Author(s): Nimesha Ranasinghe, Sahan Perera, Tharushi Wijesekara and Dinuka Fernando

Abstract: The study presents a comprehensive data-driven framework for the early detection of building façade defects through the integration of Unmanned Aerial Vehicle (UAV) photogrammetry and deep learning-based image analytics. Traditional façade inspection methods are often limited by high labor costs, safety risks, and restricted accessibility, particularly in dense urban settings. To overcome these limitations, high-resolution UAV imagery was acquired and processed through a photogrammetric reconstruction pipeline to generate three-dimensional façade models. Deep convolutional neural networks (CNNs) were trained and optimized to detect and classify common defects, including cracks, efflorescence, corrosion, and spalling, using radiometrically corrected and geometrically aligned datasets. The results demonstrated substantial improvements in precision, recall, and mean average precision (mAP), with the proposed model achieving up to 93% accuracy compared to 82% in conventional 2D approaches. Localization accuracy, measured through root mean square error (RMSE), improved by nearly 50% owing to 3D back-projection of detected anomalies onto the reconstructed mesh. Moreover, the framework achieved reduced inference time and increased throughput, highlighting its scalability for real-world deployment. The integration of UAV-based photogrammetry with artificial intelligence not only enhances the reliability of façade inspection but also provides spatial and temporal insights essential for predictive maintenance. Practical recommendations include the institutional adoption of UAV inspection workflows, development of standard operating procedures for data acquisition and annotation, periodic model retraining using domain-specific datasets, and linking the framework with building information modeling (BIM) for real-time condition monitoring. Overall, the research establishes a foundation for intelligent, cost-effective, and proactive façade management systems capable of transforming urban infrastructure maintenance into a predictive, data-driven process.

Pages: 01-05 | Views: 2 | Downloads: 1

Download Full Article: Click Here