High-resolution thermal infrared dataset for airborne person detection in SAR missions

Authors: Johannes Büttner, Kilian Führer, Andreas Fritz, Jonathan Zender and Bernd R. Pinzer
Published: 7 July 2026
Original: Scientific Data

Summary:

Operational importance: The study demonstrates why thermal detection cannot be reduced to locating bright shapes. In hot terrain, rocks, vegetation and animals may equal or exceed human surface temperature, making shape, context and repeated observation important. The benchmark better reflects small targets, unusual poses and alpine backgrounds than many existing datasets.

It supports practical training in:

• Thermal false positives and missed detections.

• Seasonal and terrain effects.

• Appropriate altitude and image resolution.

• Deliberate review of possible anomalies.

• Maintaining human confirmation when using automated detection.

Transferability: Relevant to UAV and airborne thermal-search practitioners. Terrain-specific results require adaptation outside alpine environments.

Limitations: Much of the training imagery was staged and recorded manually from cable cars, towers and bridges. The benchmark used simulated SAR exercises, not uncontrolled live searches. Improved model accuracy does not demonstrate improved operational find rates, and performance should not be assumed for other cameras, altitudes, terrains or weather conditions.

Training sections: Resource selection and tasking; UAV thermal operations; visual search; detection reliability; Search Operations; debrief and assurance.

Training decision — Reinforce: Thermal imagery should be treated as a specialist detection aid requiring trained interpretation, systematic review and confirmation—not as inherently reliable coverage.