BIM-Based 3-D Multimodal Reconstruction for Substation Equipment Inspection Images

Multimodal imaging inspection with intelligent inspection robots and drones has revolutionized the online monitoring and diagnosis of substation equipment. Nevertheless, such robots and drones generate a large amount of 2-D multimodal inspection images, increasing the workload for substation staff tasked with image analysis. To address this problem, we develop a 3-D reconstruction approach that integrates these 2-D images into intuitive 3-D multimodal models. Specifically, a novel 3-D multimodal reconstruction framework, leveraging the building information model (BIM) of the substation, is designed for recovering the high-fidelity 3-D multimodal models from the 2-D multimodal images of substation equipment. Central to this framework is an image decision strategy, used to accurately match different modal images. In addition, an image harmonization method is demonstrated to minimize discrepancies among the same modal images, maintaining color consistency for the texture mapping between the 3-D BIM and 2-D multimodal images. The quantitative and qualitative experiments conducted on self-built virtual and real-world substation platforms demonstrate the superiority of our method, which yields more complete and accurate 3-D multimodal models than state-of-the-art reconstruction techniques.