Crack Identification Method of Bearing Ring Fluorescence Magnetic Particle Flaw Detection Image Based on Interference Display Rejection

Fluorescent magnetic particle inspection is a commonly used non-destructive testing method for bearing rings. With the development of computer vision technology, more and more automatic discrimination algorithms have been applied to fluorescent magnetic particle inspection. However, due to the complex background of the detection image of fluorescent magnetic particle flaw detection, there are many interferences, which reduces the accuracy and reliability of automatic discrimination results. Therefore, starting from the analysis of the characteristics of the interference display, this paper studies the interference removal method of the fluorescent magnetic particle flaw detection image to extract the candidate frame of the crack detection, and on this basis, the ResNet50 network is used to identify the surface crack image of the bearing ring. The results show that compared with the image recognition results of fluorescent magnetic particle flaw detection without interference rejection method, the image recognition method of fluorescent magnetic particle flaw detection based on interference rejection proposed in this paper can increase the accuracy of crack recognition from 79.36% to 94.33%, and has certain application reference value for intelligent discrimination of bearing outer ring fluorescent magnetic powder flaw detection images.