Defect Segmentation of Ultrasonic Aluminum Bonding Joint Based on Region Growing and Level-Set

An optimized cascaded defect segmentation algorithm which combining with region growing and level set is proposed to segment effectively the joint defect in ultrasonic aluminum wire. Three algorithms, which include Log edge detection, neighborhood variance, and Gray-Level Co- occurrence Matrix (GLCM), to achieve the location of joint region. The redundancy region of Bonding joint is removed clearly, even there are much noise in the joint image. The modelling of region growing and level set are descried in detail, respectively. More, the defect segmentation implement is conducted in OpenCV, and the result proves that the cascaded defect segmentation combining with region growing and level set can effectively segment the defects of Bonding joints. © 2018 IEEE.