Edge-Based Monocular Thermal Odometry in Low Illumination Environments

Achieving accurate motion estimation in low-illumination environments is challenging for traditional Visual Odometry (VO) that utilizes visible cameras. In contrast, long-wave infrared (LWIR) cameras can operate independently of illumination conditions. However, the direct applicability of traditional VO methods to thermal images is limited due to poor image quality. This paper proposes an edge-based monocular thermal Visual Odometry, which achieves motion estimation by matching edge feature points and incorporates loop detection to maintain global map consistency. Experimental results demonstrate that the system can achieve accurate and robust localization in low-illumination environments.