A reliability-oriented decision-support framework for pigment indication on sooty murals from VNIR–SWIR reflectance spectra

Soot can cause strong attenuation and baseline distortion in the reflectance spectra of mural pigments. We present a reliability-oriented decision framework for interpreting VNIR–SWIR reflectance spectra using normalized first derivative features. Channel-specific effective identification thresholds for VNIR and SWIR are derived from robust statistics of soot baselines. Spectra that pass the threshold are matched with channel-specific algorithms (UASC for VNIR saliency concordance and DDSC for SWIR dual-domain shape correlation) and optionally fused when both channels are informative. A confidence threshold reports High-confidence indication, low-confidence indication, or unidentifiable outcome. Laboratory mock-ups and two mural case studies using VNIR hyperspectral imaging endmembers demonstrate that the framework improves the interpretability of spectral features under soot and provides actionable guidance for targeted follow-up.