Scholarly record

MOISTURE RISK MAPPING: MACHINE LEARNING APPROACHES USING UAV MULTISPECTRAL IMAGERY FOR CULTURAL HERITAGE MONITORING

Ivan Ivanov, Silviya Filipova, Stefan Vlaykov

Abstract

This study investigated a nondestructive, remote-sensing-based testing method for detecting harmful wetness in immovable cultural heritage structures using unmanned aerial vehicle multispectral imagery combined with machine learning algorithms. The multispectral data were captured over a historic building with DJI Mavic 3M UAV, and supervised machine learning classifiers were applied to identify moisture-risk areas across the building’s facades. The spectral range of the employed multispectral sensor did not allow direct observation of water absorption features. Therefore, moisture-related deterioration was assessed indirectly through spectral responses associated with biological colonization on building surfaces – an indicator of prolonged surface wetness. The classification results demonstrated successful identification of moisture-prone areas, which were later validated with in-situ measurements collected using a calibrated moisture meter, confirming strong spatial agreement between predicted and observed zones. In contrast to existing studies that rely on short-wave infrared or thermal sensors, that are often costly and limited to ground-based platforms, the proposed method enabled efficient and cost-effective monitoring of large historic buildings and building complexes. These findings establish the potential of the proposed method as a scalable and practical tool for moisture risk mapping and condition assessment in immovable cultural heritage.

Publication details

Back to publication list