Hyperspectral Imaging or Victim Detection with Rescue Robots

Abstract

The main task of rescue robots is to locate victims after a disaster such as an earthquake. For this task sensor data is used to localize the robots in their environment, build maps, and mark the victims in the maps. Usually, thermal and color cameras, monitored by a human operator, are used for the detection. Hyperspectral imaging techniques are today used for industrial tasks such as quality control, fast material sorting or food analysis. This paper proposes a new approach for the victim detection in rescue environments, based on hyperspectral imaging in the near infrared spectral domain. This technique involves a simultaneous recording of spatial and spectral information. Different materials can be distinguished when the spectra are analyzed. The result of the experiments show that the spectra of skin are very characteristic and that even under the impact of ash layers the spectral similarity remains very high. Thus our approach can be used for rescue robots to find human bodies autonomously, where other techniques such as color or thermal image analysis would fail.