LCP-Fusion: A Neural Implicit SLAM with Enhanced Local Constraints and Computable Prior

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Wang, Jiahui, Deng, Yinan, Yang, Yi, Yue, Yufeng
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866913572677746688
author Wang, Jiahui
Deng, Yinan
Yang, Yi
Yue, Yufeng
author_facet Wang, Jiahui
Deng, Yinan
Yang, Yi
Yue, Yufeng
contents Recently the dense Simultaneous Localization and Mapping (SLAM) based on neural implicit representation has shown impressive progress in hole filling and high-fidelity mapping. Nevertheless, existing methods either heavily rely on known scene bounds or suffer inconsistent reconstruction due to drift in potential loop-closure regions, or both, which can be attributed to the inflexible representation and lack of local constraints. In this paper, we present LCP-Fusion, a neural implicit SLAM system with enhanced local constraints and computable prior, which takes the sparse voxel octree structure containing feature grids and SDF priors as hybrid scene representation, enabling the scalability and robustness during mapping and tracking. To enhance the local constraints, we propose a novel sliding window selection strategy based on visual overlap to address the loop-closure, and a practical warping loss to constrain relative poses. Moreover, we estimate SDF priors as coarse initialization for implicit features, which brings additional explicit constraints and robustness, especially when a light but efficient adaptive early ending is adopted. Experiments demonstrate that our method achieve better localization accuracy and reconstruction consistency than existing RGB-D implicit SLAM, especially in challenging real scenes (ScanNet) as well as self-captured scenes with unknown scene bounds. The code is available at https://github.com/laliwang/LCP-Fusion.
format Preprint
id arxiv_https___arxiv_org_abs_2411_03610
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle LCP-Fusion: A Neural Implicit SLAM with Enhanced Local Constraints and Computable Prior
Wang, Jiahui
Deng, Yinan
Yang, Yi
Yue, Yufeng
Robotics
Computer Vision and Pattern Recognition
Recently the dense Simultaneous Localization and Mapping (SLAM) based on neural implicit representation has shown impressive progress in hole filling and high-fidelity mapping. Nevertheless, existing methods either heavily rely on known scene bounds or suffer inconsistent reconstruction due to drift in potential loop-closure regions, or both, which can be attributed to the inflexible representation and lack of local constraints. In this paper, we present LCP-Fusion, a neural implicit SLAM system with enhanced local constraints and computable prior, which takes the sparse voxel octree structure containing feature grids and SDF priors as hybrid scene representation, enabling the scalability and robustness during mapping and tracking. To enhance the local constraints, we propose a novel sliding window selection strategy based on visual overlap to address the loop-closure, and a practical warping loss to constrain relative poses. Moreover, we estimate SDF priors as coarse initialization for implicit features, which brings additional explicit constraints and robustness, especially when a light but efficient adaptive early ending is adopted. Experiments demonstrate that our method achieve better localization accuracy and reconstruction consistency than existing RGB-D implicit SLAM, especially in challenging real scenes (ScanNet) as well as self-captured scenes with unknown scene bounds. The code is available at https://github.com/laliwang/LCP-Fusion.
title LCP-Fusion: A Neural Implicit SLAM with Enhanced Local Constraints and Computable Prior
topic Robotics
Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2411.03610