Multi-Modal Inspection of Industrial Structures v1.0

Thermography is a Non-Destructive Testing (NDT) technology that measures the thermal distribution of a specimen by quantifying the electromagnetic radiation emitted, reflected and transmitted at lower frequency than the visible light part of the spectrum. Despite some previous studies addressing the estimation of surface and shape characterization from multiple or single active thermograms, thermography, by nature, is a bi-dimensional sensing technology unable to provide information about the specimen's texture and geometry without any preparatory process. Thus, many studies have recently focused on using multi-modal platforms to obtain extensive information about the inspected scene. Still, data availability and algorithm implementation difficulties are faced by the analysts when performing the registration of consecutive 3D data from multiple sensors and Fields of Views (FOVs). This study presents a complete solution for multi-modal inspection of industrial components, including a processing pipeline for registering consecutive multi-modal point clouds. A comparative evaluation of optimization and learning based registration methods is provided as part of the processing pipeline. Moreover, a benchmark dataset of point cloud data from different FOVs of industrial and construction components is provided (Lemanchot-points), having 5 point clouds with depth, color, and thermal information at each point. The experimental campaign conducted with different objects demonstrates the proposed solution's applicability for the multi-modal inspection of industrial components.

热成像（Thermography）是一种无损检测（Non-Destructive Testing, NDT）技术，通过量化光谱中低于可见光频段的电磁辐射的发射、反射与透射情况，实现试件热分布的测量。尽管此前已有多项研究围绕基于单幅或多幅主动热成像图开展表面与形状特征估计，但热成像本质上属于二维传感技术，若未开展相应预处理步骤，则无法提供试件的纹理与几何信息。因此，近期诸多研究聚焦于借助多模态平台获取被检场景的全面信息。然而，分析人员在对多传感器与多视场（Fields of Views, FOVs）采集的连续三维数据进行配准时，仍会遭遇数据可用性不足与算法实现难度较高的困境。本研究提出了一套面向工业构件多模态检测的完整解决方案，其中包含用于配准连续多模态点云的处理流水线。作为该处理流水线的组成部分，本文还对基于优化与基于学习的配准方法开展了对比评估。此外，本文还公开了一套涵盖工业与建筑构件不同视场的点云基准数据集（Lemanchot-points），该数据集包含5组点云数据，每组点云的每个点均附带深度、色彩与热成像信息。针对不同对象开展的实验验证表明，所提解决方案可有效应用于工业构件的多模态检测任务。