Research On Non-destructive Identification Method Of Mural Paint Based On Visible Spectrum

To recognize the pigment composition of paintings is a hot issue in image reproduction and cultural relic protection filed.The traditional three stimulus expression method cannot eliminate material composition objectively without the influence of other substances.Visible reflectance spectrum represents the characteristic of the object's absorption to light,which is the essential information of the color of object surface.Different objects have different physical and chemical structure,therefore different spectral information can be applied to recognize the type and composition of objects.Nondestructive identification method based on visible spectrum can be used to analyze the component without destroying the object.A nondestructive method based on feature extraction of visible spectrum to recognize mural pigments is proposed in this paper.Then construct a nondestructive identification system and analysis the pigment composition of Dunhuang murals.Using the non-contact type spectrophotometer PR705 in situ to obtain visible spectrum of Dunhuang mural pigments and the color patch of standard pigments nondestructively.The negative eigenvalue of the second order derivative of spectral curve was extracted as feature values to present the peaks.In order to highlight the most important feature we used median pyramid to derive three resolution representations of the feature curves.Then normalize these three curves and combined as the feature space F?F₁,F₂,F₃?.The similarity between sample points of murals and standard pigments samples was calculated by K nearest neighbor algorithm and Euclidean distance to recognize the information of murals.Finally,a recognition system was constructed according the proposed algorithm to realize automatic application of the recognition method.The research results show that the method based on feature extraction of visible spectrum can recognize the component of mural pigments automatically,thus can provide valuable information for restoration and digital archiving of murals.