Document Details

Document Type : Thesis 
Document Title :
Real-time Computer Vision-based Crowd Monitoring System on NVIDIA Jetson GPU using High-Level GPU Coder
نظام مراقبة الحشود المستند إلى رؤية الكمبيوتر في الوقت الفعلي على وحدة معالجة الرسومات NVIDIA Jetson باستخدام وحدة تشفير GPU عالية المستوى
 
Subject : Faculty of Engineering 
Document Language : Arabic 
Abstract : Crowd monitoring in Saudi Arabia is an essential task, especially in the Two Holy Mosques and the holy sites. During this scientific research, crowd flow monitoring system was designed and developed using image analysis of live CCTV camera feed in these places. The system automatically monitors crowd flow such as collecting similar flow patterns in regions i.e. segmenting the crowd based on the flow which in turn is calculated using object displacement calculations made at the pixel and object level. Various computer vision-based algorithms have been proposed in the literature for visual flow estimation that high-level algorithms use to build a larger picture of crowd movement patterns. The Lucas-Kanade and Horn-Schunck methods were used. It is widely known that pixel-level accurate optical flow with reasonable accuracy requires tremendous computational power. Furthermore, the algorithm should further estimate crowd flow patterns by aggregating the motion vectors obtained by optical flow. So, the overall system is computationally expensive and requires a lot of computing power. To achieve this, we will move this task to a GPU that can handle vector computations much more quickly than CPUs. For this purpose, we intend to use a MATLAB GPU encoder to implement the optimization algorithm for optical flow. This development environment is suitable for rapid algorithm deployment on GPU hardware. The solution that was developed to help the current crowd management system in the holy places and will not only enhance the local capabilities of the Kingdom towards solutions for crowd management in the holy places but also serve the development plan in the Kingdom. Moreover, the portability of the solution developed on the embedded GPU will allow it to be installed quickly in temporary security locations as well. 
Supervisor : Dr. Muhammad Ahmad Bilal 
Thesis Type : Master Thesis 
Publishing Year : 1444 AH
2022 AD 
Added Date : Tuesday, February 21, 2023 

Researchers

Researcher Name (Arabic)Researcher Name (English)Researcher TypeDr GradeEmail
بندر حنش القرنيAl-Qarni, Bander ResearcherMaster 

Files

File NameTypeDescription
 49003.pdf pdf 

Back To Researches Page