Method for automatically labeling appearance defect images of industrial products based on semi-supervised learning

Abstract

The invention discloses a method for automatically labeling an industrial product appearance defect image based on semi-supervised learning. The method comprises the following steps of (a) collectinga product appearance defect image to construct a data set; (b) separating a known label data image and an unknown label data image in the data set, and constructing a training set by using the known label data image; (c) constructing a deep convolutional neural network classification model and training and learning deep features of defects on a known label training set; (d) marking unknown label data by taking the trained deep convolutional network classification model as an automatic marking model, and marking pseudo labels on the unknown label data; (e) extracting a predetermined amount of data from the pseudo tag data and merging the data into a training set formed by known tag data to form a new training set; (f) continuously training the automatic labeling model by using the new training set, re-labeling the pseudo label data merged into the training set by using the trained model, and converting the pseudo label data into new known label data; and (g) repeating the step (e) and the step (f) till all the pseudo tag data is marked as known tag data. The method is advantaged in that texture features, wavelet features and the like are more abstract, and robustness is better; andunknown label data can be automatically labeled, and the method is high in efficiency, accuracy and capability of greatly reducing manual labeling cost.

Description

technical field [0001] The invention belongs to the technical field of machine vision detection, and more specifically relates to a method for automatically marking industrial product appearance defect images based on semi-supervised learning. Background technique [0002] With the continuous upgrading of market consumption, the demand for related products in 3C, home appliances and other industries has increased sharply. At the same time, the requirements for production quality have been continuously improved, especially the requirements for appearance quality are particularly prominent. The appearance defect detection of large-scale and high-precision industrial products has become the current One of the significant problems within the industry. At present, defect image detection based on machine vision has gradually replaced manual inspection as the mainstream appearance detection method for industrial product parts, but machine vision detection methods need to use a larg...

AI Technical Summary

Problems solved by technology

However, the appearance defect images of industrial products have the characteristics of inconspicuous defects, large optical noise, and image imbalance of defect categories. Therefore, the existing automatic image labeling technology cannot guarantee the accuracy of unknown data labeling, and cannot effectively replace the manual image labeling process.

Images