Research On Pigment Spectral Subsection Identification And Mapping On Surface Of Painted Artifacts

Due to environmental changes and human activities,the color on surface of painted artifacts will appear to be lost or faded,which needs digital preservation and color restoration as soon as possible.The key step for color reproduction is to judge the type of pigment used in the area to be repaired.Therefore,it is urgent to use modern scientific techniques to determine the type of pigment on the surface of painted artifacts,for retaining pigment information,providing scientific basis for restoration workers so as to improve their efficiency.In order to avoid causing secondary damage to cultural relics,the hyperspectral technology,with the advantages of non-destructive,was used for obtaining and analyzing the full-surface pigment information of painted artifacts in this paper.What's more,a typical pigment spectral library was established specifically for Chinese painting,which is helpful for the digital archive and sustainable preservation of cultural heritage.According to pigment usage and painting skills,artists often mix multiple pigments for brilliant colors,which leads to the result that the colors on Chinese paintings or frescoes observed by our naked eyes or instruments such as hyperspectral cameras are usually a mixture of multiple pigments.Most of commonly used method for spectral identification can only identify one pigment at one time,or match them after spectral unmixing,which is more complicated and requires some prior knowledge.Therefore,a spectral subsection identification approach considering ion absorption characteristics was proposed in this paper for simultaneous determining multiple pigments contained in mixed area on surface of paintings.Based on the hyperspectral images and reflection spectra collected from a Chinese painting and a fresco,the "four-step process",including endmember estimation,image clustering,spectral identification,and pigment mapping,was proposed in this paper to achieve the pigment type determination and the visualization of their spatial distribution on painting surface.The main research results are as follows:(1)Establishing the Typical Pigments of Paintings Spectral Library.Currently,the re are few spectral libraries specially constructed for typical pigments on Chinese painting surface.The typical pigments commonly used in Chinese painting were selected for samples making.The electron microscope,Raman spectrometer,scanning electron microscope,and portable X-ray fluorescence spectrometer were utilized for chemical analysis of the pigments,and the spectroradiometer was used to collect reflection spectra covering wavelength from 350 to 2500 nm,to establish typical pigment spectral library so as to be the reference of spectral identification.Furthermore,the spectral data of pigments collected from real cultural relics were also recorded in the spectral library,which made it possible to establish digital archives and realize sustainabl e preservation for these precious cultural relics.(2)Proposing the spectral subsection identification method considering ion absorption characteristics.In this paper,the spectral absorption positions of the main ions and functional groups contained in typical pigments were summarized firstly as a basis for dividing feature subsection.Secondly,these subsections were extracted from the whole unknown spectrum to perform spectral absorption coefficient matching respectively,to ensure the accurate positioning and not to miss the key absorption characteristics.And then,weights were assigned to each subsection result and the weighted indices of each pigment were calculated.Finally,the identification results were determined by a threshold.The results shown that the recognition performance of the proposed method is better than that using the whole spectrum for spectral matching,where the former one can correctly identify the main components of mixed pigments.In addition,compared with those spectral unmixing method,the subsection identification method is simpler in material recognize.(3)Proposing the "four-step process" for pigment identification and mapping.Firstly,based on the eigenvalues of hyperspectral image,the endmember number of painting was estimated,which was the reference for initial value setting before image clustering.Secondly,the unsupervised classification was performed to find homogeneous regions on the image.And then,according to the clustering results,some reflection spectra in each homogeneous region were collected,respectively,and the pigment types contained in each mixture were recognized by spectral identification methods.Finally,the pigment identification results were mapped on the classification image.The "four-step process" can presented the types and spatial distribution of pigments in each mixed area more intuitively.