Coating defect detection method based on deep learning

Abstract

The invention relates to a coating defect detection method based on deep learning, and the method specifically comprises the steps: 1, selecting different types of defect image blocks and defect-freeimage blocks as a training sample set; 2, constructing a convolutional neural network initial model; 3, carrying out data set training and model optimization; and 4, detecting and identifying the defects of the glued surface on line by using an optimized deep learning algorithm to realize automatic classification of the defects. According to the method, the defect image is directly input into thedeep convolutional network to construct a multi-layer neural network, layer-by-layer feature extraction is performed on the image, advanced features reflecting implicit in image data can be accuratelylearned under training of a large number of training data sets, and the network structure is optimized and trained to obtain an optimal parameter value; the problem of multi-defect type detection andrecognition in the coating melt-blowing process is solved.

Description

technical field [0001] The invention relates to the technical fields of machine vision and image processing, in particular to a coating defect detection system based on deep learning. Background technique [0002] In the production process of domestic coating technology, large companies all use imported equipment to monitor the change of coating thickness, which is expensive. Other small entrepreneurial companies mostly stay in the traditional manual visual inspection stage. The characteristics of traditional manual inspection do not require professional equipment and only need to guide and train professionals, but its disadvantage is that the results are greatly affected by subjective factors, which will lead to missed inspections and false inspections, and the defects can only be sampled offline after coating is completed. , unable to detect online in real time. [0003] Coating defect detection has always been a hot topic of research and discussion by scholars at home a...

AI Technical Summary

Problems solved by technology

The characteristics of traditional manual inspection do not require professional equipment and only need to guide and train professionals, but its disadvantage is that the results are greatly affected by subjective factors, which will lead to missed inspections and false inspections, and the defects can only be sampled offline after coating is completed. , unable to detect online in real time

Images