Title :
Cracklike defects sizing from co-occurrence matrix
Author :
Moysan, J. ; Corneloup, G. ; Magnin, I. ; Benoist, P. ; Basset, O.
Author_Institution :
Lab. de Controle par UltraSons, CENS, Gif sur Yvette, France
Abstract :
The inspection of austenitic welds used in the nuclear field with ultrasound involves problems of interpretation: intense grain noise makes the detection of the crack difficult. The formation B-scan images and their digital processing increase the effectiveness of testing. A segmentation method based on the cooccurrence matrix is presented. It separates defect zones and noise zones. The calculation of statistical parameters makes it possible to automate and to improve image thresholding. Defect sizing through the analysis of the modified image is the final goal. Some examples of segmentation improvement applied to artificial defects are presented
Keywords :
austenitic steel; crack detection; ultrasonic materials testing; welding; B-scan images; artificial defects; austenitic welds; cooccurrence matrix; cracklike defects sizing; defect zones; digital processing; grain noise; image thresholding; modified image; noise zones; nuclear field; segmentation method; statistical parameters; ultrasound; Histograms; Image analysis; Image segmentation; Inspection; Laboratories; Sorting; Symmetric matrices; Testing; Ultrasonic imaging; Welding;
Conference_Titel :
Ultrasonics Symposium, 1990. Proceedings., IEEE 1990
Conference_Location :
Honolulu, HI
DOI :
10.1109/ULTSYM.1990.171542
