Loading...
Drone-Based AI System for Real-Time Hazard Detection and Crowd Safety During Hajj and Umrah
Turkistani, Nareman Alamoudi, Reem Alamoudi, Reham
Turkistani, Nareman
Alamoudi, Reem
Alamoudi, Reham
Citations
Altmetric:
Type
Supervisor
Subject
Date
Collections
Research Projects
Organizational Units
Journal Issue
Abstract
Mass gatherings like Hajj and Umrah, which draw millions of pilgrims annually to Mecca, present complex
safety challenges due to extreme crowd density, high temperatures, and rapidly evolving risk environ ments. Traditional surveillance methods, like CCTV and manual monitoring, are generally inadequate
for real-time response. In this report, we present the development and evaluation of an AI-powered
drone system selected to detect threats and improve crowd safety during Hajj and Umrah. Integrat ing drone-based live monitoring and deep learning algorithms is proposed to identify critical threats
like overcrowding, associated heat stress, and medical emergencies. We developed a custom application
to perform real-time hazard classification, emergency alerting, and centralized data visualization. We
trained an augmented dataset of Hajj-related images with real and synthetic images using the deep
learning model that includes YOLOv5 for object detection, U-Net for segmentation, and ResNet50 for
severity analysis in the developed Android application. With edge deployment using Raspberry Pi and
NVIDIA Jetson Nano, the system operated for its accuracy promise in low-connectivity, high-density en vironments, with a latency of less than two seconds. Experimental results demonstrated high accuracy:
Fire hazards, crowd congestion, and medical emergencies were 92.3%, 88.7%, and 90.1%, respectively.
It consists of a scalable backend, a mobile dashboard, and an alert notification system for ease of use
and operational reliability for emergency responders and drone operators. The project fills significant
gaps in existing surveillance systems by adding real-time responsiveness, AI-based prediction, and inte gration with ground crews. This system aligns with Saudi Vision 2030’s aspiration to improve the safety
of pilgrimage and achieves a new benchmark in large-scale event safety management. Further research
includes extending hazard categories, including swarm drone intelligence, and increasing compatibility
between platforms for deployment on a larger scale internationally at mass gatherings.