Time Structural Defect Detection Using UAVs and GA-Chebyshev Neural Networks in Top-Down Construction

This paper presents a novel approach to construction defect detection by integrating drone technology with a GA-Chebyshev neural network. Drones are employed to collect real-time imagery of construction sites, which is analyzed by the neural network to identify defects such as cracks, spalling, and exposed rebar. The GA-Chebyshev network, optimized using a genetic algorithm, outperforms traditional methods like CNN and SVM in terms of classification accuracy and mean absolute error (MAE). In experiments on publicly available datasets, the proposed method achieved an accuracy of up to 90.54% for spalling detection, with a significant reduction in MAE. These results demonstrate the effectiveness of the proposed system, although further validation in real-world environments is necessary. Future work will focus on real-time implementation and deployment on active construction sites.