A Method for Enhancing Magneto-Optical Images of Cracks by Improving GAN

Abstract

The invention discloses a method for enhancing the magneto-optical image of a crack by improving GAN. Firstly, the magneto-optical image of the crack is collected and preprocessed, and then the preprocessed magneto-optical image is expanded to obtain training data for training the improved GAN model, and then The improved GAN model is trained, and finally, the magneto-optical image of the test piece is input into the improved GAN model, and the image blocks are output and stitched through the internal generation network, so as to obtain a high-contrast crack magneto-optical image, which realizes defect enhancement .

Description

technical field [0001] The invention belongs to the technical field of image processing, and more specifically relates to a method for enhancing a crack magneto-optical image by improving a generative confrontation network (GAN). Background technique [0002] Crack detection plays an extremely important role in preventing occasional dangerous events caused by crack-defective facilities. With the rapid development of modern industrial technology, industrial facilities and equipment are gradually developing towards high precision, which also has extremely high requirements on the safety of components. The appearance and expansion of cracks will not only seriously affect the performance of facilities and equipment, To a certain extent, it may even lead to the fracture of the components, causing safety problems of the facility, resulting in extremely serious consequences. Therefore, it is necessary to detect cracks in industrial components at an early stage. [0003] The curre...

AI Technical Summary

Problems solved by technology

[0007] 2) The image does not have texture features

[0009] 4) Noise has a great influence on the accuracy of crack detection

Due to the extremely high similarity between the pixel values ​​of the noise area and the crack area, the above method is difficult to filter out the noise effectively

When filtering in the frequency domain, a specific passband method is usually used to filter out the noise component, but in the magneto-optical image, there is no obvious difference in frequency distribution between the noise and the defect area in the frequency domain

In addition, in the field of image processing, methods such as edge detection and threshold segmentation are usually used to segment images, but the above methods are usually extremely dependent on artificially selected parameters, and do not have generalization capabilities, and cannot automatically detect large batches of images.

Images