Road traffic injuries claim an estimated 1.19 million people per year, a crisis that is mostly dominant in developing countries such as Zimbabwe due to high speeding and limited speed regulation enforcement hardware. Traditional speed regulation methods such as fixed speed cameras and roadblocks suffer from predictability, creating significant coverage gaps. The study introduces the Intelligent Aerial Traffic Observation System (IATOS), a solution to make use of Unmanned Aerial Vehicles (UAVs) and computer vision to automate speed enforcement. The study employed a hybrid methodology, integrating Design Science Research and quantitative experimental analysis; the system utilises edge-cloud hybrid architecture where an ESP32-CAM handles low-power image acquisition and a cloud backend performs heavy processing using You Only Look Once (YOLOv8) for object detection and DeepSORT for multi-object tracking. The system utilises a homography technique, which projects the 2D aerial video onto the road plane to calculate speed with high precision. Experimental results demonstrate a vehicle detection mean Average Precision (mAP) of 0.92 and a speed estimation accuracy of ±2.0 km/h compared to GPS ground truth. The system automatically generates standardised digital evidence packs, containing speed data, GPS location, and visual proof, for every confirmed violation.

Keywords

UAV Traffic Monitoring, Computer Vision, Real-time Speed Detection, Automated Citation, Intelligent Transport Systems