A novel computer vision-based approach for fatigue crack detection using a video stream was developed. The technology offers significant advantages over traditional inspection and fixed sensor-based methods and overcomes many of the limitations and concerns with static image and image processing based techniques.
The system tracks the surface motion of a structure under repetitive loads using a video stream and detects irregularities resulting from opening/closing of cracks. It also enables the measurement of the crack opening with submillimeter accuracy, which can facilitate fatigue damage assessment and fatigue life calculations.
Detecting structural impairments of in-service physical structures such as bridges and roadways is necessary to ensure their continued functionality and to allow for timely repairs. Early knowledge of such structural impairments also helps prevent catastrophic damage due to factors such as corrosion, fatigue, and unexpected loading conditions such as natural disasters or terrorism.
How it works:
The method involves taking a short video stream of a structure in a crack-prone area under repetitive loading. Then, feature points are detected within a region of interest (ROI) in each frame, the movements of which can be tracked throughout the video stream. By comparing movements between adjacent feature points, the presence of a crack can be detected and evaluated to determine the full extent of the crack along with its dimensions. This data can be stored and compared with previous scans to monitor crack propagation over time.
Video stream based techniques may be particularly useful in the structural health monitoring of today’s aging infrastructure, particularly when deployed on unmanned aerial vehicle (UAV) platforms for automated inspections.
Why it is better:
The method can detect cracks with high reliability and quantify the crack dimensions with high accuracy, even in poor lighting conditions and for structures with complex surface textures and features that typically result in false positives such as structural boundaries, wires, dirt, rust, or corrosion marks. The use of video allows for the robust collection of more comprehensive crack information than static images and also provides a level of redundancy to reduce uncertainty in the image processing.
The method can also be used in other applications, such as detecting localized slippage or the loosening of structural components.
Owner: University of Kansas
IP Protection Status: Pending Patent