Smart Mobility Sensing of Passenger Origin Destination Tracking

Understanding passenger movement within public transportation systems is essential for efficient transit planning, especially in resource limited communities. Traditional methods for measuring origin destination flows, such as manual driver logs, are labor intensive and often inaccurate. This study proposes an automated computer vision based system for accurately detecting and tracking OD pairs. Combining advanced modules like passenger detection, tracking, and re identification with GPS data, the framework offers a robust privacy preserving solution tailored to areas where expensive sensing systems or smart card technologies are not feasible. It highlights how low cost scalable tools can support equitable transportation planning. Using deep learning models and edge computing, the system achieves high accuracy under real world conditions, overcoming challenges such as occlusion and pose variation. The approach adapts existing algorithms to localized constraints and deploys them effectively within resource limited transit systems. Results show that the system improves OD data precision and provides actionable insights that support route optimization and enhance overall system performance. (Master Thesis)