL4SD) method achieves high processing speed.
- Demonstrated improved detection accuracy, particularly in low-resolution scenarios.
- Quantitative and qualitative validation on public domain datasets using single CCTV feeds.
- Outperformed baseline methods in both speed and accuracy.
Conclusions:
- The PETL4SD system offers an effective, real-time solution for monitoring social distancing using existing CCTV infrastructure.
- The novel integration technique significantly improves performance in challenging low-resolution environments.
- This approach can aid in managing COVID-19 and preventing future pandemics through enhanced public health surveillance.
