The condition assessment of glass cladding of highrise buildings is of paramount importance
for structural safety and long-term maintainability. Traditional inspection methods
are manual, time-consuming and costly procedures that carry significant risks for
the inspection professionals. This study presents an integrated drone-based system
that automates the crack and defect detection of glass cladding of skyscrapers using
artificial intelligence and multi-sensor data collection. The system uses a combination
of optical, LiDAR, thermographic and hyperspectral sensors, and the data is analyzed
by deep learning models (YOLOv8, ResNet50, PointNet++ + CNN). Algorithmic performance
measurements on open datasets (CrackTree200, DeepCrack) demonstrated that the multi-sensor
fusion model achieved an F1 score of 93.7%, while YOLOv8 achieved an outstanding real-time
frame rate of 28 frames/second. The results highlight the advantages of multimodal
data processing, especially in improving the accuracy of fault detection. Future development
directions include online learning, the introduction of resourceefficient network
architectures, automatic route adaptation, and the integration of 5G/6G communication
technologies. The proposed system can contribute to safer, more cost-effective, and
data-driven maintenance processes for high-rise buildings.
