EUVIMEDEUVIMED entdecken
Crossref

Local-prior–guided Segmentation and Visibility-normalized Photogrammetric Mapping of Stone-carved Heritage Defects

Xiaohu Lin, Chenlu Liang, Bei Xue, Yinan Gao, Tao Yan, Yanan Guo, Yuxiang He

Photogrammetric Engineering & Remote Sensing · 2026

DOI: 10.14358/pers.26-00125r2 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Photogrammetric Engineering & Remote Sensing, 2026. Verfasst von Xiaohu Lin und 6 weiteren Autor:innen.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Photogrammetric three-dimensional (3D) mapping supports defect documentation for stone-carved heritage. However, complex surface appearance and varying view coverage hinder reliable two-dimensional (2D)–to-3D defect mapping. To address these challenges, this study proposes a local-prior–guided semiautomatic segmentation and visibility-normalized photogrammetric mapping method. First, coarse defect boxes guide region-of-interest segmentation to produce pixel-level defect masks, while category labels retain the defect class. Second, structure-from-motion and multi-view stereo (SfM-MVS) reconstruct the point cloud and recover camera geometry. The predicted masks are projected onto the 3D surface, and depth consistency removes inconsistent projections. Finally, multi-view semantic evidence is aggregated using a joint rule that combines a minimum support count with a support ratio normalized by geometrically valid views. Experiments covered multiple heritage objects at the Qianling Mausoleum and two independent public objects. On the statue test set, 2D segmentation achieved 74.21 ± 2.25% mean intersection over union (IoU) and 83.96 ± 1.80% mean Dice similarity coefficient. With the 3D decision parameters fixed on statue A, three views of statue B were excluded from fusion and used only for evaluation. The resulting macro-F1 and macro-IoU were 88.43 ± 0.24% and 84.63 ± 0.41%, respectively. Further tests trained on one statue and tested on the other and extended complete 3D mapping to the Qianling Uninscribed Stele, providing further evidence of transfer across different objects and deterioration patterns. These results indicate that local priors support fine-grained 2D defect delineation, while geometric verification and visibility-normalized multi-view aggregation support reliable 3D semantic mapping within the evaluated SfM-MVS photogrammetric setting.

Zitieren

APA

Lin, X., Liang, C., Xue, B., Gao, Y., Yan, T., Guo, Y., & He, Y. (2026). Local-prior–guided Segmentation and Visibility-normalized Photogrammetric Mapping of Stone-carved Heritage Defects. Photogrammetric Engineering & Remote Sensing. https://doi.org/10.14358/pers.26-00125r2

Vancouver

Lin X, Liang C, Xue B, Gao Y, Yan T, Guo Y, et al. Local-prior–guided Segmentation and Visibility-normalized Photogrammetric Mapping of Stone-carved Heritage Defects. Photogrammetric Engineering & Remote Sensing. 2026. doi:10.14358/pers.26-00125r2

BibTeX
@article{Lin2026,
  author = {Lin, Xiaohu and Liang, Chenlu and Xue, Bei and Gao, Yinan and Yan, Tao and Guo, Yanan and He, Yuxiang},
  title = {Local-prior–guided Segmentation and Visibility-normalized Photogrammetric Mapping of Stone-carved Heritage Defects},
  journal = {Photogrammetric Engineering & Remote Sensing},
  year = {2026},
  doi = {10.14358/pers.26-00125r2}
}
Kostenlos starten →