A mildew spectral index of ancient Chinese silk paintings based on hyperspectral imaging data

Painting and calligraphy possess significant historical value and play a crucial role in the transmission of human cultural heritage. However, the presence of mildew greatly impacts the preservation of painting and calligraphy, thereby affecting their cultural value and legacy. Traditional mildew detection methods rely on manual visual inspection and/or chemical analysis, which are limited by inefficiency, subjectivity, the need for nondestructiveness, and low accuracy. Here, we overcome these limitations by developing a new mildew spectral index (MSIndex) using hyperspectral imaging technology and the mildew characteristics of ancient Chinese silk paintings. This approach provides support for the rapid, accurate, and nondestructive extraction and identification of mildew in ancient Chinese silk paintings. We first analyzed the mildew spectra on ancient Chinese silk paintings and optimized the spectral characteristics of mildew based on the hyperspectral data. Then, using this analysis, we constructed the MSIndex to detect mildew. We tested the proposed mildew detection method on the hyperspectral dataset of Shen Qinglan Tieluo on Qing Dynasty (1796–1805), followed by the evaluation of its generalization ability using the hyperspectral dataset of the Portrait of Pañcika Arhat (dated 1756) as an independent validation set. The results suggested that the proposed MSIndex was robust and effective with an overall accuracy of 94.17 % in mildew detection. The MSIndex was also capable of detecting mildew regions even in complex environments, such as those involving other pigments or diseases. This method can help professionals make accurate restoration plans for ancient Chinese silk paintings and support the preservation of cultural heritage.