CrowdSync: An IoT and Computer Vision-Based Framework for Real-Time Bus Occupancy Monitoring and Passenger Tracking

CrowdSync demonstrates that a fully functional, real-time bus occupancy monitoring and passenger tracking system can be built using open-source technologies at a cost that is actually viable for Indian municipal transport operators. The combination of edge-deployed computer vision, geospatial cloud infrastructure, and purpose-built mobile and web interfaces creates a complete platform that addresses the three core problems — passenger uncertainty, reactive fleet management, and absence of planning data — in a single, integrated solution. The experimental results confirm that the system achieves over 93% counting accuracy under typical operating conditions, delivers real-time updates to all connected clients within 5 seconds, and triggers overcrowding alerts within 2 seconds of a threshold breach. These numbers are not just technically satisfying — they are operationally meaningful. An alert delivered in 2 seconds allows a dispatcher to respond before a situation worsens, and occupancy information delivered to a commuter 4 minutes before arrival gives them a real, actionable choice. The system's privacy-first design — where video is processed entirely on the edge and never leaves the bus — ensures that deployment does not require navigating complex data protection concerns around biometric or surveillance data. Future work will focus on upgrading the detection model to YOLOv8 for improved accuracy under challenging conditions, integrating predictive demand forecasting using LSTM models trained on historical occupancy patterns, and scaling the backend to a Kubernetes-based deployment for city-wide fleet management. Integration with digital ticketing and automated fare collection systems is also planned to expand CrowdSync into a unified smart transit platform.

[1] N. Wojke, A. Bewley, D. Paulus, "Simple Online and Realtime Tracking with a Deep Association Metric (DeepSORT)," IEEE ICIP, 2017. [2] M. Imran, H. Hassan, M. Hasan, "A Review of Travel Time and Arrival-Time Prediction Methods on Road Networks," Applied Sciences, vol. 11, no. 22, 2021. [3] Y. Zhang, H. Liu, Q. Qiu, "Real-Time Transit Information Systems and Passenger Satisfaction in Urban Networks," Transportation Research Part C, 2019. [4] OpenRouteService API Documentation, Heidelberg Institute for Geoinformation Technology, 2024. [Online]. Available: https://openrouteservice.org/ [5] PostGIS Documentation — Spatial and Geographic Objects for PostgreSQL, 2024. [Online]. Available: https://postgis.net/ [6] Expo Documentation — React Native Development Framework, 2024. [Online]. Available: https://docs.expo.dev/ [7] V. Mandal, L. Adu-Gyamfi, and U. Uddin, "Low-Cost IoT-Based Passenger Counting System Using Infrared Sensors for Public Transportation," IEEE Sensors Journal, vol. 21, no. 14, pp. 16183–16192, 2021. [8] J. A. Guerrero-Ibanez, S. Zeadally, and J. Contreras-Castillo, "IoT-Based Intelligent Transportation Systems: A Survey of Architectures, Protocols, and Applications," IEEE Internet of Things Journal, vol. 9, no. 10, pp. 7508–7527, 2022. [9] S. Kang, J. Park, and H. Kim, "Crowd Density Estimation and People Counting in Public Spaces Using Deep Learning," IEEE Access, vol. 8, pp. 141892–141903, 2020. [10] J. Wen, H. Liu, and T. Zhang, "Benchmarking YOLOv5 and MobileNet-SSD for Pedestrian Detection on Edge Devices in Transit Environments," IEEE Transactions on Intelligent Vehicles, vol. 8, no. 2, pp. 1923–1934, 2023.