A Weld Defect Recognition Method Based on Improved Convolutional Neural Network

Abstract

The invention discloses a weld defect recognition method based on an improved convolutional neural network, establishes a pooling model that comprehensively considers the pooling domain and its feature map feature distribution, and introduces a correction factor μ to correct the maximum pooling feature and the ReliefF algorithm Combined with neural network as a feature selection method; construct a deep convolutional neural network with the above-mentioned pooling model and feature selection method, iterate with the goal of minimizing the cost function, and form a neural network for weld defect recognition through training to realize weld seam Identification of defect types. The method of the invention avoids the process of manually extracting features in the traditional welding seam defect recognition method, and further improves the defect recognition rate of the traditional improved convolutional neural network model.

Description

technical field [0001] The invention belongs to the technical field of automatic identification of welding seam defects, and in particular relates to an identification method of welding seam defects based on an improved convolutional neural network. Background technique [0002] In the field of automatic identification of weld defects, the traditional method inevitably has to go through the process of manual selection and feature extraction, which is time-consuming and laborious, and whether the selection of features is reasonable or not is highly subjective, which has a great impact on the accuracy of recognition . [0003] The classic pooling models of convolutional neural networks (maximum pooling model and average pooling model) lack dynamic adaptability when extracting features from pooling domains with different feature distributions, resulting in inaccurate feature extraction. [0004] Weld defect data belongs to small samples and non-massive data. In the case of non...

AI Technical Summary

Problems solved by technology

[0005] The technical problem to be solved by the present invention is to provide a method for identifying weld defects based on the improved convolutional neural network, avoiding the manual feature selection process of the traditional method; for the classic pooling model of the neural network By making improvements with the feature selection method, a weld defect recognition method based on the improved convolutional neural network is obtained, and the image of the weld to be recognized is sent to the trained deep neural network to realize the automatic recognition of the weld defect type

Images