Automated Signal Processing for Multimodal Spectroscopic Pigment Identification in Cultural Heritage Diagnostic

Reliable and rapid interpretation of spectroscopic data is a key requirement for the development of smart diagnostic systems in cultural heritage and material analysis. This work presents an automated and intelligent signal-processing tool for the identification of historical pigments, integrating visible reflectance spectrometry and Raman spectroscopy. Although these techniques provide complementary chromatic and molecular information, conventional workflows remain largely manual, time-consuming, and strongly dependent on operator expertise. The proposed system operates entirely downstream of analysis on raw data and is designed to minimize user intervention while ensuring robustness and repeatability. The study involved 100 historical pigments from the ®KREMER Pigmente collection, prepared according to traditional recipes. After acquisition, the workflow automatically performs signal processing and feature extraction through custom Python pipelines specifically tailored to each analytical technique. Reflectance data are processed using Savitzky-Golay smoothing, derivative-based maxima identification, and CIELab1976 colorimetric conversion. Raman spectra undergo Asymmetric Least Squares (ALS) baseline correction, Savitzky-Golay smoothing, and first-derivative-based peak detection. The heterogeneous information-colorimetric parameters, maxima of the first derivative of the spectral reflectance factor (SRF%), and Raman peaks-is encoded into compact and standardized descriptors in the form of structured binary barcode-like images, enabling efficient comparison between unknown samples and a reference dataset. The workflow demonstrated its capability to automatically extract diagnostic features and provide rapid feedback for material identification, representing a concrete step toward intelligent, non-invasive, and real-time diagnostic tools for cultural heritage applications.