Evaluating the Impact of Fog on 3D Digital Twinning and Semantic Segmentation: A Synthetic Fog Modeling Approach

Unmanned aerial vehicles (UAVs) are being extensively used to gather large volumes of visual data, playing a crucial role in monitoring construction sites, managing projects, and conducting safety assessments. However, much of the prior research has concentrated on analyzing visual data under optimal conditions, often presuming clear visibility and high-quality datasets. These assumptions fail to address the challenges posed by real-world environmental factors, such as fog. This study addresses this challenge by generating synthetic fog to replicate varying levels of fog density, enabling an assessment of its effects on 3D digital twinning methodologies and 2D semantic segmentation performance metrics. By measuring the degradation caused by fog in both 3D reconstruction and 2D segmentation processes, this research provides practitioners with valuable insights into potential data loss under foggy conditions, enabling them to avoid visual data collection during dense fog and enhance the reliability of data-driven analyses.