Face mask identification with enhanced cuckoo optimization and deep learning-based faster regional neural network.

Area of Science:

• Computer Vision
• Artificial Intelligence
• Public Health Surveillance

Background:

• Uncontrolled outdoor environments pose challenges for accurate face mask detection from aerial imagery.
• Existing surveillance methods may lack the dynamic capabilities for real-time monitoring of public health guidelines.

Purpose of the Study:

• To develop and evaluate an Unmanned Aerial Vehicle (UAV)-based system for identifying face mask usage and monitoring social distancing.
• To enhance image quality from UAVs for improved detection accuracy in outdoor settings.

Main Methods:

• Image preprocessing techniques including grayscale conversion and contrast enhancement (Optimum Wavelet-Based Masking, Enhanced Cuckoo Methodology).
• Feature extraction using Gabor-Transform (GT) and Stroke Width Transform (SWT).
• Classification using Weighted Naive Bayes (WNBC) and a Faster Region-Based Convolutional Neural Networks (R-CNN) with Adaptive Galactic Swarm Optimization (AGSO) for mask detection and social distance monitoring.

Main Results:

• The system successfully identifies appropriate and incorrect face mask wear and monitors social distancing in crowded areas.
• Alerts are sent to medical personnel and police for unmasked individuals, and automated audio alerts promote social spacing.
• The methodology achieved high performance metrics, validated using a large dataset and 10-fold cross-validation.

Conclusions:

• The proposed UAV-based system offers an effective solution for public health surveillance, particularly in monitoring face mask compliance and social distancing.
• The integration of advanced image processing and deep learning algorithms enhances detection accuracy and system efficacy in real-world outdoor conditions.