Application Model of Hyperspectral Technology Based on Novel Spectral Indices for Salinity Assessment in Soil Heritage Sites

Fang Liu, Yikang Ren2024 · DOI

Abstract The Dunhuang murals are a precious treasure of China's cultural heritage, yet they have long been affected by salt damage. Traditional methods for detecting salt content are costly, inefficient, and may cause physical harm to the murals. Among current techniques for measuring salt content in murals, hyperspectral remote sensing technology offers a non-invasive alternative, circumventing issues of high costs and low efficiency. Building on this, our study developed a high-spectral feature inversion model for mural phosphate content using Fractional Order Differentiation (FOD), a novel three-band spectral index, and the Partial Least Squares Regression (PLSR) algorithm. The study's findings include: 1) Model performance reached its optimum at a 0.3 order of differentiation for both PSR and PNDI data, with a determination coefficient ($R^2$) of 0.728. 2) Utilizing PLSR with the optimal six-order three-band spectral index combination, we successfully constructed an inversion model with an $R^2$ of 0.815. This provides an effective technical means for monitoring the salt damage conditions of precious cultural heritage.

The Threat of Salt Damage to Cultural Heritage

The Dunhuang murals, located in the Mogao Caves, are a UNESCO World Heritage site under constant threat from "salt damage." Fluctuations in environmental humidity cause salts within the mural plaster to crystallize and expand. This physical process leads to severe structural issues, including efflorescence, cracking, and the eventual detachment of pigment layers.

Figure 1: Geographic location of Dunhuang City and visual evidence of salt damage on murals, including structural cracking (a, b) and pigment flaking (c, d).

To preserve these treasures, conservators require precise data on salt concentrations. However, traditional chemical analysis is destructive, requiring physical samples to be scraped from the murals. While hyperspectral remote sensing offers a non-invasive solution, the spectral signatures of salts are often faint and masked by the complex mineral composition of the mural plaster. This research addresses the need for a more precise mathematical framework to extract these subtle signals.

Methodology: Enhancing Spectral Sensitivity

The study implemented a three-tiered mathematical approach to improve detection accuracy:

  1. Fractional Order Differentiation (FOD): Unlike standard first or second-order derivatives, FOD uses fractional steps (e.g., 0.1, 0.2, 0.3). This acts as a "fine-tuning knob," capturing subtle spectral variations that integer derivatives might smooth over or obscure with noise.
  2. Phosphate Three Simple Ratio (PTSR): The authors developed a new three-band index specifically for phosphates. By incorporating a third sensitive waveband ($R_c$), the model significantly improved its resistance to environmental background interference compared to traditional dual-band indices.
  3. Partial Least Squares Regression (PLSR): To manage the high dimensionality of hyperspectral data, PLSR was used to construct the final inversion model, utilizing the best-performing spectral indices as explanatory variables against mural plaster electrical conductivity.

Experimental Setup and Data Collection

To avoid damaging actual heritage artifacts, the researchers created simulated mural plaster samples using a mixture of kaolinite clay, sand, and wheat straw, mimicking the original materials of the Mogao Caves. These samples were treated with five different concentrations of disodium hydrogen phosphate dodecahydrate.

Figure 4: Experimental workflow including (a) sample preparation in high-precision containers, (b) samples showing varying levels of salt erosion, and (c) the data collection process for spectral and conductivity measurements.

Reflectance data was captured using an ASD FieldSpec 4 Hi-Res spectroradiometer (350–2500 nm) under controlled conditions. The study identified a "sweet spot" for data processing at the 0.3 order of differentiation, suggesting that traditional calculus methods are too aggressive for soil heritage data and may destroy useful information.

Key Results and Model Performance

The integrated PTSR-FOD-PLSR model demonstrated superior performance over traditional methods:

  • Optimal Preprocessing: Applying 0.3 order FOD to dual-band indices (PSR and PNDI) yielded an $R^2$ of 0.728.
  • Final Model Accuracy: The full three-band inversion model achieved an $R^2$ of 0.815 and a Root Mean Square Error (RMSE) of 0.327.
  • Improvement: This approach represents a 10.4% improvement in accuracy compared to previous models, providing a robust, non-contact tool for quantitative salt assessment.

Conclusion and Future Directions

While the PTSR-FOD-PLSR model provides a powerful new tool for heritage conservation, several challenges remain for field application. Future research must bridge the "lab-to-field" gap by accounting for the complex mineral pigments found on real murals, which may interfere with plaster signatures. Additionally, the researchers suggest that integrating machine learning and deep learning could further enhance computational efficiency and accuracy for real-time monitoring at heritage sites.

Summary updated

Abstract

The Dunhuang murals are a precious treasure of China's cultural heritage, yet they have long been affected by salt damage. Traditional methods for detecting salt content are costly, inefficient, and may cause physical harm to the murals. Among current techniques for measuring salt content in murals, hyperspectral remote sensing technology offers a non-invasive , circumventing issues of high costs, low efficiency. Building on this, our study developed a high-spectral feature inversion model for mural phosphate content using Fractional Order Differentiation (FOD), a novel three-band spectral index, and Partial Least Squares Regression (PLSR) algorithm. The specific research contents include: 1) Exploring the absorption mechanism of phosphates and their characteristic bands, combined with the optimal spectral index to construct a univariate linear regression model, providing a basis for rapid quantitative measurement of mural phosphate content. 2) By comparing the accuracy of the PSR and PNDI spectral indices based on the linear regression model, the first six orders of the highest accuracy spectral index were selected as the optimal three-band spectral index combination, used as explanatory variables, with mural plaster electrical conductivity as the response variable, employing the PLSR method to construct the mural phosphate content high-spectral feature inversion model. The study's findings include: 1) By comparing the outcomes of different orders of fractional differentiation, it was found that the model performance reached its optimum at a 0.3 order of differentiation for both PSR and PNDI data, with a determination coefficient (R2) of 0.728. 2) Utilizing PLSR, this study employed the previously determined optimal six-order three-band spectral index combination as explanatory variables, with salt content as the response variable, successfully constructing the high-spectral feature inversion model for mural phosphate content with a determination coefficient (R2) of 0.815. This provides an effective technical means for monitoring the salt damage conditions of precious cultural heritage such as murals.
BibTeX
@inproceedings{liu2024application,
  author = {Fang Liu and Yikang Ren},
  title = {{Application Model of Hyperspectral Technology Based on Novel Spectral Indices for Salinity Assessment in Soil Heritage Sites}},
  booktitle = {˜The},
  year = {2024},
  doi = {10.5194/isprs-archives-xlviii-2-2024-477-2024},
}
Figure