Backgroundremovalandw_省略_gydistributiono.pdfVIP

Background rem ova l and weld defect detection ba sed on energy d istr ibution of image 3 Chi D azhao, Gang Tie and Gao Shuangsheng 迟大钊 , 刚 　铁 , 高双胜 33 A bstrac t　The la tera l w ave in ultrasonic TO FD ( tim e of f light d iffraction) im age has a tail in transit tim e, w hich d isturbs the detection and eva luation of sha llow w eld defect. M eanw hile, the la tera l w ave and back2w all echo that act as background add redundant data in d ig ita l im age processing. In order to separa te defect w ave from la tera l w ave and prepare the w ay for follow ing im age processing, an a lgorithm of background rem oval m ethod nam ed as m ean2subtraction is developed. B ased on this, an im proved m ethod by sta tistic of the energy distribu tion in the im age is proposed. The results show that by choosing proper threshold value according to the axial energy distribu tion of the im age, the background can be rem oved au tom atica lly and the defect section becom es predom inan t. M eanw hile, d iffractive w ave of sha llow w eld defect can be separated from la tera l w ave effectively. Key words　time of flight diffraction ( TOFD) , digital image p rocessing, background removal, defect detection 0　 In troduction U ltrasonic TOFD is a newly developed technique that has p roved highly effective for the inspection of thick p late weld and has started to take its way to rep lace the other testing techniques [ 1 - 3 ] . U ltrasonic TOFD has a lot of ad2 vantages which make it the p referable technique in weld testing [ 4 - 5 ] . It gives accurate sizing, positioning and char2 acterising of weld defects with a high p robability of detec2 tion. Currently, most of the TOFD image data evaluation is done in raw image, requiring operator skill, experience and most significantly time. In light of the industrial p res2 sure, the recent trend is to automate defect detection p rocess. The p rocedure could be imp roved by advanced digital image p rocessing by utilising computational tools.