عليرضا عالي نژاد

The intensification of congestion in metropolitan areas, resulting from the accelerated pace of urban development an‎d the increase in motor vehicles, has underscored the necessity for a precise analysis of the performance of signalized intersections. This matter holds particular significance for oversaturated intersections. Moreover, in Iran, the complexity of intersection analysis is amplified by the non-lane-based traffic behaviors of vehicles, incomplete adherence to regulations, an‎d disorderly traffic flow. The accessibility of aerial imagery via Unmanned Aerial Vehicles (UAVs), which allows for the simultaneous observation of all approaches leading to an intersection an‎d the recording of vehicle trajectories categorized by type, has presented an opportunity for a more comprehensive analysis compared to conventional data collection methods at intersections. In this research, by utilizing a UAV as a modern tool, an‎d while presenting a systematic framework for the automated extraction of vehicle trajectories an‎d other traffic indices, the performance of the Tohid signalized intersection in Tehran was examined. To this end, a computer vision-based process was developed, encompassing the stages of pre-processing, frame stabilization an‎d geo-referencing, vehicle detection an‎d tracking, an‎d the extraction of traffic variables. Accordingly, after extracting the trajectories of approximately 18,500 vehicles based on a trained object detection model with a mean Average Precision (mAP@0.5-0.95) of 0.902, key traffic performance metrics were calculated an‎d analyzed, disaggregated by approach an‎d vehicle type. These variables included signal timing, queue length, trajectory, speed, discharge flow rate an‎d deman‎d, saturation flow rate, delay, Level of Service (LOS), an‎d green time efficiency. The analysis of these factors provided a foundation for a comprehensive assessment of the intersectionʹs performance. The findings of this study indicate that UAV imagery can offer an efficient platform for the estimation of traffic variables an‎d the performance analysis of signalized intersections.

تقاطع چراغ‌دار , پهپاد , بينايي كامپيوتر , استخراج خودكار متغيرهاي ترافيكي

Signalized Intersection , Drone (UAV) , Computer Vision , Automated Extraction of Traffic Parameters

Alireza Aalinejad

Afshin Shariat Mahaymany