Probabilistic Volumetric Fusion for Dense Monocular SLAM

被引:13
作者
Rosinol, Antoni [1 ]
Leonard, John J. [1 ]
Carlone, Luca [1 ]
机构
[1] MIT, Cambridge, MA 02139 USA
来源
2023 IEEE/CVF WINTER CONFERENCE ON APPLICATIONS OF COMPUTER VISION (WACV) | 2023年
关键词
D O I
10.1109/WACV56688.2023.00311
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
We present a novel method to reconstruct 3D scenes from images by leveraging deep dense monocular SLAM and fast uncertainty propagation. The proposed approach is able to 3D reconstruct scenes densely, accurately, and in real-time while being robust to extremely noisy depth estimates coming from dense monocular SLAM. Differently from previous approaches, that either use ad-hoc depth filters, or that estimate the depth uncertainty from RGB-D cameras' sensor models, our probabilistic depth uncertainty derives directly from the information matrix of the underlying bundle adjustment problem in SLAM. We show that the resulting depth uncertainty provides an excellent signal to weight the depth-maps for volumetric fusion. Without our depth uncertainty, the resulting mesh is noisy and with artifacts, while our approach generates an accurate 3D mesh with significantly fewer artifacts. We provide results on the challenging Euroc dataset, and show that our approach achieves 92% better accuracy than directly fusing depths from monocular SLAM, and up to 90% improvements compared to the best competing approach.
引用
收藏
页码:3096 / 3104
页数:9
相关论文
共 50 条
[41]   Real-time dense map fusion for stereo SLAM [J].
Pire, Taihu ;
Baravalle, Rodrigo ;
D'Alessandro, Ariel ;
Civera, Javier .
ROBOTICA, 2018, 36 (10) :1510-1526
[42]   Unsupervised Monocular Depth Estimation Based on Dense Feature Fusion [J].
Chen Ying ;
Wang Yiliang .
JOURNAL OF ELECTRONICS & INFORMATION TECHNOLOGY, 2021, 43 (10) :2976-2984
[43]   Real-Time Dense Monocular SLAM With Online Adapted Depth Prediction Network [J].
Luo, Hongcheng ;
Gao, Yang ;
Wu, Yuhao ;
Liao, Chunyuan ;
Yang, Xin ;
Cheng, Kwang-Ting .
IEEE TRANSACTIONS ON MULTIMEDIA, 2019, 21 (02) :470-483
[44]   Dense Mapping from Feature-Based Monocular SLAM Based on Depth Prediction [J].
Duan, Yongli ;
Zhang, Jing ;
Yang, Lingyu .
2018 IEEE CSAA GUIDANCE, NAVIGATION AND CONTROL CONFERENCE (CGNCC), 2018,
[45]   A Front-End for Dense Monocular SLAM using a Learned Outlier Mask Prior [J].
Zhang, Yihao ;
Leonard, John J. .
2021 IEEE INTERNATIONAL CONFERENCE ON ROBOTICS AND AUTOMATION (ICRA 2021), 2021, :11732-11738
[46]   DeepRelativeFusion: Dense Monocular SLAM using Single-Image Relative Depth Prediction [J].
Loo, Shing Yan ;
Mashohor, Syamsiah ;
Tang, Sai Hong ;
Zhang, Hong .
2021 IEEE/RSJ INTERNATIONAL CONFERENCE ON INTELLIGENT ROBOTS AND SYSTEMS (IROS), 2021, :6641-6648
[47]   NR-SLAM: Nonrigid Monocular SLAM [J].
Rodriguez, Juan J. Gomez ;
Montiel, Jose M. M. ;
Tardos, Juan D. .
IEEE TRANSACTIONS ON ROBOTICS, 2024, 40 :4252-4264
[48]   Dense Reconstruction from Visual SLAM with Probabilistic Multi-Sequence Merging [J].
Zhang, Hanxiang ;
Xu, Chang ;
Gu, Jason .
2022 IEEE CANADIAN CONFERENCE ON ELECTRICAL AND COMPUTER ENGINEERING (CCECE), 2022, :33-40
[49]   Camera-Agnostic Monocular SLAM and Semi-dense 3D Reconstruction [J].
Ruenz, Martin ;
Neuhaus, Frank ;
Winkens, Christian ;
Paulus, Dietrich .
PATTERN RECOGNITION, GCPR 2016, 2016, 9796 :285-296
[50]   HI-SLAM: Monocular Real-Time Dense Mapping With Hybrid Implicit Fields [J].
Zhang, Wei ;
Sun, Tiecheng ;
Wang, Sen ;
Cheng, Qing ;
Haala, Norbert .
IEEE ROBOTICS AND AUTOMATION LETTERS, 2024, 9 (02) :1548-1555