WO2022109949A1 - 不良图片缺陷等级识别的方法、装置及存储介质 - Google Patents

不良图片缺陷等级识别的方法、装置及存储介质 Download PDF

Info

Publication number
WO2022109949A1
WO2022109949A1 PCT/CN2020/131918 CN2020131918W WO2022109949A1 WO 2022109949 A1 WO2022109949 A1 WO 2022109949A1 CN 2020131918 W CN2020131918 W CN 2020131918W WO 2022109949 A1 WO2022109949 A1 WO 2022109949A1
Authority
WO
WIPO (PCT)
Prior art keywords
defect
size
bad
determined
level
Prior art date
Application number
PCT/CN2020/131918
Other languages
English (en)
French (fr)
Inventor
周全国
徐开琴
邹佳洪
张国林
黄巡
张青
周丽佳
王志东
沈鸿翔
唐浩
程久阳
Original Assignee
京东方科技集团股份有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 京东方科技集团股份有限公司 filed Critical 京东方科技集团股份有限公司
Priority to US18/038,197 priority Critical patent/US20230419466A1/en
Priority to PCT/CN2020/131918 priority patent/WO2022109949A1/zh
Priority to CN202080003018.0A priority patent/CN115023603A/zh
Publication of WO2022109949A1 publication Critical patent/WO2022109949A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/0002—Inspection of images, e.g. flaw detection
    • G06T7/0004—Industrial image inspection
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/0002—Inspection of images, e.g. flaw detection
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
    • G01N21/84—Systems specially adapted for particular applications
    • G01N21/88—Investigating the presence of flaws or contamination
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T3/00—Geometric image transformations in the plane of the image
    • G06T3/40—Scaling of whole images or parts thereof, e.g. expanding or contracting
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20084—Artificial neural networks [ANN]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30108—Industrial image inspection
    • G06T2207/30121—CRT, LCD or plasma display
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30168—Image quality inspection

