Image processing and wavelets transform for sizing of weld defects using ultrasonic TOFD images

Abstract

Ultrasonic time-of-flight diffraction (TOFD) is rapidly gaining prominence as a reliable nondestructive testing technique for weld inspection in steel structures, providing highly accurate positioning and sizing of flaws. A number of signal and image processing tools have been specifically developed for use with TOFD data and adapted to function autonomously without the need for continuous intervention, configuring automatically according to the nature of the data and the data acquisition settings. This paper presents the results of several innovative procedures based on Multiresolution analysis such as wavelet transforms and texture analysis for detecting edges of planar defects. The approach is based on the decomposition by packets of wavelets of the image while taking into account of the underimage content textural after each level of decomposition. The reconstruction of the image is done by eliminating the underimages of poor textural information and at the end the segmented image is got by a binarisation operation done on the image rebuilt histogram. The automation of sizing and positioning of weld flaws in TOFD data as an essential stage of a comprehensive TOFD inspection and interpretation aid is developed and implemented. Key words: Ultrasonic Time-Of-Flight Diffraction., Texture Analysis, Wavelet Transform.