Reliable 3D person detection is critical for rescue robotics, but the paper notes that most multimodal detection research is oriented toward automotive scenarios, where pedestrians appear in a narrow range of conditions. The new arXiv paper introduces TriSAR-3D, a dataset combining LiDAR, RGB, and thermal data for search-and-rescue environments, and PaintFusion, a method for 3D person detection from those modalities.
According to the abstract, the work is motivated by the need to give first responders situational awareness. The dataset and method are presented together, with the authors arguing that rescue settings differ from driving scenes. As this is a single source, there are no conflicting findings to compare.