Hyperspectral Image Classification of Pigments Based on Multiscale
Spatial-Spectral Feature Extraction
TANG Bin1, LUO Xi-ling1, WANG Jian-xu1, FAN Wen-qi2*, SUN Yu-yu1, LIU Jia-lu2, TANG Huan2, ZHAO Ya3*, ZHONG Nian-bing1
1. Chongqing University of Technology Key Laboratory of Optical Fiber Sensing and Photoelectric Detection in Chongqing, Chongqing 400054, China
2. Key Scientific Research Base of Pest and Mold Control of Heritage Collection (Chongqing China Three Gorges Museum), State Administration of Cultural Heritage, Chongqing 400060, China
3. School of Economics and Management, Chongqing Jiaotong University, Chongqing 400074, China
Abstract：Pigments not only endow cultural relics with color and aesthetic value but also carry rich historical, cultural, and technical information. Accurate classification and identification of pigments are essential for the restoration, preservation, and academic study of ancient painted artworks. Identifying the types and compositions of pigments helps determine the creation period, regional characteristics, and craftsmanship style, providing scientific guidance for restoration and cultural value research. However, traditional pigment analysis faces challenges due to limitations in sample size, surface flatness, and the destructive nature of some analytical methods, which may cause irreversible damage to the artifacts. Hyperspectral Imaging (HSI), with its non-destructive nature, wide-area scanning, and capability of capturing complete spectral information, has become a powerful tool for pigment detection. HSI overcomes the limitations imposed by uneven surfaces and small sample sizes, enabling the extraction of fine-grained spectral and spatial features from pigments. This study aims to utilize HSI for the precise classification and detailed feature extraction of ancient painting pigments, addressing challenges in complex scenarios. We propose a multi-scale spatial-spectral feature fusion method to integrate information at different levels. A spectral-spatial attention mechanism is employed to capture fine details. At the same time, the Vision Transformer (ViT) extracts high-level semantic information from the entire image, enhancing the representation of complex pigment features and improving classification performance. The experimental results show that the proposed method significantly outperforms traditional and other deep learning models in the classification of simulated painting samples: it improves classification accuracy by 34.35% compared to the Support Vector Machine (SVM) and by 8.93% and 5.6% compared to HyBridSN and Spectral-Spatial Residual Network (SSRN), respectively. This study not only improves the accuracy of pigment detection but also provides non-destructive, reliable technical support for the scientific restoration and cultural value preservation of ancient paintings, contributing to the intelligent development of cultural heritage conservation.
汤 斌，罗希玲，王建旭，范文奇，孙玉宇，刘家路，唐 欢，赵 雅，钟年丙. 基于多尺度空间-光谱特征提取的颜料高光谱图像分类方法[J]. 光谱学与光谱分析, 2025, 45(08): 2364-2372.
TANG Bin, LUO Xi-ling, WANG Jian-xu, FAN Wen-qi, SUN Yu-yu, LIU Jia-lu, TANG Huan, ZHAO Ya, ZHONG Nian-bing. Hyperspectral Image Classification of Pigments Based on Multiscale
Spatial-Spectral Feature Extraction. SPECTROSCOPY AND SPECTRAL ANALYSIS, 2025, 45(08): 2364-2372.
[1] YANG Wen-zong, TANG Xing-jia, ZHANG Peng-chang, et al(杨文宗, 唐兴佳, 张朋昌, 等). Sciences of Conservation and Archaeology(文物保护与考古科学), 2023, 35(4): 11.
[2] YIN Qin-li, WANG Xiao, GENG Dan, et al(尹琴丽, 王 笑, 耿 丹, 等). Study on Natural and Cultural Heritage(自然与文化遗产研究), 2019, 4(6): 108.
[3] LI Guang-hua, CHEN Yao, SUN Xue-jian, et al(李广华, 陈 垚, 孙雪剑，等). Infrared(红外), 2024, 45(2): 42.
[4] ZHANG Yan, SHEN Jin-yue, SHAO Yu-yi, et al（张 岩, 沈金悦, 邵钰奕, 等）. Computer & Digital Engineering(计算机与数字工程), 2024, 52(3): 898.
[5] TAN Yun-fei, LI Ming, LUO Yong-hang, et al(谭云飞, 李 明, 罗勇航, 等). Computer Technology and Devolopment(计算机技术与发展), 2024，34(9)：147.
[6] Ding Y, Zhao X, Zhang Z, et al. IEEE Geoscience and Remote Sensing Letters, 2022，19: 5504205.
[7] ZHONG Jia-ping, LI Yun-song, XIE Wei-ying, et al(钟佳平, 李云松, 谢卫莹, 等). Acta Electronica Sinica(电子学报), 2024, 52(5): 1716.
[8] SUN Yu-tong, LÜ Shu-qiang, HOU Miao-le(孙宇桐, 吕书强, 侯妙乐). Spectroscopy and Spectral Analysis(光谱学与光谱分析), 2024, 44(8): 2357.
[9] ZHAO Feng, GENG Miao-miao, LIU Han-qiang, et al(赵 凤, 耿苗苗, 刘汉强，等). Journal of Electronics & Information Technology(电子与信息学报), 2024, 46(5): 2237.
[10] YANG Xing, CHI Yue, ZHOU Ya-tong, et al(杨 星, 池 越, 周亚同, 等). National Remote Sensing Bulletin，2023, 27(11): 2565.
[11] Sun L, Zhao G, Zheng Y, et al. IEEE Transactions on Geoscience and Remote Sensing, 2022, 60: 5522214.
[12] Mou L，Zhu X X. IEEE Transactions on Geoscience and Remote Sensing, 2020, 58(1)：110.
[13] Mohla S, Pande S, Banerjee B, et al. FusAtNet: Dual Attention Based Spectro Spatial Multimodal Fusion Network for Hyperspectral and LiDAR Classification, 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, 2020.
[14] Wang Z, Chen J，Hoi S C H. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2021，43(10)：3365.
[15] Pal M, Foody G M. IEEE Transactions on Geoscience & Remote Sensing, 2010, 48(5): 2297.
[16] Zhong Z L, Li J, Luo Z M, et al. IEEE Transactions on Geoscience and Remote Sensing, 2018, 56(2): 847.
[17] Roy S K, Krishna G, Dubey S R，et al. IEEE Geoscience and Remote Sensing Letters, 2020, 17(2): 277.
