Strip steel surface defect detection method and system

Abstract

The invention relates to a detection method and system for strip surface defects, comprising the following steps: obtaining strip surface defect samples; constructing an improved ACGAN model based on residual optimization; training the discriminator network and the generator network of the improved ACGAN model to obtain the discriminant The discriminant loss and classification loss of the generator and the generator; the gradient penalty mechanism is introduced in the discriminator, the parameters of the discriminator network model are optimized, and the optimized discriminator network model is obtained; the generator discriminant loss and classification loss are combined to update the generator Network model parameters to obtain the optimized generator network model; iterate to make the generator and discriminator reach Nash equilibrium, and improve the ACGAN model to converge to the best; input the test set data into the optimized improved ACGAN model to realize the strip surface Defect detection. It improves the training stability of the ACGAN model, and greatly improves the accuracy of strip surface defect detection in a small sample environment.

Description

technical field [0001] The invention relates to the technical field of strip steel surface defect detection, in particular to a strip steel surface defect detection method and system. Background technique [0002] Steel is one of the essential raw materials in industry, and is widely used in automobile manufacturing, aerospace and electric energy and other fields. However, due to complex factors such as manufacturing process and production environment, various defects are prone to appear on the surface. These defects not only affect the appearance of the product, but also adversely affect its performance and safety. Therefore, it is particularly important to detect steel surface defects to control its quality. [0003] The automatic detection method based on machine vision has the advantages of real-time, high efficiency, economy and non-contact. The accuracy depends to a certain extent on the features designed by human experts, and is very sensitive to changes in the app...

AI Technical Summary

Problems solved by technology

However, the application of deep learning depends on the drive of big data. Most of the classifiers trained based on traditional deep learning are established when there are sufficient training samples. However, it is difficult to obtain surface defect data in actual industrial production. Collecting a large number of labeled samples is time-consuming and expensive. However, the accuracy of defect recognition in existing methods is low in a small-sample environment, and there are even problems such as misidentification.

Images