Automated Traffic Measurement with UAVs and Image Classification

Realistic traffic data is a critical factor in simulations, as aspects like density, flow, and temporal patterns heavily influence system performance. However, obtaining such data is often challenging, as real-world datasets typically provide only average values for specific road sections and timeframes. This limitation hinders the ability to test systems under diverse and dynamic conditions.

To address this, modern UAVs equipped with high-resolution cameras and advanced machine learning capabilities of consumer hardware offer a promising solution. These drones can capture real-time traffic data with sufficient accuracy, enabling the extraction of detailed traffic patterns. By leveraging image classification and automation, this data can be transformed into realistic traffic simulations, allowing for testing under varied scenarios (e.g., weekdays, holidays, peak hours).

Goal of the thesis

This thesis explores how to automate the process of evaluating camera data from UAVs, classifying traffic patterns, and generating SUMO simulations. By doing so, it aims to bridge the gap between real-world data and simulation environments, enabling more accurate and scalable testing of traffic management systems.

Requirements

• Experience with C++, machine learning

Contact

• J. Bigge
• C. Sommer