Classification method of surface defect of aluminum plate based on bp neural network and support vector machine

Abstract

The invention discloses an aluminum plate surface defect classification method based on a BP neural network and a support vector machine, extracting the characteristic value of the aluminum plate surface defect as the input of the BP neural network classification model, and constructing the BP neural network as the oil spots and the first type of defects as the output Classification model; the first type of defect uses a one-to-one classification method to construct multiple support vector machine classification models; obtain learning samples and train the BP neural network classification model and support vector machine classification model; use the BP neural network classification model for oil spot For the classification of the first type of defects, the BP neural network classification model is considered to be the test sample of oil spots and eliminated, and the support vector machine classification model is used to classify the remaining first type of defects; finally, the classification results are obtained through statistics. This method not only improves the identification and classification rate of oil spots on the surface of cold-rolled aluminum sheets, but also improves the overall recognition rate of surface defects on cold-rolled aluminum sheets, and can also be used for identification and classification of other metal surface defects, and is simple and easy to implement.

Description

technical field [0001] The invention relates to the technical field of pattern recognition, in particular to a method for classifying surface defects of an aluminum plate based on a BP neural network and a support vector machine. It is suitable for metal surface defect detection system to identify and classify various types of metal surface defects including oil spots. Background technique [0002] The surface defect detection and classification of cold-rolled aluminum sheets has experienced three development stages: manual visual classification, traditional non-destructive testing classification and machine vision-based detection and classification. Poor sex. Traditional nondestructive testing classification methods include eddy current testing classification, infrared testing classification, magnetic flux leakage testing classification, laser testing classification, etc. These methods detect few types of defects, and the real-time detection and classification are not stro...

AI Technical Summary

Problems solved by technology

This scheme can improve the recognition and classification rate of individual defects on the surface of cold-rolled aluminum sheets and the overall recognition and classification rate to a certain extent, but it increases the complexity of the hardware and increases the cost, which is not conducive to the small and medium-sized cold-rolled aluminum sheet surface defect detection system. Aluminum plate production enterprises to promote the use

Images