[DC] Exploring Virtual Photogrammetry Techniques and Applications For Advancement of Digital Twin Generation

Jacob Rubinstein

Abstract: We explore challenges and opportunities in the use of photogrammetry for digital twin generation, namely we aim to improve our understanding of what makes good input data for photogrammetry and to quantify the different traits of various photogrammetry processes. We propose the use of virtual photogrammetry - utilizing synthetic 2D images from pre-existing 3D models as input - to aid in this goal. Our approach aims to create a pipeline for generating datasets of synthetic images which can be used to evaluate and improve camera pose/ intrisics estimation as well as to assess the impact of errors on 3D reconstruction accuracy. By leveraging the advantages of this synthetic data, we aim to evaluate the resilience and accuracy of photogrammetry systems, leading to the higher quality results from non-virtual photogrammetry in the future.