FAUNO: A Machine Learning-Based Methodology for Monitoring and Predictive Maintenance of Structures in Archaeological Parks Through Image Analysis

Archaeological parks represent one of the most important cultural resources for a nation. Archaeological parks safeguard, conserve, manage, and defend their archaeological, architectural, environmental, and landscape heritage. An Archaeological Park experiences very complex operational dynamics: for example, visitors often come into direct contact with artistic artifacts and, even unintentionally, can damage them. In addition, Archaeological Parks are usually outdoors; therefore, their cultural assets are subjected to environmental influences such as rain, wind, sun, and weeds. New technologies can effectively support monitoring activities. This paper will present an approach that aims to identify and classify potential critical issues within an Archaeological Park in an automated manner and to communicate these issues to the various actors in the reference scenarios. The reference scenario for this activity is the Archaeological Park of Pompeii, which cyclically produces high-definition images of its artistic resources using aerial drones. In particular, on a fortnightly basis, high-definition orthophotos of the park and its artistic artifacts are produced. It is understood that the entire operation is time-consuming and does not allow for Real-Time maintenance. Therefore, the research's objective is to automate this process by introducing Machine Learning techniques for automatically identifying and classifying problems in the image and implementing a mechanism for notifying park personnel. The methodology is based on using deep learning systems capable of identifying and classifying potential issues. The YoloV7 library was used to detect the issues in the images.