A Study On Mural Painting Pigment Identification Based On Spectral Database

China has created rich cultural achievements and scored great artistic success.Among them,the murals are the most gorgeous,with artistic value and historical significance.The pigment is the material carrier of mural expression,and it is also important for mural protection and research.Accurate identification of pigments is a prerequisite for heritage conservation,and it is of great significance for us to understand the skills of ancient paintings,pigment processing techniques,art and historical values of murals.Traditional sampling analysis methods can cause irreversible damage to precious mural artifacts,and hyperspectral technology has received extensive attention as a fast and non-destructive pigment identification method in recent years.At present,there is no comprehensive spectral database for ancient Chinese mural painting pigments,and there is no systematic research on pigment recognition.In this paper,through the establishment of a complete mural pigment spectral database,a spectral matching algorithm suitable for different pigment recognition scenes is proposed.Finally,a practical mural pigment spectral database system is developed.The main research results are as follows:1.Cooperating with the experts of the Shaanxi History Museum,we invested and determined 28 typical mural pigments that used in and before the Tang Dynasty.According to the name of these pigments,we collected 30 pigment samples.The reflectance spectra of the powder and painted pigment samples at 350-2500 nm were measured using an ASD FieldSpec 3 spectroradiometer,and the reflectance spectra of the painted pigment samples at 385-1032 nm were measured using a Dualix GaieField imaging spectrometer.And the pigment spectral library was established including three types of data: standard spectra,painted spectra and imaging spectra of these pigments.2.Based on the standard spectra in the library,spectral features were determined by analyzing the spectrum of every pigment.Investigating the effects of particle size and glue on the pigment spectrum and the relationship between the imaging spectrum and the standard spectrum,we get the following conclusions:(1)The effect of particle size on the pigment spectrum is only reflected in the amplitude but the shape of the curve remains unchanged.(2)The effect of glue on pigment spectrum is mainly reflected in the absorption within SWIR bands,and the characteristics within VNIR bands remain.(3)Except for white pigments,imaging spectrum is mainly affected by the purple edge effect,and the basic characteristics are still maintained.The invariant feature of pigment spectrum provides a theoretical basis for pigment identification in different scenes using standard spectra.3.Using the standard spectra as the reference,7 classical spectral matching algorithms were used to identify the powder,painted and imaging pigment spectral data.With the exception of CCSM(Cross Correlogram Spectral Mapping)and SFF(Spectral Feature Fitting),the remaining algorithms are well-recognized for powder pigments.However,these methods fail to identify the painted and imaging pigments due to differences from the standard spectra.For painted pigments,the overall recognition accuracy of SAM(Spectral Angle Mapper),SCM(Spectral Correlation Matching)and SID(Spectral Information Divergence)is about 40%,which is higher than the other 4 methods.For imaging pigments,they are also superior to other methods.Among them,SCM has the highest accuracy of 33%.Removing the SWIR bands of painted pigments and the purple edge bands of imaging pigments improves the recognition accuracy.However,black and white pigments having no spectral features could not be recognized.Therefore,taking into account the intrinsic spectral feature of pigments,we proposed the SFCM(Spectral Feature Correlation Matching)algorithm.SFCM performs well in the identification of the other 18 pigments.The overall recognition accuracy of powder pigments,painted pigments,and imaging pigments is 100%,97% and 95% respectively,and some pigments that are misrecognized before can be successfully identified.4.In order to facilitate researchers to apply hyperspectral technology to mural painting analysis,we developed a set of pigment spectral database system based on Qt C++.The system can manage and query all the pigments in the library,and provide basic functions such as the spectral display,analysis,and matching.It provides a powerful tool for relevant researchers to apply hyperspectral technology to achieve fast and nondestructive identification of mural painting pigment,improving the level of research and protection techniques of mural paintings.