Industrial product appearance defect detection method and device

Abstract

The invention relates to the technical field of industrial quality inspection. In order to solve the technical problem of poor detection effect of industrial product appearance defects, a method and device for detecting industrial product appearance defects are provided. The method includes: labeling defect categories on a plurality of defect samples , to obtain the first sample set with N labeled categories; input the first sample set into the feature extraction network to obtain the feature vector corresponding to each defect sample, and form a clustering data set; perform a density maximum on the clustering data set Clustering to classify multiple defect samples into K cluster categories to obtain a second sample set with K cluster categories; train the detection network through the second sample set to obtain a defect detection model; The product image is input into the defect detection model to obtain the output result; the sub-images whose output result is a multi-class output defect are fine-grained classification, and the output result is an output result in which the multi-class output defect belongs to the corresponding label category.

Description

technical field [0001] The invention relates to the technical field of industrial quality inspection, in particular to a method for detecting industrial product appearance defects, a device for detecting industrial product appearance defects, a computer device and a non-temporary computer-readable storage medium. Background technique [0002] In the field of industrial intelligent quality inspection, deep learning has been widely used and received a lot of attention. Especially for detection tasks such as industrial appearance defects, due to the wide variety of defects and the various forms, using deep learning can often achieve better results. [0003] The general appearance defect detection process based on deep learning algorithm mainly has the following steps: the first step is to manually label the collected data according to the specified defect category; the second step is to select a target detection or instance segmentation network , to set a series of parameters,...

AI Technical Summary

Problems solved by technology

According to the above-mentioned general detection method, if the classification is carried out on the pre-specified categories, the network will face the problem of small differences between classes and large differences within classes. The training is difficult, and the performance of the trained model is often not enough in actual detection. Well, the false positive rate is high

Images