Towards automatic analysis of ultrasonic time-of-flight diffraction data using genetic-based Inverse Hough Transform

Ultrasonic Time-of-Flight Diffraction (TOFD) is a nondestructive inspection technique that has proved to be very effective for the detection, localisation and sizing of buried crack defects in steel structures. However, it produces a huge amount of data that are manually processed and interpreted. This process is time consuming and painstaking. Moreover, it requires the skill, alertness and experience of the operator. Consequently, it is subject to human errors. In order to save time, effort and inspection cost while at the same time increasing the detection rate, automatic analysis tools need to be developed. This paper presents thus an application of image processing techniques to the B-scan image representation of ultrasonic TOFD data so as to take advantage of the power of image representation of information. In a B-scan image, crack defects are characterised by multiple arcs of diffraction. In order to detect these multiple arcs of diffraction and thus reveal the presence of a crack in the structure under inspection, some methods based on conventional Hough Transform (HT) were proposed in the literature. The main problems related to conventional HT are its large data storage requirements and expensive computation times. To cope with these problems, we propose the use of the Inverse Hough Transform (IHT) where the voting process is performed in the image space rather than the parameter space. With the IHT, the local peak detection problem in the parameter space is converted to a parameter optimisation problem that is solved using Genetic Algorithms. The proposed Genetic-Based Inverse Voting Hough Transform GBIVHT algorithm allows thus the automatic detection of the arcs of diffraction, and therefore the crack defects, while avoiding the computational complexity as well as the huge storage requirement of conventional HT.