Since its foundation in 2019, the Institute for the Protection of Terrestrial Infrastructures has been dedicated to the protection and security of critical infrastructures on Earth. We develop concepts, processes and technologies that strengthen and improve the resilience of organisations and systems. In doing so, we focus on people, technology and the system as a whole.

What to expect

A key objective of our institute is the development of concepts, processes, and technologies that enhance the resilience of terrestrial infrastructures. In this context, the growing threat posed by drones in both military and civilian sectors highlights the need for advanced detection systems. Modern detection technologies leverage a variety of sensor data, including radar, RGB cameras, infrared (IR), acoustic signals, and radio frequencies, each with unique strengths and limitations. While RGB cameras combined with state-of-the-art deep learning (DL) techniques offer a cost-effective solution, they struggle in low-light condition.  In contrast, IR cameras capture thermal radiation independently of ambient lighting conditions, facilitating drone detection even in low-light environments. As such, IR sensors serve as an ideal complement to RGB-based systems. Similar to RGB data, IR data is typically processed using DL models to extract significant features. For reliable detection systems, these models require large-scale, diverse, and accurately annotated training data. However, the availability of real-world data is often limited, and manual annotations are expensive. Therefore, leveraging synthetic data for training IR-based drone detection systems offers great potential.

Drones pose significant threats in both military and civilian settings, driving the need for advanced detection systems. Modern detection technologies leverage a variety of sensor data, including radar, RGB cameras, infrared (IR), acoustic signals, and radio frequencies, each with unique strengths and limitations. While RGB cameras combined with state-of-the-art deep learning (DL) techniques offer a cost-effective solution, they struggle in low-light condition. In contrast, IR cameras capture thermal radiation independently of ambient lighting conditions, facilitating drone detection even in low-light environments. As such, IR sensors serve as an ideal complement to RGB-based systems. Similar to RGB data, IR data is typically processed using DL models to extract significant features. For reliable detection systems, these models require large-scale, diverse, and accurately annotated training data. However, the availability of real-world data is often limited, and manual annotations are expensive. Therefore, leveraging synthetic data for training IR-based drone detection systems offers great potential.

Your tasks

The objective of this work is to assess the viability of synthetic IR data for DL-based drone detection.

This includes:

• employing game engine-based simulations for generating diverse, high-quality IR data that accurately represent various environmental conditions
• selecting and training advanced DL models for drone detection using the (synthetic) IR data
• conducting an in-depth evaluation of the data's effectiveness, with a focus on identifying its strengths, potential benefits, and inherent limitations

Your profile

• currently enrolled as a Master student in computer science, mathematics, optical engineering or similar major
• experiences in working with Unreal or Unity (or something comparable), experience with Microsoft AirSim advantageous but not required
• experience in deep learning, especially in terms of object detection, is advantageous
• basic Python skills (ideally with respect to deep learning applications, e.g., PyTorch or Tensorflow

 We look forward to getting to know you!

If you have any questions about this position (Vacancy-ID 765) please contact:

Tobias Koch 
Tel.: +49 2241 20148 55