Development of Three JS-based 3D Scene with Seamless Visualization of Gaussian Splatting and Transformation to Global Coordinates

Abstract

Existing scholarly literature on the Gaussian Splatting algorithm has predominantly concentrated on improving the rendering and reconstruction of three-dimensional objects, as well as exploring its applications in various academic disciplines, such as medicine, robotics, and mapping, while being limited to local coordinate systems. This study describes the development of a 3D scene modeled using the Gaussian Splatting algorithm, featuring accurate distance and position geometry based on Three JS. The developed 3D scene was then evaluated with precise position and distance coordinates in the field and compared to the established SfM-MVS (Structure from Motion-Multi View Stereo) algorithm. The findings demonstrate that the proposed development successfully generated Three JS-based 3D scenes with global coordinate compatibility utilizing the Gaussian Splatting algorithm, achieving the same level of position and distance accuracy as the SfM-MVS algorithm, with a 95% confidence interval using T-Test. This research concludes that the developed approach is successful and can be further expanded for various scientific fields that require accurate position and distance information using Gaussian Splatting Algorithm.