Decay Detection and Classification on Architectural Heritage Through Machine Learning Methods Based on Hyperspectral Images: An Overview on the Procedural Workflow

Imaging spectroscopy is a well-established technology that allows non-destructive remote analysis of objects in order to detect defects or imperfections in a wide range of the electromagnetic spectrum. In the field of cultural heritage, and especially in the architectural one, the interest for its application is increasing, since it allows to carry out decay assessment surveys in a more accurate way. In combination with machine learning (ML) techniques, hyperspectral imaging (HSI) allows semi-automatic evaluations, overcoming time-consuming operations. In this article we present a general framework for the acquisition and processing of hyperspectral images in order to obtain semi-automatically generated decay maps of historical buildings. Starting from the presentation of basic concepts of spectral imaging, we discuss capturing tools and methods for data collection campaigns. Likewise pre-processing operations and classification algorithms are illustrated, along with datasets, which are usually required for the execution of such elaborations. This work aims to become a reference for those are intended to improve research in the field of conservation of architectural heritage with the application of HSI, presenting a typical workflow to follow for surveys and analysis.