Crack length measurement in the surface of vessel structures based on computer vision


Measuring crack length is critical for data acquisition in fatigue tests. Especially for large, complex marine structural tests with multiple cracks, efficient and inexpensive crack length measurement techniques are a major advantage. A computer vision based method for measuring the crack length on the surface of complex vessel structures is developed with the goals of reasonable accuracy with extreme ease-of-use. The method involves digital image acquisition with automated crack detection and then dimension measurement. In order to capture a clear image by a GoPro camera with a 12.5X macro lens, UV dyes are applied to increase contrast between cracks and structures surfaces. The images are transferred to grey scale and the crack is detected by an edge filter. Noise points are filtered out and the crack is simulated and measured on pixel level. A MATLAB code is created for the proposed method to automate the processing of images. To evaluate the method, a laboratory experiment is conducted on a hexagon tension specimen with four cracks. Details of the experimental testing, figure capturing, figure processing and accuracy are presented.

Trends in the Analysis and Design of Marine Structures