Method for quality inspection during the execution of facades based on uas images and machine learning algorithms

This research highlights the importance of inspecting building facades for damage during construction. Manual inspections are inefficient, costly and unsafe, requiring digital solutions. The study proposes an inspection method using images from Unmanned Aerial Systems (UAS) and Machine Learning (ML) algorithms, utilizing ResNet and AlexNet networks on a web platform to detect defects in concrete facades during execution, supporting the Quality Management System (QMS). Design Science Research was the chosen method. It involved three empirical studies that integrated UAS-collected images, analyzed them using ML algorithms, and implemented the findings into the QMS to expedite information generation during facade execution. The method identified four construction anomalies, with an ML model correctly recognizing anomalies with up to 98.80% precision in detection and classification during testing. The generated information was integrated into the QMS via images, reports, action plans and meetings. This research contributes to the automated inspection of facades, offering optimized data acquisition using UAS, facilitating the identification of defects and supporting decision-making through comprehensive QMS integration.