基于嵌套U型结构网络的古代壁画破损区域检测

doi:10.3778/j.issn.1002-8331.2309-0069

摘要/Abstract

摘要： 由于年代久远和自然因素等影响，许多古代壁画表面存在着不同程度的损坏，精准地检测这些破损区域对于保护文化遗产具有重要研究价值。然而，现有的方法在古代壁画的破损检测中存在检测不全，对微小细节的破损区域检测不准确等问题。针对以上问题，提出基于嵌套U型结构网络的壁画破损区域检测方法U2-DUANet，该方法采用了一种新的深度监督聚合模块，能够更有效地融合侧边输出的细节信息。提出了一种像素级上下文及通道注意力（pixel-wise contextual and channel attention，PCCA）机制，来选择性地关注每个像素上的信息上下文特征以及通道信息，更精确地捕捉图像的重要特征，提高模型的准确性。自主创建了一个古代破损壁画图像的数据集，并对其破损区域进行了人工标注。在壁画数据集上的实验结果表明，U2-DUANet相比于U-Net，在精确率和F-measure上分别提升了10.5个百分点和8.2个百分点，并且具有更好的鲁棒性和泛化能力。

关键词: 嵌套U型结构, 深度监督聚合, 注意力机制, 破损区域检测

Abstract: Due to the effects of long history, natural factors, etc., many ancient murals have varying degrees of damage on their surfaces. Accurately detecting these damaged areas holds significant research value for the preservation of cultural heritage. However, the existing methods have some problems in detecting the damage of ancient murals, such as incomplete detection and inaccurate detection of damaged areas in minute details. To solve this problem, a new method for detecting damaged areas of murals, U2-DUANet based on nested U-structure network, is proposed. This method uses a new deeply supervised aggregation module, which can more effectively fuse the side output details. An attention module PCCA(pixel-wise contextual and channel attention) is designed to selectively attend to informative context locations and channel information at each pixel. Important features of the image can be captured more precisely, improving the accuracy of the model. A data set of ancient damaged murals is created independently and their damaged areas are manually labeled. The experimental results on the murals dataset show that compared with U-Net, U2-DUANet improves the precision and F-measure by 10.5 percentage points and 8.2 percentage points, meanwhile, U2-DUANet has better robustness and generalization ability.

Key words: nested U-structure, deeply supervised aggregation, attention mechanism, damaged areas detection

陈安, 余映, 赵辉荣, 王信超. 基于嵌套U型结构网络的古代壁画破损区域检测[J]. 计算机工程与应用, 2025, 61(2): 283-292.

CHEN An, YU Ying, ZHAO Huirong, WANG Xinchao. Damaged Areas Detection of Ancient Murals Based on Nested U-Structure Network[J]. Computer Engineering and Applications, 2025, 61(2): 283-292.

