NMF在古书画混合颜料光谱分析中的应用

Application of NMF in the Analysis of Mixed Pigments in Ancient Calligraphy and Painting

颜料分析是古书画保护修复的重要环节。高光谱技术由于其快速、无损等特点广泛应用在颜料分析中,但混合颜料分析仍然是难点。文章提出了一种基于非负矩阵分解(non-negative matrix factorization,NMF)的混合颜料光谱分析算法。该算法将混合颜料光谱视为多源信号的线性混合,首先利用NMF将混合光谱分离为若干个端元光谱,然后将端元光谱与颜料光谱库进行对比,确定端元光谱的颜料种类。模拟仿真数据以及实测混合颜料样本数据的实验结果表明,该算法能够准确地分离出混合颜料光谱的组成成分,为古书画混合颜料种类的鉴定提供依据。

Pigment analysis plays an important role in the protection and restoration of ancient calligraphy and painting. Hyperspectral technology is widely used in pigments analysis because of its fast and non-destructive characteristics, but the mixed pigment analysis is still difficult. A new spectral analysis algorithm for mixed pigments based on Non-negative Matrix Factorization (NMF)is proposed in this paper. In this algorithm, the mixed pigments spectrum is regarded as a linear mixing of multi-source signals. First, the mixed spectrum is separated into several end-element spectra by NMF, and then the end-element spectra are compared with the pigments spectral library to determine the types of pigments in the end-element spectra. The experimental results of simulated data and measured sample data of mixed pigments show that the algorithm can accurately separate the spectral components of mixed pigments, which can provide a basis for the identification of mixed pigments in ancient calligraphy and painting.