Novel computing model can detect structural damage with no physical intervention

Researchers from Drexel University in Pennsylvania have developed a system that is believed to efficiently identify and evaluate problematic areas by integrating a novel machine learning approach with visual inspection technologies.

The team argues that the rapid decline of the built environment makes the maintenance by physical, and manual inspect impossible. The latter observation is based on recent incidents, e.g. building collapses and structural failures in roads and bridges.

Their innovative system merges computer vision with deep learning algorithms to find potential positions with a structural problem. Once an area of interest is identified, a robotic arm scans it with a laser scanner, creating a three-dimensional image of the damaged area. At the same time, a Lidar camera scans the surrounding structure. Merging these data produces a digital model presenting crack dimensions and facilitating tracking of changes over time.

More specifically, the system employs a high-resolution stereo-depth camera feed processed by a convolutional neural network. This network, trained on crack samples, identifies crack-like patterns in images collected by a robotic system, making the detection process more efficient.

While human inspectors retain the final decision-making authority, the robotic assistants can substantially reduce their work time and minimize oversights and subjective errors. By focusing data collection on areas requiring attention, the system ensures comprehensive and reliable assessments.

With an impressive excess of repairs estimated at $786 billion for US roads and bridges, and a shortage of skilled infrastructure workers, the need for efficient inspection and maintenance solutions is urgent.

The researchers envision integrating this multi-scale monitoring system into a broader autonomous framework, incorporating drones and other autonomous vehicles. Real-world testing and collaboration with industry and regulatory bodies are crucial for practical application and ongoing improvement of the technology.