Extraction And Evaluation Of The Disease Of The Mural Paint Loss

Known for its huge color murals inside the temple,the Temple of qutan is known as the“Northwest Small Forbidden City”and is the second batch of national key cultural relics protection units.The temple murals were made by the court masters of the Ming and Qing Dynasties.The number of the murals is rich,the pictures are rich in content and the painting skills are exquisite.It can be described as the precious art of the temple.According to field investigations,due to the influence of natural factors and human factors,some murals in the 51 wall gallery in the temple have been fading,cracking,curling,lifting,and the paint loss,these diseases seriously affect the cultural value and artistic value of the mural paintings of the temple.Especially when the pigment layer falls off,it is undoubtedly an irreversible loss for non-renewable murals.Therefore,in the process of repairing murals,the identification and evaluation of mural diseases becomes very important.However,the traditional method of using artificial direct search for disease is time-consuming and labor-intensive.Therefore,in terms of the identification,extraction,analysis and evaluation of the paint loss disease,the use of modern computer technology can improve the efficiency and avoid the secondary damage that the mural painting may suffer,which is of great significance for the permanent retention of the mural paintings.The data in this paper is the hyperspectral image data of the West Corridor,and after the data preprocessing and image disease information extraction.Regression analysis in mathematical statistical model is used to evaluate the severity of mural hyperspectral image data,which provides a reference for mural restoration work.The specific work is as follows:(1)Establishment of mean spectrum and endmember spectrum.For the paint loss disease of the mural,using the difference of the spectral curves of the shedding area and other areas,based on the high-resolution spectral features of the hyperspectral image,two techniques are explored to establish the mean and endmember spectra of different pigments respectively.The former uses the spectral information of different pigments on the hyperspectral image to obtain the spectral curves of different pigment targets,and establishes the mean spectrum;the latter uses the Sequential Maximum Angle Convex Cone(SMACC)method to extract the endmember spectral information of different pigments,and establishes the endmember spectrum.The two kinds of spectral information provide spectral information training samples for the identification and extraction of the disease of the mural paint loss,which promotes the application of hyperspectral technology in mural cultural relics information extraction.(2)Disease information extraction in the area where the paint loss.Based on the spectral information and the end-spectrum spectra,using the different thresholds of spectral angle mapping(SAM)to classify and extract the paint lossdiseases,experiments show that the SAM setting threshold is 0.1,which is better.At the same time,the spectral information is combined with the spatial information to avoid the influence of the same spectrum of different substances,and the region of interest(ROI)is selected from the image as the training sample.Using SAM method and Support Vector Machine(SVM)method to identify the classification of mural diseases respectively,the experiment proves that the SVM classification method has a good effect on the extraction of paint loss diseases,so the SVM method is used to identify and extract the paint loss diseases in other experimental areas,and prepare the basic data for the quantitative evaluation of the disease degree in the next step.(3)Evaluation of the degree of the paint loss.The degree of damage to the murals is closely related to the damaged area,quantity and density of the disease.According to the analysis,based on the vector data of paint loss,the Arcgis software is used to obtain the quantitative data that affects the damage factors of the mural.By analyzing the correlation between different influencing factors and proposing the indicators,they are the damage rate index(DRI),the density index(DI),the boundary density index(BDI)and the fragmentation index(FI).The Least Squares(LS)method was used to select four index elements as independent variables for multivariate regression modeling,and the regression equation and regression coefficients were tested for significance.Using the step regression method for the insignificant variables,the model with the smallest AIC value was selected by akaike information criterion(AIC)as the optimal estimation model for the mural paint loss disease index,which was established by two index elements of DRI and DI.The research shows that the model with two index factors as independent variables is the best with the AIC criterion.The R~2 and standard error of the model adjustment are 0.9314 and 0.0526,respectively.And use experimental data to verify the regression equation,the results show that it is feasible to use the LS-SR-AIC method to analyze the disease index of the mural paint loss.It provides an effective method for the evaluation of the disease degree of murals,and provides a scientific basis for the later scientific workers to repair murals.