TACSTechnology and Application of Computer Science2998-89262998-8934Art and Technology10.61369/TACS.2025140058Article一种面向动态场景的3DGS-SLAM 方法https://artdesignp.com/journal/TACS/2/14/10.61369/TACS.2025140058李梓鹏20252142025-07-28为克服现有3DGS-SLAM 算法在动态环境中跟踪稳定性不足及单目视觉绝对尺度缺失的局限性，本文提出一种具有高精度位姿估计与高保真建图能力的动态场景3DGS-SLAM 方法。首先，采用贝叶斯概率框架融合多源信息生成动态掩模，有效过滤动态干扰特征；其次，引入尺度感知的位姿联合优化策略以解决单目尺度缺失问题；最后，结合动态感知的3D 高斯建图与渲染框架，确保高质量渲染与几何一致性。实验结果表明，本文方法在位姿估计精度、渲染质量方面均优于其它对比方法，初步验证其在动态场景下的有效性与重建鲁棒性。动态场景,视觉SLAM,3D 高斯绘制,动态掩模

[1] 王俊超，张金山，杨璐嘉，等. 基于神经辐射场与高斯溅射的3D 虚拟人重建与驱动方法综述 [J/OL]. 计算机辅助设计与图形学学报, 1-31[2026-03-31][2]Mur-Artal R, Tardós J D.ORB-SLAM3: An accurate open-source library for visual, visual-inertial, and multi-map SLAM[J]. IEEE Transactions on Robotics,2021,37(6):1874-1894.[3]Engel J, Koltun V, Cremers D. DSO: Direct sparse odometry[J]. IEEE transactions on pattern analysis and machine intelligence, 2017, 40(3): 616-634.[4]Wang W, Chen H, Li Y, et al. NICE-SLAM: Neural Implicit Scalable Encoding for SLAM[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition.2022:13039-13048.[5]Keetha N, Sunderhauf N, Milford M. SplaTAM: Splatting into a Neural Radiance Field for Real-Time 3D Reconstruction with a Camera[J].arXiv preprint arXiv:2308.09661，2023.[6] 高翔，张涛，刘毅，等．视觉SLAM 十四讲— 从理论到实践[M]．北京：电子工业出版社，2017:17-22.[7]Kerbl B，Kopanas G，Leimkühler T，et al.3D Gaussian Splatting for Real-Time Radiance Field Rendering[J].ACM Transactions on Graphics (TOG),2023,42(4):1-14.