Metal surface defect detection method based on U-NET convolutional neural network

Abstract

The invention relates to a metal surface quality detection method, in particular to a U-NET-based metal material surface defect detection method, which comprises the following steps: acquiring metal material image data; performing data enhancement on defect type images with a small number by using an ACGAN image generation network, preprocessing the acquired image data, and dividing a data set into a test data set and a training and verification data set; establishing a U-NET semantic segmentation network, performing down-sampling feature extraction on the images by using a deep convolutional neural network, and fusing multi-scale feature maps by using multilayer deconvolution; and inputting the fused feature maps into a classifier module to carry out metal material surface defect positioning and classification. The model can quickly judge whether a surface image of a metal material has defects or not and give the defect category and the position of the defect, so that the automatic analysis of the quality in the surface image of the metal material is completed.

Description

technical field [0001] The invention relates to a metal surface quality detection method, in particular to a U-NET-based metal material surface defect detection method, which belongs to the field of image processing and analysis. Background technique [0002] Metal materials are important industrial products, and their surface quality directly affects their marketing and even engineering safety. With the continuous improvement of my country's total industrial production value, various production enterprises have put forward higher requirements for the surface quality of products. At the same time, real-time detection of product surface quality problems during the production process can not only warn workers to repair production equipment in time, but also reduce waste and increase the authenticity rate of products. Therefore, real-time detection of material surface quality, early detection of damage and timely maintenance of production equipment have become a basic task in ...

AI Technical Summary

Problems solved by technology

The visual inspection method is labor-intensive, dangerous, labor-intensive, low-efficiency, and the measurement results are subject to subjective influence; the ultrasonic flaw detection method has high operating costs, and at the same time has low classification accuracy and low detection speed; The excitation signal makes the system structure and signal processing more complex, and the detection efficiency is relatively low

Images