Local optimal threshold technique for the segmentation of ultrasonic time-of-flight diffraction image

摘要

An ultrasonic time-of-flight diffraction (TOFD) image is composed of a series of digitised A-scan signals collected by scanning the interior of the material. However, its low contrast and grey-level discontinuities make it difficult to fully detect the crack edges in an ultrasonic TOFD image using global optimal threshold segmentation. This paper presents a local optimal threshold technique to extract the crack edges in an ultrasonic TOFD image completely. This method begins with the estimation of the distributed energy in a TOFD image and then divides the image into multiple blocked images based on the estimated energy distribution. An optimal threshold for each blocked image is then calculated through Otsu 's method and the multiple local optimal thresholds are obtained from the blocked image. Finally, the crack edges are detected through these local optimal thresholds in the ultrasonic TOFD image. Experimental results show that the edges of defects can be extracted and cracks can be identified in the ultrasonic TOFD image.