This project has been created as part of my university senior project requirements.

Academic Advisor: Dr. Souhila Nada

April 2017

• Problem Statement
• Project Scope
  • Fact Finding Technique
• Proposed System
  • Corwd Monitoring System Interface (Prototype)
• Implementation
  • System Results
• Conclusion
• References
  • Appendix

Managing the crowds in big venues and events has been an issue for many decades for organizers and officials. Crowds gathering in one place or events such as sporting even or religious rituals like at Hajj pilgrimage gathering, can create crushes and stampedes, whether triggered by natural disaster or misguided crowd managers.

In this project, machine learning technology will be implemented by developing a software system that uses the computer vision technique to understand and monitor the live footage of the crowds at Hajj pilgrimage gathering, providing the best possible solution when a stampede might happen. The main purpose of this project is to help securing the Hajj sites and ease the rituals for all pilgrims.

After conducting a comprehensive research, the study discovered that using a crowd vision system for Hajj Pilgrimage is significantly more advanced than the current methods used in Hajj recently (dated 2017). Therefore, problems happening with the current methods such as unregistered pilgrims, able to be controlled only in a single environment (The Holy Mosque) and the flaw in the planning and management of clearing areas may overflow the crowd stampede. Nevertheless, computer vision technique can assess in securing the Hajj by monitoring the crowds and analyzing their movements, saving time of years of planning and being able to react immediately when a stampede could happens.

Based on the research, and the conclusion of adopting vision based solution, the system can be split into 3 phases:

• Crowd density and dynamics analysis from video streams
• Estimation of threshold value for normalization
• Gate security system for crowd flow regulation

The proposed system will consist of the following component:

1. Closed Circuit Television (CCTV) Cameras
2. Server
3. Monitors that processed the footage feed
4. Smartphone application (Android)

The below is a prototype of the interface of the Crowd Monitoring software system:

The Monitoring Crowd System is implemented using Python 2.7

In addition, multiple packages and libraries has been used such as OpenCV and NumPy as detailed below:

• Numpy
• OpenCV
• Optical Flow
• The Lucas-Kanade Algorithm

Here’s a screenshot of a video at Mina, Makkah where a stampede is happening before running the system:

Following are the results when applying the parameters of the (image) criteria: The system is detecting people with a red arrow above each person in the scene.

The system then will split the video into four parts to better understand the video and crowd flow density, indicating a stampede is happening when one fourth of a video pop up on the left side of the window:

The stampede at Mina, Hajj is a recent example of a terrible crowd disaster where, in spite of all precautions, many people died during a mass event. Pedestrian dynamics have been studied intensively for more than four decades. Multiple researches have been conducting in studying the computer vision for detection and tracking pedestrian in the crowded scenes. In conclusion, the demonstration proves the concept of monitoring crowds using computer vision technique which allows to track people using the optical flow and to detect if there’s a stampeding situation happening. An effective proven results is shown of the algorithm implemented on different stampede scenarios.

1. Incidents during Hajj https://en.wikipedia.org/wiki/Incidents_during_the_Hajj
2. 2015 Mina Stampede https://en.wikipedia.org/wiki/2015_Mina_stampede
3. Sharley Kulkarni, PG Student, Dept. of E&TC, and S. K. Shah HOD PG, STES's Smt. Kashibai Navale College of Engineering, Pune, Maharashtra, India, Monitoring and Safety of Pilgrims Using Stampede Detection and Pilgrim Tracking, International Journal of Advanced Research in Electrical, Electronics and Instrumentation Engineering, 2015.
4. Almoaid A. Owaidah, Thesis, Hajj Crowd Management via a Mobile Augmented Reality Application: A case of The Hajj event, Saudi Arabia, University of Glasgow 2014.
5. Mantoro T, Hajj Locator: A Hajj pilgrimage tracking framework in crowded ubiquitous environment, Multimedia Computing and Systems (ICMCS), 2011 International Conference, April 2011.

• Fishbone Diagram https://realtimeboard.com
• Crowd Monitoring System UI (Prototype) https://www.sketch.com/