Ground-Based Remote Sensing and Machine Learning for in Situ and Noninvasive Monitoring and Identification of Salts and Moisture in Historic Buildings

Abstract

Historical buildings are prone to deterioration due to moisture and salt activity. Salt weathering affects the appearance of monuments, resulting in mechanical degradation. Many laboratory-based studies have been performed focusing on understanding salt formation in building materials and the resulting damage mechanisms. However, large-scale in situ monitoring is necessary to understand salt activity in realistic situations. Here, we present a novel methodology for in situ and noninvasive identification and monitoring of moisture and salts, following a complementary remote sensing approach. The study is based on ground-based remote short-wave infrared (SWIR) spectral imaging and remote Raman spectroscopy at stand-off distances of order 10 m. SWIR spectral imaging was used for scanning large wall surfaces at high resolutions (angular resolution of 45 μrad), which gave spatial distributions of moisture and salts in their various hydration states, visualized using an artificial neural-network based spectral clustering method. Remote Raman spectroscopy in each cluster area confirmed the identification of the salts.