Dense 3D Mapping in Extremely Dark Environments Based on Colored LLAH Descriptor Using Phosphorescent Materials

Abstract

In this paper, we propose a dense three-dimensional (3D) mapping approach for extremely dark environments based on structure from motion (SfM) and multi-view stereo (MVS) using a general optical camera and phosphorescent materials in extremely dark environments. Conventional methods that use phosphorescent emission as a visual feature suffer from unstable feature correspondence when gradient-based descriptors are used, owing to the temporal decay of the emission intensity. Maintaining uniform brightness requires repeated ultraviolet irradiation, which is impractical in power-limited environments such as lunar caves. To address this issue, we apply locally likely arrangement hashing as a novel framework to find the corresponding feature points for SfM by leveraging the geometric arrangement and luminous color of phosphorescent materials. The experimental results demonstrated that the proposed SfM–MVS framework can construct a geometrically accurate dense 3D map even in extremely dark environments.

Overview of the proposed framework for dense 3D mapping in extremely dark environments