1.武汉工程大学电气信息学院,湖北 武汉 430074
2.中冶武勘工程技术有限公司,湖北 武汉 430081
3.武汉科技大学计算机科学与技术学院,湖北 武汉 430065
4.智能信息处理与实时工业系统湖北省重点实验室,湖北 武汉 430065
时 愈(1985-),女,副教授,E-mail:shiyu0125@163.com.
收稿:2026-03-18,
网络首发:2026-10-17,
移动端阅览
时愈,曾子其,刘照等.基于混合三平面高斯溅射的高效单视图车辆重建[J].华中科技大学学报(自然科学版),
SHI Yu,ZENG Ziqi,LIU Zhao,et al.Efficient single-view vehicle reconstruction via hybrid triplane gaussian splatting[J].J.Huazhong Univ.of Sci.& Tech.(Natural Science Edition),
时愈,曾子其,刘照等.基于混合三平面高斯溅射的高效单视图车辆重建[J].华中科技大学学报(自然科学版), DOI:10.13245/j.hust.260130.
SHI Yu,ZENG Ziqi,LIU Zhao,et al.Efficient single-view vehicle reconstruction via hybrid triplane gaussian splatting[J].J.Huazhong Univ.of Sci.& Tech.(Natural Science Edition), DOI:10.13245/j.hust.260130.
为解决现有车辆重建存在的精细纹理丢失和体渲染计算量大等问题,提出一种基于混合三平面高斯溅射的高效单视图车辆重建方法,即GSCar。GSCar是一种用于单视图车辆重建的前馈推理方法,在训练和验证阶段均无需额外视图或三维真实数据。GSCar构建了混合三平面高斯表示实现单视图车辆重建,该特征结合了显式点云与隐式高斯属性,既支持高效的三维重建,也提升了渲染质量。通过三平面高斯编码器将二维图像特征映射到三维空间,GSCar实现了形状与高斯潜码的解耦,从而优化了全局形状拟合与局部纹理细节。此外,GSCar利用车辆的对称性来稳健地推断被遮挡区域。实验结果表明,GSCar在三维车辆重建上表现出色,显著提高了车辆重建的质量和速度。
To address issues such as the loss of fine texture details and the high computational cost associated with volumetric rendering in existing vehicle reconstruction methods
this paper proposes an efficient single-view vehicle reconstruction via hybrid triplane gaussian splatting
termed GSCar.GSCar is a feed-forward inference method for single-view vehicle reconstruction,without requiring additional views or 3-dimensional (3D) ground truth during both training and validation phases.GSCar constructs a hybrid triplane Gaussian representation to achieve single-view vehicle reconstruction.The representation combines explicit point clouds with implicit Gaussian attributes,enabling both efficient 3D reconstruction and improved rendering quality.By using a triplane Gaussian encoder to map 2-dimensional image features into 3D space,GSCar decouples shape and Gaussian latent codes,thereby optimizing both global shape fitting and local texture details.Additionally,GSCar leverages vehicle symmetry to robustly infer occluded regions.Experimental results show that GSCar performs excellently in 3D vehicle reconstruction
significantly improving the quality and speed of vehicle reconstruction.
MÜLLER N , SIMONELLI A , PORZI L , et al . AutoRF:learning 3D object radiance fields from single view observations [C ] // Proc of the IEEE/CVF Conference on Computer Vision and Pattern Recognition . New Orleans : IEEE , 2022 : 3971 - 3980 .
YU A , YE V , TANCIK M , et al . PixelNeRF:neural radiance fields from one or few images [C ] // Proc of the IEEE/CVF Conference on Computer Vision and Pattern Recognition . Virtual : IEEE , 2021 : 4578 - 4587 .
JANG W B , AGAPITO L . CodeNeRF:disentangled neural radiance fields for object categories [C ] // Proc of the IEEE/CVF International Conference on Computer Vision . Virtual : IEEE , 2021 : 12949 - 12958 .
TOCHILKIN D , PANKRATZ D , LIU Z X , et al . TripoSR:fast 3D object reconstruction from a single image [EB/OL ] .[ 2026-01-23 ] . https://github.com/VAST-AI-Research/TripoSR https://github.com/VAST-AI-Research/TripoSR .
JIANG H W , HUANG Q X , PAVLAKOS G . Real3D:towards scaling large reconstruction models with real images [C ] // Proc of the IEEE/CVF International Conference on Computer Vision . Honolulu : IEEE , 2025 : 5821 - 5833 .
陈芬 , 陈学煌 , 李旭 , 等 . 面向六自由度视频应用的动态场景重建 [J ] . 华中科技大学学报(自然科学版) , 2025 , 53 ( 3 ): 48 - 55 .
CAI Y H , ZHANG H , ZHANG K , et al . Baking gaussian splatting into diffusion denoiser for fast and scalable single-stage image-to-3D generation and reconstruction [C ] // Proc of the IEEE/CVF International Conference on Computer Vision . Honolulu : IEEE , 2025 : 25062 - 25072 .
MIAO H , LU F , LIU Z , et al . Robust 2D/3D vehicle parsing in arbitrary camera views for CVIS [C ] // Proc of the IEEE/CVF International Conference on Computer Vision . Montreal : IEEE , 2021 : 15611 - 15620 .
彭刚 , 刘星宇 , 刘杰 , 等 . 结合深度梯度与对齐融合策略的目标检测方法 [J ] . 华中科技大学学报(自然科学版) , 2025 , 53 ( 11 ): 15 - 20 .
庞立泰 , 谭鑫 . 基于深度估计进行高斯初始化的三维占用预测 [J ] . 华中科技大学学报(自然科学版) , 2025 , 53 ( 12 ): 178 - 184 .
GEIGER A , LENZ P , URTASUN R . Are we ready for autonomous driving? The kitti vision benchmark suite [C ] // Proc of the IEEE Conference on Computer Vision and Pattern Recognition . Las Vegas : IEEE , 2012 : 3354 - 3361 .
CAESAR H , BANKITI V , LANG A H , et al . Nuscenes:a multimodal dataset for autonomous driving [C ] // Proc of the IEEE/CVF Conference on Computer Vision and Pattern Recognition . Seattle : IEEE , 2020 : 11621 - 11631 .
ZHANG R , ISOLA P , EFROS A A , et al . The unreasonable effectiveness of deep features as a perceptual metric [C ] // Proc of the IEEE Conference on Computer Vision and Pattern Recognition . Salt Lake City : IEEE , 2018 : 586 - 595 .
HEUSEL M , RAMSAUER H , UNTERTHINER T , et al . GANs trained by a two time-scale update rule converge to a local nash equilibrium [C ] // Proc of the 31st International Conference on Neural Information Processing Systems . Long Beach : Curran Associates Inc , 2017 : 6626 - 6637 .
0
浏览量
0
下载量
0
CSCD
关联资源
相关文章
相关作者
相关机构
京公网安备11010802024621