State-of-the-Art Techniques in 3D Industrial Reconstruction: A Detailed Survey
摘要
Three-dimensional (3D) reconstruction involves extracting 3D information about real-world objects from limited sensory data such as images, depth maps, or point clouds. Recently, 3D reconstruction has been widely used in industry for tasks such as object measurement, defect detection, and quality control. This paper investigates and summarizes the recent research on 3D reconstruction in industry from perspectives of surface reconstruction and internal reconstruction. In our paper, surface reconstruction is further categorized into geometric methods and deep learning methods. Geometric methods comprise contact methods and non-contact methods which include laser scanning reconstruction, stereo vision, and structured light methods. Furthermore, internal reconstruction methods are grouped into electron backscatter diffraction (EBSD) analysis and computed tomography (CT) reconstruction. We introduce the representative research works in each category, compare their advantages and limitations, and analyze their applicability to different industrial challenges. Ultimately, we clarify the future research direction in industrial 3D reconstruction.