用3D高斯点云实现厘米级相机定位，精度远超现有方法。
GSplatLoc: Ultra-Precise Camera Localization via 3D Gaussian Splatting
- 基于可微渲染优化相机位姿，直接对齐深度图差异。
- 在Replica数据集上平移误差小于0.01厘米，旋转误差接近零。
- 适合需要高精度实时定位的机器人与增强现实场景。
我们提出GSplatLoc，一种利用3D高斯点云可微渲染能力实现超精密位姿估计的相机定位方法。通过将位姿估计建模为梯度优化问题，最小化预构建3D高斯场景渲染的深度图与实际观测深度图像之间的差异，该方法在Replica数据集上实现了小于0.01厘米的平移误差和近乎为零的旋转误差，显著优于现有方法。在Replica和TUM RGB-D数据集上的评估表明，该方法在具有复杂相机运动的挑战性室内环境中仍具备鲁棒性。GSplatLoc为密集映射中的定位设立了新基准，对机器人学与增强现实等需高精度实时定位的应用具有重要意义。
原文摘要 · Abstract (English)
We present GSplatLoc, a camera localization method that leverages the differentiable rendering capabilities of 3D Gaussian splatting for ultra-precise pose estimation. By formulating pose estimation as a gradient-based optimization problem that minimizes discrepancies between rendered depth maps from a pre-existing 3D Gaussian scene and observed depth images, GSplatLoc achieves translational errors within 0.01 cm and near-zero rotational errors on the Replica dataset - significantly outperforming existing methods. Evaluations on the Replica and TUM RGB-D datasets demonstrate the method's robustness in challenging indoor environments with complex camera motions. GSplatLoc sets a new benchmark for localization in dense mapping, with important implications for applications requiring accurate real-time localization, such as robotics and augmented reality.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。