Early detection of facing-masonry surface biodeterioration through convolutional neural networks

Preventive conservation is the proposed and recommended approach to preserve historic building heritage from deterioration problems caused by several types of actions. It is based on data collection, steady monitoring, inspections, and control of environmental agents. Architectural heritage is subjected to many deterioration issues caused by different types of pathologies, among which attention must certainly be paid to the growth of living microorganisms (bio-colonization). Monitoring actions able to represent the evolution of buildings’ deterioration state have been proposed and implemented towards the creation of predictive models based on machine learning methods with the aim to reduce the need for major interventions. In this paper is proposed a method for the early detection of microalgae growth on facing-masonry surfaces. Images representing the microalgae growth process on facing-masonry facades, collected during experimental activities in controlled conditions, were used for training and testing a convolutional neural network. The trained model can ensure an accuracy of 83% and is able to recognize the starts of the bio-colonization process on different types of clay bricks. The work shows that, by processing these images with the trained convolutional neural network, it is possible to disclose the first stage of bio-deterioration phenomena. This work is part of a more extensive research for the early detection of different types of building facade damages and could be implemented for real cases monitoring.