Method and device for detecting a substrate

Abstract

The invention discloses a detection method and device of a substrate. The detection method includes collecting a to-be-detected image of the substrate to be detected; based on the to-be-detected image, acquiring the detecting value of at least one recognition parameter, relative to the optical property, of the substrate to be detected; correspondingly inputting the detecting value of the recognition parameter into a pre-trained input node of a neural network-based recognition module; acquiring and outputting a first recognition result, recognized by the recognition module according to the input data of the input node, of the substrate to be detected; based on a training image collected from the substrate to be detected with recognition result known, performing training to acquire the neural network. According to the arrangement, detection outcome bias caused by artificial subjective factors is reduced, while the detection accuracy is increased, workers are fewer, and cost is lowered.

Description

technical field [0001] The invention relates to the technical field of detection, in particular to a substrate detection method and detection device. Background technique [0002] Thin film transistor liquid crystal display TFT-LCD is used as the display element of products such as mobile phones, Pads, notebook computers, monitors and televisions, and has become indispensable in today's life. [0003] In order to ensure the performance of the display, various tests are carried out during the production and sales of the display, including the test of the display performance (such as flicker, mura, etc.) of the display panel. [0004] At present, human eye inspection is required in the process of substrate production in the TFT-LCD industry, that is, human eye inspection is performed on the substrate after the current process is completed. [0005] However, the existing human eye detection needs to consume a lot of manpower and material resources, as explained below. [0006...

AI Technical Summary

Problems solved by technology

[0005] However, the existing human eye detection requires a lot of manpower and material resources, as explained below

[0006] In a mature substrate production factory, the defect rate of macro defects is about 1 in 10,000. It takes about 3 minutes for the human eye to detect a substrate macroscopically, so in order to detect less than ten of the 10,000 substrates with macro defects A defective color filter substrate takes about 30,000 minutes, or more than 500 hours of inspection time

It will take more than 50 hours to arrange 10 inspectors to conduct the inspection

[0007] However, if the random inspection method is adopted, it is almost impossible to inspect a few substrates with defective defects from 10,000 substrates

Images