CN-122023875-A - Panel defect detection system, panel defect detection method and electronic equipment

CN122023875ACN 122023875 ACN122023875 ACN 122023875ACN-122023875-A

Abstract

The application discloses a panel defect detection system, a panel defect detection method and electronic equipment, and belongs to the field of defect detection. The system comprises a reference focusing module, an image acquisition module, a defect layering module and a defect extraction module, wherein the defect layering module comprises a deep learning classification model, the defect layering module is used for inputting a detection image set into the deep learning classification model to obtain a target structural layer output by the deep learning classification model, the target structural layer is a structural layer with highest defect definition in a plurality of structural layers, the defect extraction module comprises a deep learning segmentation model, and the defect extraction module is used for inputting a detection image corresponding to the target structural layer into the deep learning segmentation model to obtain a defect detection result of a panel to be detected output by the deep learning segmentation model. The system does not need a complex manual parameter adjusting process, greatly improves the detection speed and the flexibility of layered judgment, and remarkably improves the rechecking efficiency and the usability of the system.

Inventors

LU YINGBIN
GUO SHAOZHENG
SHI GUANGJUN
ZHAO YAN

Assignees

苏州凌云光工业智能技术有限公司
凌云光技术股份有限公司

Dates

Publication Date: 20260512
Application Date: 20251226

Claims (10)

1. A panel defect detection system, comprising: the reference focusing module is used for acquiring the reference coordinate position of a reference structural layer in the panel to be tested, wherein the panel to be tested comprises a plurality of structural layers which are sequentially laminated, and the reference structural layer is one of the plurality of structural layers; The image acquisition module is connected with the reference focusing module and is used for acquiring a detection image set of the panel to be detected based on the reference coordinate position, wherein the detection image set comprises a plurality of detection images corresponding to the structural layers; The defect layering module is connected with the image acquisition module and comprises a deep learning classification model, and is used for inputting the detection image set into the deep learning classification model to obtain a target structural layer output by the deep learning classification model, wherein the target structural layer is the structural layer with highest defect definition in the plurality of structural layers; The defect extraction module is connected with the defect layering module and comprises a deep learning segmentation model, and the defect extraction module is used for inputting a detection image corresponding to the target structural layer into the deep learning segmentation model to obtain a defect detection result of the panel to be detected, which is output by the deep learning segmentation model.
2. The panel defect detection system of claim 1, wherein the deep learning segmentation model comprises a first processing unit, a second processing unit, and a third processing unit connected in sequence; the first processing unit is used for extracting features of the detection image corresponding to the target structural layer and generating a feature thermodynamic diagram; the second processing unit is used for generating a defect segmentation mask map based on the characteristic thermodynamic diagram; the third processing unit is configured to generate the defect detection result based on the defect segmentation mask map.
3. The panel defect detection system of claim 1, wherein the defect detection result is a defect map, the defect map comprising a detection image and defect labeling information located on the detection image.
4. The panel defect detection system of claim 3 wherein the defect labeling information includes at least one of a defect bounding box, a defect type, a pixel size, and a defect size.
5. A panel defect detection system according to claim 3 wherein the defect map is a three-dimensional image.
6. The panel defect detection system of any of claims 1-5, wherein the image acquisition module is configured to acquire detection images of at least two first structural layers in the panel under test if a total number of structural layers of the panel under test is greater than or equal to a first preset threshold; The defect layering module is used for inputting the detection images of the at least two first structural layers into the deep learning classification model, and determining candidate structural layers from the at least two first structural layers; The image acquisition module is used for acquiring detection images of the structure layers adjacent to the candidate structure layer; The defect layering module is used for inputting the detection images of the candidate structural layers and the structural layers adjacent to the candidate structural layers into the deep learning classification model to obtain a target structural layer output by the deep learning classification model, wherein the target structural layer is one of the candidate structural layers and the structural layers adjacent to the candidate structural layers.
7. The panel defect detection system of claim 6, wherein the first preset threshold is greater than or equal to 6.
8. The panel defect detection system of any of claims 1-5, comprising an end-to-end trained composite model comprising the deep-learning classification model and the deep-learning segmentation model, an output of the deep-learning classification model being connected to an input of the deep-learning segmentation model.
9. A method for detecting a defect in a panel, comprising: acquiring a reference coordinate position of a reference structural layer in a panel to be tested, wherein the panel to be tested comprises a plurality of structural layers which are sequentially laminated, and the reference structural layer is one of the plurality of structural layers; acquiring a detection image set of the panel to be detected based on the reference coordinate position, wherein the detection image set comprises a plurality of detection images corresponding to the structural layers; Inputting the detection image set into a deep learning classification model to obtain a target structural layer output by the deep learning classification model, wherein the target structural layer is the structural layer with highest defect definition in the structural layers; And inputting the detection image corresponding to the target structural layer into a deep learning segmentation model to obtain a defect detection result of the panel to be detected, which is output by the deep learning segmentation model.
10. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor implements the panel defect detection method of claim 9 when the program is executed by the processor.

