Virtual Booth

Concrete structures are susceptible to cracks due to environmental exposure, material fatigue, and structural loading. Conventional inspection methods rely heavily on manual visual inspection, which is time-consuming, labor-intensive, and potentially unsafe when inspecting high or inaccessible structures. This project presents an autonomous AI-powered wall-climbing robotic system designed for real-time concrete crack inspection and classification. The system integrates a wall-climbing robot equipped with a camera module and a Deep Convolutional Neural Network (DCNN)-based crack detection model to enable automated structural health monitoring. The robotic platform employs a hybrid adhesion mechanism combining an Electric Ducted Fan (EDF) and polyurethane rubber wheels to ensure stable movement on inclined and vertical surfaces. Image data captured by the onboard camera are processed using DCNN models to classify crack patterns. The proposed framework utilizes ResNet50 enhanced with Squeeze-and-Excitation (SE) blocks for feature extraction and classification, while MobileNetV2 transfer learning is incorporated to optimize computational efficiency. Experimental results demonstrate reliable classification performance and effective real-time crack detection. The proposed system improves inspection safety, reduces maintenance costs, and supports sustainable infrastructure monitoring. Furthermore, the technology offers strong commercialization potential for infrastructure inspection services, smart city monitoring systems, and automated structural maintenance platforms.