CRACK-LIKE DEFECT DETECTION AND SIZING FROM IMAGE SEGMENTATION THROUGH COOCCURRENCE MATRIX ANALYSIS

The inspection of austenitic welds used in the nuclear industry using ultrasound poses problems in interpretation: strong grain noise makes the detection of the crack top and the crack bottom difficult. Since corresponding echoes enable defect sizing, defect sizing also becomes difficult. The formation of two-dimensional images (BSCAN) and their processing enable an increase in the effectiveness of testing. This paper presents a new segmentation method, based on the analysis of the co-occurrence matrix. A threshold is automatically calculated by the extremum of a curve, called the Average Grey-level Variance Measure. This curve is an analysis of the distribution of the matrix coefficients. Examples of segmentation improvement applied to BSCAN images of artificial defects are presented.