Thermal-to-Color Image Translation Using Heat Conduction for 3D Mapping in Low-Visibility Environments

For safe firefighting and rescue operations at the scene of a fire, it is extremely important to accurately assess the situation inside a building. Currently, however, it is very difficult to generate a three-dimensional map in an environment filled with smoke and dust. To solve this problem, this study attempts to generate a three-dimensional map using a thermal camera, which is less susceptible to smoke. The thermal image acquired by a thermal camera is a grayscale image based on temperature gradients, which limits the number of feature points that can be obtained. In particular, it is difficult to extract feature points in environments where there is little temperature variation, making map generation difficult. In order to enable effective map generation even in such environments, this research aims to improve map generation accuracy by converting thermal images from thermal cameras into RGB images using deep learning.