Inversion of Salt Content in Simulated Mural Based on Hyperspectral
Mural Salt Index
GUO Zhou-qian1, 2, LÜ Shu-qiang1, 2, HOU Miao-le1, 2*, SUN Yu-tong1, 2, LI Shu-yang1, 2, CUI Wen-yi1
1. School of Geomatics and Urban Spatial Informatics, Beijing University of Civil Engineering and Architecture, Beijing 102616, China
2. Beijing Key Laboratory for Architectural Heritage Fine Reconstruction & Health Monitoring, Beijing 102616, China
Abstract:Disruption and salt efflorescence are mainlycaused by the soluble salts found in the mural paintings, which are irreversible and hurt the frescoes' health. Quantifying the inversion of soluble salts contained in murals using non-contact hyperspectral techniques is significant. A mural salt content inversion model based on a hyperspectral salt index is proposed to address the problems of high cost, low timeliness, and the need for field sampling for mural salt detection. Indoors, mock mural samples with varying salt content gradients (salt soil ratio: 0~1%) were made with yellow sandy soil and wheat straw fine hemp mixed with anhydrous Na2SO4. The ASD-FieldSpec4HI-RES spectrometer was used for spectral acquisition. The sample spectral set was created after break pointcorrection and averaging, which was randomly divided into a calibration subset and validation subset in the ratio of 7∶3. The original reflectance (R) was subjected to four enhancement processes: first-order differentiation (R+1D), first-order differentiation after Continuum Removal (CR+1D), first-order differentiation after Logarithm Reciprocal (LR+1D), and first-order differentiation after Savitzky-Golay smoothing (SG+1D). The original reflectance and enhanced spectral data were correlated with the salt concentration, and the top three strong correlation bands with high contribution were extracted. Single-band regression models were established by linear and parabolic fitting with the strongest correlation bands, respectively. The first three correlation bands were used to create a hyperspectral mural salinity index (MSI) which was then compared to the normalized salinity index (NDSI), three salinity indices (SI1 SI2 SI3), and the brightness index (BI) for accuracy evaluation. The evaluation metrics are the coefficient of determination (R2), root mean square error (RMSE), and slope of the fit scatter line with intercept. The results show: (1) with the increase of salt content, the overall reflectance spectral curve first decreases and then increases. The reflectance is lowest in the range of salt content of 0.3%~0.6% for the mural samples. (2) The sensitive bands of Na2SO4 in the murals are 1 420, 1 940 and 2 210 nm, and there are also some sensitive bands in the visible range. (3) The first-order derivative transformed spectra are strongly correlated to salt concentration, with the highest R2 enhancement of 0.646. (4) The R-1D-MSI inversion model has the highest accuracy, with R2C and RMSEC of 0.857 and 0.116, respectively. This study can provide a new technical means for the rapid and nondestructive detection of salt content in murals.
国洲乾,吕书强,侯妙乐,孙宇桐,李淑阳,崔雯漪. 基于高光谱壁画盐分指数的模拟壁画盐含量反演[J]. 光谱学与光谱分析, 2023, 43(10): 3272-3279.
GUO Zhou-qian, LÜ Shu-qiang, HOU Miao-le, SUN Yu-tong, LI Shu-yang, CUI Wen-yi. Inversion of Salt Content in Simulated Mural Based on Hyperspectral
Mural Salt Index. SPECTROSCOPY AND SPECTRAL ANALYSIS, 2023, 43(10): 3272-3279.