2026-03-11T09:47:20Z https://ipsj.ixsq.nii.ac.jp/oai

oai:ipsj.ixsq.nii.ac.jp:00231940 2025-01-19T10:35:38Z 1164:4619:11539:11571

NeDDFを用いた煙霧環境に適用可能なニューラル場単眼Visual SLAM Monocular Visual SLAM Applicable to Smoggy Environments Using NeDDF 髙橋, 響熙 上田, 樹 謝, 淳 北原, 格 Hibiki, Takahashi Itsuki, Ueda Chun, Xie Itaru, Kitahara We propose a monocular Visual SLAM system based on NeDDF (Neural Density-Distance Field). In contrast to previous methods using point clouds or meshes, our approach has an advantage to reconstruct objects with indistinct boundaries, such as steam, smoke, dust, and similar phenomena. Our method utilizes NeDDF as a 3D shape representation and introduces a photometric error to account for the challenging smoke environment. This approach enables the reconstruction of indoor scenes featuring low-density objects using a monocular RGB camera. To evaluate the effectiveness of our proposed method, we conduct experiments using a simulated RGB dataset of an indoor scene with smoke. We propose a monocular Visual SLAM system based on NeDDF (Neural Density-Distance Field). In contrast to previous methods using point clouds or meshes, our approach has an advantage to reconstruct objects with indistinct boundaries, such as steam, smoke, dust, and similar phenomena. Our method utilizes NeDDF as a 3D shape representation and introduces a photometric error to account for the challenging smoke environment. This approach enables the reconstruction of indoor scenes featuring low-density objects using a monocular RGB camera. To evaluate the effectiveness of our proposed method, we conduct experiments using a simulated RGB dataset of an indoor scene with smoke. 情報処理学会 2024-01-18 jpn technical report https://ipsj.ixsq.nii.ac.jp/records/231940 2188-8701 AA11131797 研究報告コンピュータビジョンとイメージメディア（CVIM） 2024-CVIM-236 18 1 6 https://ipsj.ixsq.nii.ac.jp/record/231940/files/IPSJ-CVIM24236018.pdf application/pdf 3.4 MB 2026-01-18