Neural rendering-enabled 3D modeling for rapid digitization of in-service products

Abstract

Rapid digitization of physical objects enables monitoring, analysis, and maintenance of in-service products, of which an up-to-date CAD model is not available. It provides designers with the products’ actual response to the real-world usage, which provides a reference base for design optimization. This paper presents neural rendering as a novel method for rapid digital model building. It learns a radiance field from RGB images to determine the characteristics of the physical object. Textured mesh can be generated from the learned radiance field for efficient 3D modeling. The effectiveness of the method is demonstrated by an engine component.