Aerial Imagery-Based Validation of an Infrastructure-Based Highway Traffic Perception System

Abstract

There is an increasing need for robust validation methods for smart roads (road infrastructure integrated with sensors and communication technologies), digital twins (virtual replicas of traffic participants and infrastructure), and advanced driver assistance systems (ADAS) to ensure their reliability within intelligent transportation systems (ITS). Traditional validation methods, relying on sensor-equipped reference vehicles, are constrained by their high costs and limited scalability. To address this validation gap, we introduce DroneTruth, a lightweight, aerial imagery-based position estimation technology. Unlike traditional, resource-intensive setups, DroneTruth offers a comprehensive, robust, and cost-effective evaluation approach, capturing all traffic participants without occlusion. To demonstrate its capability, we performed a measurement campaign on a smart highway section near Budapest, successfully evaluating its precision, positional accuracy and tracking performance. First, we cross-validated DroneTruth against a reference vehicle equipped with a high-precision positioning unit. After that, we recorded the traffic on the relevant highway section via drone-mounted cameras. From the aerial footage, we generated reference trajectories and evaluated the accuracy of the digital twin model provided by the ITS.