Speaker
Description
This article addresses the need for dense and spatially coherent three-dimensional thermal information in short-range thermal digitization processes, where small- and medium-sized objects require an accurate representation of both their geometry and thermal distribution. Conventional two-dimensional thermography techniques, although widely used for object inspection and thermal analysis, provide only a partial view of thermal phenomena due to the lack of complete geometric information and three-dimensional spatial referencing. In this context, 3D thermal point clouds emerge as a suitable solution for the unified integration of geometric and thermal information, enabling a more complete and accurate characterization of digitized objects in controlled environments.
This work presents a system capable of generating dense 3D thermal point clouds as a preliminary step toward the creation of complete thermal models. The system is based on the fusion of three imaging sensors: two RGB cameras and one thermal camera, all integrated into a portable and easily deployable setup. The combination of visible and thermal cameras allows the high geometric resolution provided by RGB imagery to be associated with the thermal information captured by the infrared camera, resulting in an enriched three-dimensional representation. The system design is oriented toward medium-scale environments, both indoor and outdoor, while maintaining a balance between portability, accuracy, and acquisition capability.
The article describes in detail the proposed methodology for generating the thermal point cloud, with particular emphasis on the calibration process between the different sensors. Accurate calibration is a critical aspect of such systems, as small errors in geometric or thermal alignment can lead to significant inconsistencies in the final model. The proposed approach incorporates an automatic calibration method that introduces novel elements with respect to previously published works, improving the accuracy, robustness, and repeatability of the system while reducing dependence on manual procedures or environment-specific configurations.
One of the main contributions of the system is its ability to generate thermal point clouds in real time. This capability significantly expands the range of potential applications, enabling its use in dynamic scenarios, rapid inspection tasks, continuous monitoring processes, or integration with robotic platforms and mobile acquisition systems. The immediate availability of three-dimensional thermal information is particularly relevant in contexts where real-time decision-making or iterative thermal evaluation is required.
The experimental validation of the system was carried out in different indoor and outdoor scenarios, covering environments with varying geometric and thermal conditions. The obtained results demonstrate the feasibility of the proposed approach and its ability to generate dense, coherent, and spatially consistent thermal point clouds. Overall, the experiments highlight the potential of the system as a foundational tool for the generation of advanced thermal models and its application in areas such as technical inspection, energy analysis, environment digitization, and three-dimensional thermal monitoring.
