Method for constructing convolutional neural network model for surface defect detection and application thereof

Abstract

The invention discloses a method for constructing a convolutional neural network model for surface defect detection and application thereof. The method comprises the following steps: (1) collecting and importing an original picture; (2) preprocessing the original picture, and determining the preprocessed original picture as an original training sample and storing the original training sample; (3)labeling the original training sample to generate a labeled sample; (4) transforming the original training sample to generate a new training sample, and enhancing the training sample; and (5) taking the training sample as input data, marking the sample, carrying out corresponding processing, and taking the processed sample as reference output: after the convolutional neural network model is generated, storing the convolutional neural network model until the stable convergence accuracy is achieved through multiple iterations. According to the method, various adverse effects caused by interference factors such as random product defect form, complex texture and low contrast ratio under the condition that a small number of samples are input are overcome, so that the defect recognition rate isimproved.

Description

technical field [0001] The technical field of image detection of the present invention particularly relates to a method for constructing a convolutional neural network model for surface defect detection and its utilization. Background technique [0002] Usually, the surface of industrial products will have defects such as cracks, dirt, impurities, and lack of appearance due to various factors in the production process. Traditional production lines use manual online judgment, but manual work is also limited by fatigue and human eyes. etc. lead to a decline in product yield, affect the quality of shipments, and fail to meet the requirements of end users. [0003] In recent years, some system integrators have adopted machine vision methods to detect surface defects of industrial products. First, image the surface defects of the magnetic circuit through a special optical path to obtain defect information on the product surface, and then use morphological processing, geometric an...

AI Technical Summary

Problems solved by technology

Due to the complex surface texture of industrial products, the shape of surface defects is very random, the contrast is low, and the stability on the production line is also difficult to guarantee. The existing machine vision algorithms have great limitations in the process of analyzing defects. The rate is very high, it is difficult to meet the actual production requirements

Moreover, the existing machine vision algorithms need to constantly adjust and optimize the algorithm when facing various random defects, which has very poor adaptability and will also lead to a waste of time.

Images