Material-Segmented Per-Pixel Emissivity Correction for Thermographic Anomaly Detection in Cultural Heritage Digital Twins

2026-08-03

AI Summary

This study addresses the limitation of conventional thermal imaging, which assumes a globally constant emissivity and consequently introduces spurious temperature artifacts on heterogeneous cultural heritage surfaces, obscuring subsurface anomalies. To overcome this, the authors propose a training-free, material-aware thermal imaging pipeline that leverages open-vocabulary semantic segmentation (SAM 3.1) to identify material regions from RGB images. By integrating literature-based emissivity values with co-registered RGB–long-wave infrared (LWIR) data, the method performs per-pixel emissivity correction and temperature retrieval via the inverse Planck’s law. This work presents the first integration of open-vocabulary segmentation with a physics-based thermal radiation model, reducing the mean absolute error from 1.97 K to 0.91 K under a 20 K temperature difference in synthetic data. Real-world validation demonstrates that correction primarily enhances accuracy in low-emissivity regions, clarifying the method’s applicability limits for outdoor weathered heritage.

Abstract

Quantitative longwave thermography of heritage surfaces is limited by the global-constant emissivity assumption in inverse-Planck temperature retrieval; on heterogeneous surfaces emissivity varies within one field of view, producing apparent-temperature artifacts that mimic and mask subsurface anomalies. We present a training-free pipeline that derives per-pixel emissivity by applying SAM 3.1 open-vocabulary segmentation to a colocated, co-calibrated RGB channel, mapping segments to a material-keyed LWIR emissivity table compiled from primary measurement literature, and propagating the field into a per-pixel inverse-Planck solve on raw radiometric data. Lacking any public dataset with raw radiometry, a temperature reference, and a colocated RGB camera, we evaluate on a physics-based synthetic benchmark and four real datasets. On the benchmark, under a palette spanning the low-emissivity exceptions, the correction cuts mean absolute error from 1.97 K to 0.91 K at 20 K contrast and, with an accurate table, beats the best fitted global constant on every layout; on a heritage-realistic emissivity distribution it does not. We contribute a quantified operating-regime map, and a measurement-backed finding that tempers the heritage claim: weathered outdoor heritage emissivities cluster near the conventional default, so the correction is small on typical surfaces and concentrated on genuine low-emissivity exceptions. We characterize the dominant failure mode, in which open-vocabulary segmentation matches appearance rather than material, and the contraindicated regime in which emissivity-defined anomalies are suppressed.

Problem

• emissivity variation
• thermographic anomaly detection
• cultural heritage
• heterogeneous surfaces
• temperature retrieval artifacts

Innovation

• per-pixel emissivity correction
• open-vocabulary segmentation
• thermographic anomaly detection
• cultural heritage digital twins
• inverse-Planck temperature retrieval