参考文献

[1] 雷桐, 魏晓阳. 数字媒介环境下陕西寺观壁画的数字化保护与传播研究[J]. 新媒体研究, 2022, 8(22): 25-27.
LEI T, WEI X Y. Research on digital protection and dissemination of temple murals in Shaanxi under digital media environment[J]. New Media Research, 2022, 8(22): 25-27.
[2] 陈永, 艾亚鹏, 郭红光. 改进曲率驱动模型的敦煌壁画修复算法[J]. 计算机辅助设计与图形学学报, 2020, 32(5): 787-796.
CHEN Y, AI Y P, GUO H G. Inpainting algorithm for Dunhuang mural based on improved curvature-driven diffusion model[J]. Journal of Computer-Aided Design & Computer Graphics, 2020, 32(5): 787-796.
[3] 胡颖. 壁画类文物保护技术研究[J]. 东方收藏, 2022(5): 104-106.
HU Y. Study on protection technology of mural cultural relics[J]. Oriental Collection, 2022(5): 104-106.
[4] 梁涛, 夏国芳, 胡春梅. 壁画病害边缘自动提取方法研究[J]. 石窟与土遗址保护研究, 2022, 1(1): 61-69.
LIANG T, XIA G F, HU C M. Research on automatic edge extraction of mural diseases[J]. Research on the Conservation of Cave Temples and Earthen Sites, 2022, 1(1): 61-69.
[5] 张子迎, 税午阳, 周明全, 等. 数字壁画病害提取与修复算法研究[J]. 计算机应用研究, 2021, 38(8): 2495-2498.
ZHANG Z Y, SHUI W Y, ZHOU M Q, et al. Research on disease extraction and inpainting algorithm of digital grotto murals[J]. Application Research of Computers, 2021, 38(8): 2495-2498.
[6] DENG X C, YU Y. Automatic calibration of crack and flaking diseases in ancient temple murals[J]. Heritage Science, 2022, 10(1): 1-17.
[7] 张叶娥, 吴利刚. 基于深度学习的石窟壁画破损检测[J]. 云冈研究, 2022, 2(1): 85-90.
ZHANG Y E, WU L G. The detection of cave mural damage based on deep learning[J]. Yungang Research, 2022, 2(1): 85-90.
[8] 吴利刚, 张梁. 基于轻量化神经网络的石窟壁画破损检测方法[J]. 信息与控制, 2024, 53(1): 108-119.
WU L G, ZHANG L. Damage detection method for grotto murals based on lightweight neural network[J]. Information and Control, 2024, 53(1): 108-119.
[9] HUANG R, FENG W, FAN M Y, et al. Learning multi-path CNN for mural deterioration detection[J]. Journal of Ambient Intelligence and Humanized Computing, 2017, 11: 3101-3180.
[10] 罗启明, 吴昊, 夏信, 等. 基于Dual Dense U-Net的云南壁画破损区域预测[J]. 图学学报, 2023, 44(2): 304-312.
LUO Q M, WU H, XIA X, et al. Damaged areas in Yunnan murals prediction using Dual Dense U-Net[J]. Journal of Graphics, 2023, 44(2): 304-312.
[11] 吕书强, 王诗涵, 侯妙乐, 等. 基于改进U-Net的壁画颜料层脱落病害区域提取[J]. 地理信息世界, 2022, 29(1): 69-74.
LYU S Q, WANG S H, HOU M L, et al. Extraction of mural paint loss diseases based on lmproved U-Net[J]. Geomatics World, 2022, 29(1): 69-74.
[12] BODNAR J L, MOUHOUBI K, VALLET J M. Examples of SVD decomposition contributions to the non-destructive testing of cultural heritage mural paintings using stimulated infrared thermography[J]. The European Physical Journal Applied Physics, 2022, 97: 1-10.
[13] LUO P, ZHANG R, REN J, et al. Switchable normalization for learning-to-normalize deep representation[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2021, 43(2): 712-728.
[14] CHEN S, TAN X, WANG B, et al. Reverse attention for salient object detection[C]//Proceedings of the 15th European Conference on Computer Vision, 2018: 236-252.
[15] ZHOU Z, SIDDIQUEE M M R, TAJBAKHSH N, et al. UNet++: a nested U-Net architecture for medical image segmentation[J]. IEEE Transactions on Medical Imaging, 2019, 39(6): 1856-1867.
[16] CHEN Z, XU Q, CONG R, et al. Global context-aware progressive aggregation network for salient object detection[C]//Proceedings of the AAAI Conference on Artificial Intelligence, 2020: 10599-10606.
[17] HE K, GKIOXARI G, PIOTR D, et al. Mask R-CNN[C]//Proceedings of the IEEE International Conference on Computer Vision, 2017: 2980-2988.
[18] SAINING X, ZHUOWEN T. Holistically-nested edge detection[C]//Proceedings of the IEEE International Conference on Computer Vision, 2015: 1395-1403.
[19] LIU Y, CHENG M M, ZHANG X, et al. DNA: deeply-supervised nonlinear aggregation for salient object detection[J]. IEEE Transactions on Cybernetics, 2022, 52(7): 6131-6142.
[20] WOO S, PARK J, LEE J Y, et al. CBAM: convolutional block attention module[C]//Proceedings of the 15th European Conference on Computer Vision, 2018: 3-19.
[21] LIU N, HAN J, YANG M H. PiCANet: pixel-wise contextual attention learning for accurate saliency detection[J]. IEEE Transactions on Image Processing, 2020, 29: 6438-6451.
[22] QIN X, ZHANG Z, HUANG C, et al. U2-Net: going deeper with nested U-structure for salient object detection[J]. Pattern Recognition, 2020, 106: 107404.
[23] LIU J J, HOU Q, CHENG M M, et al. A simple pooling-based design for real-time salient object detection[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2019: 3912-3921.