A study for detecting disaster victims using multi-copter drone with a thermographic camera and image object recognition by SSD

Abstract

It is dangerous for people to approach a building damaged by an earthquake when there is a risk of it collapsing, and the likelihood of survival of victims of building collapse decreases over time. Various rescue robots have been developed, most of which are ground-based. However, these robots tend to be large and heavy, and they move slowly on rough terrain. This study assessed the ability of a small, lightweight, fast-moving multicopter drone to locate disaster victims from the air. An image recognition system for identifying victims in an automated manner was proposed. We mounted a wide-angle camera on the drone, which is easy to control in a narrow building space. In some experiments, the drone was equipped with a thermographic camera to facilitate detection of victims. We investigated the effects of each and looked for ways to detect victims more accurately.