Description

Panel defect detection system, panel defect detection method and electronic equipment Technical Field The application belongs to the field of defect detection, and particularly relates to a panel defect detection system, a panel defect detection method and electronic equipment. Background With the rapid development of the liquid crystal panel towards ultra-high resolution, ultra-thin and large screen, the requirements of terminal products such as mobile phones, vehicle-mounted display screens, televisions and the like on the panel quality are becoming severe. The method provides extremely high standards for the detection precision and the rechecking efficiency of micro-point defects such as bright spots, dark spots, dust embedded spots, liquid crystal impurity spots and the like. At the front end of a production line, initial inspection is generally performed by an Automatic Optical Inspection (AOI) system, but the AOI can only judge whether point defects exist or not, cannot accurately position a specific physical layer where the defects are located, and also cannot measure the actual physical size of the defects, so that the follow-up process optimization lacks critical data support. In order to break through the limitation of AOI primary inspection, an electron microscope system with high-power microscopic imaging and accurate interlayer focusing capability is adopted for defect rechecking and analysis. However, the existing electron microscope system has the problems of long detection time, complex parameter adjustment and stiff layered logic in practical application, and severely limits the rechecking efficiency and the system usability. Disclosure of Invention The present invention aims to solve at least one of the technical problems existing in the prior art. Therefore, the invention provides the panel defect detection system, the panel defect detection method and the electronic equipment, which can improve the detection speed, do not need complex parameter adjustment flow, have flexible layering logic and obviously improve the rechecking efficiency and the usability of the system. In a first aspect, the present application provides a panel defect detection system comprising: the reference focusing module is used for acquiring the reference coordinate position of a reference structural layer in the panel to be tested, wherein the panel to be tested comprises a plurality of structural layers which are sequentially laminated, and the reference structural layer is one of the plurality of structural layers; The image acquisition module is connected with the reference focusing module and is used for acquiring a detection image set of the panel to be detected based on the reference coordinate position, wherein the detection image set comprises a plurality of detection images corresponding to the structural layers; The defect layering module is connected with the image acquisition module and comprises a deep learning classification model, and is used for inputting the detection image set into the deep learning classification model to obtain a target structural layer output by the deep learning classification model, wherein the target structural layer is the structural layer with highest defect definition in the plurality of structural layers; The defect extraction module is connected with the defect layering module and comprises a deep learning segmentation model, and the defect extraction module is used for inputting a detection image corresponding to the target structural layer into the deep learning segmentation model to obtain a defect detection result of the panel to be detected, which is output by the deep learning segmentation model. According to the panel defect detection system, the defect layering module and the defect extraction module are arranged, the defect layering module adopts the deep learning classification model to perform defect layering and output a target structural layer with the clearest defects, the defect extraction module adopts the deep learning segmentation model to perform defect detection and output a defect detection result, a complex manual parameter adjustment process is not needed, the detection speed and the layering judgment flexibility are greatly improved, the rechecking efficiency and the system usability are remarkably improved, and the severe requirements of panel mass production on stability and precision can be met. According to one embodiment of the application, the deep learning segmentation model comprises a first processing unit, a second processing unit and a third processing unit which are sequentially connected; the first processing unit is used for extracting features of the detection image corresponding to the target structural layer and generating a feature thermodynamic diagram; the second processing unit is used for generating a defect segmentation mask map based on the characteristic thermodynamic diagram; the third processing unit is configured to g