Crowd density estimation using deep learning for Hajj pilgrimage video analytics.

Area of Science:

• Computer Vision
• Artificial Intelligence
• Crowd Dynamics

Background:

• High-density crowd management is critical for public safety, especially during large-scale events such as the Hajj pilgrimage in Makkah.
• Traditional video analysis methods struggle with the scale and density of Hajj crowds.
• Advanced computer vision and surveillance are increasingly vital for pilgrimage security.

Purpose of the Study:

• To develop an advanced crowd counting and density estimation algorithm tailored for high-density scenarios.
• To enhance the safety and security of pilgrims through precise crowd analysis.
• To introduce a novel dataset specifically for evaluating crowd estimation models in Hajj-like conditions.

Main Methods:

• A Convolutional Neural Networks (CNN) model architecture was developed for detecting individuals and estimating crowd density.
• The model utilizes head detection with bounding boxes for accurate person counting.
• A new dataset, HAJJ-Crowd, was created for training and evaluating the proposed algorithm.

Main Results:

• The proposed CNN algorithm demonstrated superior performance compared to state-of-the-art methods.
• Achieved a Mean Absolute Error of 200, showing an average improvement of 82.0.
• Attained a Mean Square Error of 240, with an average improvement of 135.54.

Conclusions:

• The developed algorithm offers a significant advancement in crowd density estimation for challenging environments.
• The HAJJ-Crowd dataset provides a valuable resource for future research in crowd analysis.
• The findings contribute to improved safety measures for large-scale gatherings through enhanced video analytics.