TRASE: A Traffic Surveillance System for Adaptive Speed Estimation of Vehicles from Aerial Videos

This research presents the development of an intelligent transportation system utilizing drone video to automatically track and identify speeding vehicles called TRASE. The system comprises multiple modules, including a YOLOv8-based detection module, a BoT-SORT-based tracking module, an adaptive speed estimation module, and a license plate recognition module. The project focuses on the design, implementation, and evaluation of these modules, with an emphasis on their performance and integration into a cohesive pipeline. We utilize prior information on road marking sizes to develop a calibration-free approach for speed estimation. Experimental results demonstrate promising performance in terms of detection accuracy, tracking reliability, speed estimation precision, and license plate recognition efficiency. Through the integration of automated algorithms and human-in-the-loop interventions, this system demonstrates its potential as a solution for traffic management authorities, offering improved decision-making capabilities and enhanced operational efficiency in the field of traffic management.