XGRIDS Builds World's Largest Air-Ground Integrated 3D Gaussian Model of Minning Town with 6.2 Billion Gaussians and 15 Billion Point Clouds

XGRIDS used its self-developed spatial capture equipment and software, combined with its independently developed large-scale 3DGS data format LCC, to reconstruct a 23.5 square kilometer area of Minning Town into a continuous 3D model containing 6.2 billion Gaussian ellipsoids. The model was trained and generated from 186,000 aerial and ground images and 15 billion radar point cloud data points, making it the largest air-ground integrated 3D Gaussian scene to date.

3D Gaussian Splatting (3DGS) is a 3D reconstruction technique that has emerged in recent years in the fields of computer vision and graphics. Its core relies on a large-scale data-driven parameter optimization process, closely related to machine learning methods. Applying 3DGS to reconstruct real-world scenes of such a large area poses extremely high demands on data capture, storage, and rendering pipelines.

The model has been applied to the production of CCTV's key documentary "The Minning Avenue," providing viewers with an immersive digital space they can freely navigate. This marks the transition of 3DGS technology from academic research and small-scale demonstrations to practical application scenarios such as film and television content production.

From a technical perspective, the scale of 6.2 billion Gaussian ellipsoids means the model has an extremely large number of parameters, posing challenges to training compute power and rendering efficiency. XGRIDS' self-developed LCC data format and capture equipment reflect the sustained investment by domestic teams in spatial intelligence infrastructure, and also provide engineering references for future large-scale scene reconstruction.

This page provides an editorial summary based on publicly available information. It is not a republished article. Use the source link below for the original report.