Mura defect detection method and device

Abstract

The invention discloses a Mura defect detection method and a Mura defect detection device. The method comprises the following steps: executing multiple different image enhancement processing on a training sample image to obtain multiple training sample enhanced images corresponding to the training sample image; inputting the training sample enhanced image into a convolutional neural network modelfor training, and outputting a classification model; performing multiple image enhancement processing the same as the training sample image on the to-be-detected test sample image to obtain multiple test sample enhanced images; and inputting the enhanced image of the test sample into the classification model, and outputting a defect detection result. According to the method and device, multiple Mura defects can be effectively detected at the same time.

Description

technical field [0001] The invention belongs to the technical field of panel defect detection, and more specifically relates to a method and device for detecting Mura defects. Background technique [0002] In automatic optical inspection (AOI) defect detection, mura defect detection is very important, which directly affects the quality of the final display panel and the grade output result of the finished product. There are two traditional mura detection methods: one is to detect mura defects through human eye observation. However, the process of human eye detection is highly subjective. At the same time, with the extension of working hours, people will also experience fatigue, which will lead to the detection efficiency. The other is to design a detection algorithm for specific mura defects to detect mura defects. In traditional algorithm detection, different detection algorithms need to be designed for each type of mura. Irregular and relatively weak, for example, many Mu...

AI Technical Summary

Problems solved by technology

There are two traditional mura detection methods: one is to detect mura defects through human eye observation. However, the process of human eye detection is highly subjective. At the same time, with the extension of working hours, people will also experience fatigue, which will lead to the detection efficiency. The other is to design a detection algorithm for specific mura defects to detect mura defects. In traditional algorithm detection, different detection algorithms need to be designed for each type of mura. Irregular and relatively weak, for example, many Mura have only 1 to 2 gray levels in the image, and traditional algorithms are difficult to achieve effective detection

[0003] In view of the problems existing in the traditional mura defect detection method, a method of using deep learning algorithm for mura defect detection is proposed in the prior art, and a classification model is obtained by performing deep learning on training samples to detect mura defects. However, due to the wide variety of mura, The shape is diverse and irregular, and relatively weak. The trained classification model cannot effectively detect multiple mura defects at the same time, and the detection effect is unstable and the versatility is not strong.

Images