关键词: 机器视觉, 荧光磁粉, 裂纹分割, 图像差分

Abstract: Fluorescent magnetic particle flaw detection is widely used in non-destructive surface testing of ferromagnetic parts,and in the process of realizing automated machine inspection,the reflective area and background noise often affect the recognition accuracy of cracks. A joint imaging method of fluorescent magnetic powder under multiple light sources is proposed, by comparing the different characteristics of fluorescent magnetic powder imaging under the irradiation of different light sources (red light and ultraviolet light),and by performing the difference and threshold mask processing on the two types of images,the effects of reflections and background noise are effectively reduced,and the crack characteristics are highlighted to improve the accuracy of detection and reduce the rate of false alarms. By building an experimental platform for image acquisition and defect segmentation,the results show that the proposed method is able to accurately segment cracks under the conditions of background noise interference and reflection masking.

Key words: machine vision, fluorescent magnetic particle, crack segmentation, image difference