This talk focuses on the specific contribution of long wave infrared (LWIR) thermal imaging to robustness and ODD expansion. Thermal cameras provide strong human/animal contrast independent of visible light, improving pedestrian and animal detection at night, and deep shadow where visible cameras saturate and conventional pipelines underperform. We summarize evidence from industry and research showing thermal’s advantages for nocturnal VRU detection and PAEB initiation.
Beyond illumination stressors, we present qualitative and quantitative case studies covering rain, fog, smoke and dust, partially occluded pedestrians, sunrise/sunset, snow, and tires/debris on the road. We document how atmosphere optical transmission affect active and passive sensors in precipitation and fog, while thermal imaging often preserves usable semantic contrast—particularly in smoke/dust. We also note thermal’s passive nature (no emissions), which avoids eye safety constraints and mutual interference issues that can affect active sensors in dense traffic.
A central portion of the talk addresses sensor fusion. We report results where RGB thermal fusion improves VRU detection relative to single modality baselines. Performance is analyzed through detection/false alarm trade offs, time to collision.
We close with an ODD landscape versus sensor performance. The result is a pragmatic path for leveraging thermal imaging to extend AEB/ADAS capability while maintaining safety margins in the real world.
