Frequency-aware and depth-guided neural radiance fields for 3D reconstruction of complex disaster scenes

Smoke occlusion, weak textures, non-Lambertian reflectance, and structural fragmentation in complex disaster scenes undermine multi-view photometric consistency and exacerbate depth ambiguity under sparse-view settings, resulting in geometric floaters, blurred boundaries, and the loss of fine details in neural radiance field reconstruction. To address these challenges, we propose a neural radiance field reconstruction method that integrates frequency-aware gating with depth-guided geometric priors. The proposed method constructs a hierarchical spatial representation using multi-resolution hash encoding, adaptively fuses low-frequency geometric features with high-frequency residual features through frequency-aware gating, and incorporates cross-level geometric consistency regularization and dynamic residual learning to suppress structural drift caused by hash collisions and enhance local responses in spectral transition regions. To mitigate geometric degradation under sparse-view conditions, monocular depth priors are further introduced, together with scale-and-shift alignment and an uncertainty-aware spectral consistency constraint, to reduce the adverse influence of erroneous priors on geometry optimization. Experimental results show that the proposed method improves the PSNR by 1.865 dB over the baseline model. Compared with the model using only depth priors, the proposed method reduces the depth MSE by 88.99%, demonstrating its effectiveness in improving pixel-level reconstruction accuracy, structural similarity, and depth consistency in complex disaster scenes.

Authors

Institutions

Publication Details

Journal
Scientific Reports
Published
2026-11-15
DOI
https://doi.org/10.1038/s41598-026-80278-7
Primary Topic
3D Shape Modeling and Analysis
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Frequency-aware and depth-guided neural radiance fields for 3D reconstruction of complex disaster scenes

Xinlong Chen, Dewen Tang, Junjia Chen, Zhong Wang et al.
Scientific Reports
3D Shape Modeling and Analysis
article

Frequency-aware and depth-guided neural radiance fields for 3D reconstruction of complex disaster scenes

Xinlong Chen, Dewen Tang, Junjia Chen, Zhong Wang, Yi Chen, Liu HM
article en

Abstract

Smoke occlusion, weak textures, non-Lambertian reflectance, and structural fragmentation in complex disaster scenes undermine multi-view photometric consistency and exacerbate depth ambiguity under sparse-view settings, resulting in geometric floaters, blurred boundaries, and the loss of fine details in neural radiance field reconstruction. To address these challenges, we propose a neural radiance field reconstruction method that integrates frequency-aware gating with depth-guided geometric priors. The proposed method constructs a hierarchical spatial representation using multi-resolution hash encoding, adaptively fuses low-frequency geometric features with high-frequency residual features through frequency-aware gating, and incorporates cross-level geometric consistency regularization and dynamic residual learning to suppress structural drift caused by hash collisions and enhance local responses in spectral transition regions. To mitigate geometric degradation under sparse-view conditions, monocular depth priors are further introduced, together with scale-and-shift alignment and an uncertainty-aware spectral consistency constraint, to reduce the adverse influence of erroneous priors on geometry optimization. Experimental results show that the proposed method improves the PSNR by 1.865 dB over the baseline model. Compared with the model using only depth priors, the proposed method reduces the depth MSE by 88.99%, demonstrating its effectiveness in improving pixel-level reconstruction accuracy, structural similarity, and depth consistency in complex disaster scenes.

Scientific Reports
University of South China (CN), Hunan Institute of Technology (CN)
3D Shape Modeling and Analysis
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon