RPC-GS: Gaussian Splatting with native RPC Rendering for Satellite Imagery

2026-04-13 · Source: cs.CV updates on arXiv.org · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Geospatial Technology · Depth: Expert, extended

Summary

RPC-GS is introduced as the first Gaussian Splatting framework designed for satellite imagery that natively integrates Rational Polynomial Camera (RPC) models. Unlike prior methods that approximate RPCs with perspective or affine camera models, RPC-GS directly projects Gaussian means and covariances through the RPC model. This approach embeds the RPC model within a geo-coordinate transformation chain and uses a numerically robust Jacobian-based covariance projection. Benchmarking against perspective and affine approximations on DFC2019 and IARPA2016 datasets, RPC-GS consistently achieves lower reconstruction errors. It reduces mean altitude error by 29.6% and 63.8% over perspective and affine approximations on DFC2019, and by 9.9% and 37.9% on IARPA2016, demonstrating superior geometric accuracy.

Key takeaway

For Computer Vision Engineers developing 3D reconstruction pipelines for satellite imagery, you should prioritize native RPC model integration over approximations. This approach, as demonstrated by RPC-GS, significantly reduces geometric errors, improving mean altitude accuracy by up to 63.8% on datasets like DFC2019. Adopting RPC-native methods will yield more precise digital surface models, crucial for applications requiring high-fidelity geospatial data.

Key insights

Native RPC integration in Gaussian Splatting significantly improves satellite imagery 3D reconstruction accuracy.

Principles

• Approximating RPC models introduces systematic geometric errors.
• Direct RPC integration enhances 3D reconstruction fidelity.
• Geo-coordinate transformations are crucial for satellite data.

Method

RPC-GS projects Gaussian means and covariances through a geo-coordinate transformation chain (Scene -> ENU -> ECEF -> Geodetic -> Image) and uses a Jacobian-based covariance projection with a metric ray-based depth for alpha compositing.

In practice

• Use RPC-native rendering for high-precision satellite 3D models.
• Implement Jacobian-based covariance propagation for nonlinear transforms.
• Employ ray-based depth for accurate RPC camera ordering.

Topics

• Gaussian Splatting
• Rational Polynomial Camera
• Satellite Imagery
• 3D Reconstruction
• Geo-coordinate Transformations
• Digital Surface Models

Best for: AI Scientist, Computer Vision Engineer, Research Scientist

Related on AIssential

• RPC-GS: Gaussian Splatting with native RPC Rendering for Satellite Imagery — RPC-GS improves satellite imagery 3D reconstruction by natively integrating Rational Polynomial Camera models into Gaussian Splatting, avoiding geometric approximations.
• NG-GS: NeRF-Guided 3D Gaussian Splatting Segmentation — NG-GS enhances 3D Gaussian Splatting segmentation by integrating NeRF-guided continuous feature fields to resolve boundary artifacts.
• GlobalSplat: Efficient Feed-Forward 3D Gaussian Splatting via Global Scene Tokens — GlobalSplat uses a global latent scene representation for efficient, compact, and consistent 3D Gaussian Splatting.
• Less Gaussians, Texture More: 4K Feed-Forward Textured Splatting — LGTM enables 4K novel view synthesis by decoupling geometry from resolution using textured Gaussian primitives.
• Three-dimensional Damage Visualization of Civil Structures via Gaussian Splatting-enabled Digital Twins — Gaussian Splatting enhances 3D damage visualization in civil infrastructure digital twins, improving accuracy and efficiency.
• Beyond Visual Fidelity: Benchmarking Super-Resolution Models for Large-Scale Remote Sensing Imagery via Downstream Task Integration — Traditional SR metrics often fail to predict real-world performance in downstream remote sensing tasks.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by cs.CV updates on arXiv.org.