LAUSR

This paper presents a novel framework for damage identification in industrial building grid structures by integrating perceptual fuzzy systems with unmanned aerial vehicle (UAV)-assisted visual inspection. The system employs drones equipped with high-resolution sensors to capture structural surface data, simulating human visual cognition to detect and evaluate defects such as cracks, corrosion, and deformation in grid components. At the core of the methodology is a multi-layered perceptual fuzzy model that emulates human attention mechanisms, enabling adaptive focus on critical regions under complex environmental conditions. The BING (Binarized Normed Gradients) objectness measure is utilized for efficient initial detection of potential damage areas across varying scales. To enhance feature representation, time-sensitive and manifold-guided aggregation techniques are incorporated, ensuring robust handling of noise and variability in real-world inspection scenarios. A low-rank active detection (LAD) algorithm is introduced to prioritize informative regions, minimizing computational cost while maximizing detection accuracy. The system further employs an attention-guided inspection path (AIP) to streamline data acquisition and processing, integrating high-dimensional convolutional neural network (CNN) features for detailed defect characterization. A fuzzy inference system, combined with support vector machine (SVM) classification, categorizes defects based on severity, location, and structural impact. Validation experiments conducted on a dedicated grid structure defect dataset and six public structural health monitoring benchmarks demonstrate the systemąŕs superiority in identifying subtle material defects and capturing interactions between environmental factors and structural integrity. The proposed approach significantly improves the automation, accuracy, and intelligence of damage assessment for industrial grid structures, contributing to safer and more efficient infrastructure maintenance practices.