Application of Thermal Imaging Cameras for Visual Marker Detection and Tracking in Limited Visibility Conditions Using Artificial Intelligence Methods

This article presents a system for detecting ArUco markers in thermal infrared (IR) images, designed to overcome the typical limitations of this imaging modality, such as low resolution, thermal noise, gradient banding, and the absence of distinct edges. Conventional fiducial marker detection algorithms, effective in RGB images, fail in IR environments due to the specific nature of thermal radiation. The author developed a multi-stage preprocessing pipeline including posterization, gradient smoothing, edge enhancement, and banding elimination, enabling accurate identification of binary ArUco patterns. Additionally, methods for estimating position and orientation in 3D space were implemented, along with adaptive binary thresholding with error tolerance tailored to thermal disturbances. The developed interactive system allows users to adjust image processing parameters in real time, significantly improving ArUco marker detection performance in varying environmental conditions. Tests conducted showed an approximately 26 % improvement in detection accuracy compared to standard approaches.