COVID-19 contact tracking based on person reidentification and geospatial data

Area of Science:

• Public Health
• Computer Science
• Geospatial Analysis

Background:

• Current COVID-19 contact tracing methods are limited by manual processes, self-reporting accuracy, and privacy concerns with digital tools.
• Existing mobile and Bluetooth-based tracing solutions face challenges due to privacy issues and reliance on personal data.

Purpose of the Study:

• To propose a privacy-preserving, efficient contact tracing method using geospatial big data.
• To overcome the limitations of traditional and digital contact tracing approaches.

Main Methods:

• Developed a real-time person reidentification model for tracking individuals across multiple surveillance cameras.
• Integrated surveillance data with geographic information and mapped it onto a 3D geospatial model to reconstruct movement trajectories.
• Utilized big data analytics for geospatial mapping and trajectory analysis.

Main Results:

• Achieved a first-accuracy rate of 91.56% and a first-five accuracy rate of 97.70% in person reidentification.
• Obtained a mean average precision of 78.03% with a fast inference speed of 13 ms per image.
• Demonstrated the method's effectiveness through real-world verification.

Conclusions:

• The proposed geospatial big data method offers an effective and privacy-conscious alternative for contact tracing.
• This approach avoids reliance on personal information, mobile phones, or wearable devices, addressing key limitations of current methods.
• The findings have significant implications for public health strategies, particularly in the post-COVID-19 era.