Definitions

  • the present disclosure relates to the field of image recognition, and in particular, to a method, device and storage medium for identifying defect levels of bad pictures.
  • the defects occurring in each process of the product can be detected by automatic optical inspection (Auto Optical Inspection, AOI), and the defective pictures containing the defects can be taken.
  • Automatic optical inspection Auto Optical Inspection, AOI
  • the present disclosure provides a method, device and storage medium for identifying defect levels of bad pictures, which are used to solve the technical problem in the prior art that bad pictures that have a substantial impact on product yield cannot be quickly and accurately determined.
  • the technical solution of a method for identifying defect levels of bad pictures is as follows:
  • the defect level of the defect is determined; wherein, the defect level is the degree of influence of the defect on the product yield.
  • a possible implementation, to determine the defect size of the defect from the bad picture includes:
  • the maximum length of the defect in the first direction and the second direction is determined as the defect size; wherein the first direction is perpendicular to the second direction.
  • a possible implementation manner according to the defect type to which the defect belongs, and the size relationship between the defect size and the design size, determine the defect level of the defect, including:
  • the grade defect is determined according to the judgment result.
  • the bad type when the bad picture is a bad picture in the display panel, the bad type includes bad particles and bad breakage of the passivation layer.
  • the defect type to which the defect belongs is determined by the neural network model in the defect automatic classification system, and the defect level judgment condition used for the defect is determined, including:
  • the defect type is particle defect
  • the value range of the set magnification is 1.3-1.7;
  • the failure type is failure of passivation layer fracture
  • the conditions for determining poor particles include:
  • the defect level is determined to be defective Quality risk; if the ratio of the defect size to the design size of the channel is greater than the set magnification, the defect level is determined to be bad without quality risk;
  • the defect level is determined to be defective with quality risk; if the ratio of the defect size to the design size of the intersection position is greater than the set magnification, the defect level is determined to be Bad quality without risk.
  • the conditions for determining the poor fracture of the passivation layer include:
  • the defect level is bad and repairable
  • the design distance between the defect size and the two adjacent data lines is greater than the set magnification, it is determined that the defect level is non-repairable.
  • a possible implementation manner, after determining the defect level of the bad picture, further includes:
  • the defects are marked with different marks for different defect grades.
  • an embodiment of the present disclosure provides a device for identifying defect levels of bad pictures, including:
  • Defect size determination unit which is used to determine the defect size of defects from bad pictures
  • a design size determination unit configured to determine the design size of the pattern corresponding to the adjacent parts at the location of the defect according to the product model corresponding to the defective picture;
  • a defect level determination unit configured to determine the defect level of the defect according to the defect type to which the defect belongs and the size relationship between the defect size and the design size; wherein, the defect level is the defect-to-product yield rate degree of influence.
  • the defect size determination unit is used for:
  • the maximum length of the defect in the first direction and the second direction is determined as the defect size; wherein the first direction is perpendicular to the second direction.
  • the defect level determination unit is used for:
  • the grade defect is determined according to the judgment result.
  • the bad type when the bad picture is a bad picture in the display panel, the bad type includes bad particles and bad breakage of the passivation layer.
  • the defect level determination unit is further used for:
  • the defect type is particle defect
  • the value range of the set magnification is 1.3-1.7;
  • the failure type is failure of passivation layer fracture
  • the conditions for determining poor particles include:
  • the defect is determined The grade is bad with quality risk; if the ratio of the defect size to the channel is greater than the set magnification, the defect grade is determined to be bad without quality risk;
  • the defect level is determined to be defective with quality risk; if the ratio of the defect size to the design size of the intersection position is greater than the set magnification, the defect level is determined to be Bad quality without risk.
  • the PVX Open bad judgment condition includes:
  • the defect level is bad and repairable
  • the design distance between the defect size and the two adjacent data lines is greater than the set magnification, it is determined that the defect level is non-repairable.
  • the defect level determination unit is further used for:
  • the defects are marked with different marks for different defect grades.
  • an embodiment of the present disclosure provides an apparatus for automatically identifying a defect level of a bad picture, including: at least one processor, and
  • a memory connected to the at least one processor
  • the memory stores instructions that can be executed by the at least one processor, and the at least one processor executes the above-mentioned automatic identification method for defective picture defect levels by executing the instructions stored in the memory.
  • an embodiment of the present disclosure further provides a device for automatically identifying defect levels of bad pictures, including:
  • a memory connected to the at least one processor
  • the memory stores instructions executable by the at least one processor, and the at least one processor executes the method according to the first aspect above by executing the instructions stored in the memory.
  • an embodiment of the present disclosure further provides a readable storage medium, including:
  • the memory is used to store instructions that, when executed by the processor, cause an apparatus including the readable storage medium to perform the method as described in the first aspect above.
  • FIG. 1 is a flowchart for identifying a defect level of a bad picture according to an embodiment of the present disclosure
  • FIG. 2 is a schematic diagram of determining a defect size of a defect from a bad picture according to an embodiment of the present disclosure
  • FIG. 3 is a schematic diagram of particle defects located in a channel of a thin film transistor according to an embodiment of the present disclosure
  • FIG. 4 is a schematic diagram of particle defects located at the intersections of scan lines and data lines according to an embodiment of the present disclosure
  • FIG. 5 is a schematic diagram illustrating that the particle defect provided by the embodiment of the present disclosure is a poor fracture of the passivation layer
  • FIG. 6 is a schematic structural diagram of a defect level identification of a bad picture according to an embodiment of the present disclosure.
  • Embodiments of the present disclosure provide a method, a device, and a storage medium for automatically identifying the defect level of a bad picture, so as to solve the technology in the prior art that cannot quickly and accurately determine a bad picture that has a substantial impact on product yield. question.
  • an embodiment of the present disclosure provides a method for automatically identifying a defect level of a bad picture, and the processing process of the method is as follows.
  • Step 101 Determine the defect size of the defect from the bad picture.
  • the maximum length of the defect in the first direction and the second direction can be determined as the defect size; wherein the first direction is perpendicular to the second direction.
  • FIG. 2 a schematic diagram of determining a defect size of a defect from a bad picture according to an embodiment of the present disclosure.
  • the defect in Fig. 2 is a particle (Particle, PT) defect, as shown in the black continuous area in the dotted line area in Fig. 2.
  • the maximum length of the PT defect is determined as X1 in the first direction (X) shown in Fig. 2.
  • the two directions (Y) determine the maximum length of the PT defect as Y1, and take the larger of X1 and Y1 as the defect size of the PT defect.
  • step 101 can be executed.
  • Step 102 Determine the design size of the pattern corresponding to the adjacent parts at the location of the defect according to the product model corresponding to the defective picture.
  • the pattern corresponding to the adjacent components at the location of the defect may be, for example, the pattern corresponding to components such as components, traces, and film layers located around the defect in the inspected product (eg, display panel, circuit board, chip).
  • the pattern corresponding to the adjacent components at the location of the defect can be Thin Film Transistor (TFT), Data Line, and Gate Line.
  • TFT Thin Film Transistor
  • Data Line Data Line
  • Gate Line You can obtain the relevant design dimensions of TFT, data lines, and scan lines, such as the design dimensions of the channel (Channel) in the TFT, the width of the data lines, and the design spacing.
  • step 103 After determining the design size of the pattern corresponding to the adjacent component at the location of the defect, step 103 can be executed.
  • Step 103 Determine the defect level of the defect according to the defect type to which the defect belongs and the relationship between the defect size and the design size; wherein, the defect level is the degree of influence of the defect on the product yield.
  • the bad type of the defect in the bad picture may include bad particle (PT bad), bad passivation layer fracture (Passivation Open, PVX Open) and the like.
  • the defect level may be, for example, defective without quality risk, defective with quality risk, defective can be repaired, defective can not be repaired, and the like. Of course, these defect levels can also be quantified. For example, 0 means bad without quality risk, 3 means bad can be repaired, 8 means bad cannot be repaired, and 9 means bad with quality risk; or graded, such as high, The risk levels are medium, low, and no. No means bad and no quality risk, low means bad can be repaired, high means bad cannot be repaired, and high means bad has quality risk, etc. Defect levels can be designed according to actual needs, and are not limited to the above examples.
  • the defect level of the defect in the bad picture is determined according to the bad type to which the defect belongs in the bad picture, and the size relationship between the defect size and the design size, which can be achieved in the following ways:
  • the defective type of the defect in the defective image is judged, and the defect level judgment condition used for the defect in the defective image is determined; The ratio of the size and the size relationship of the set magnification are judged to obtain the judgment result; the grade defect is determined according to the judgment result.
  • ADC Automatic Defect Classification
  • the neural network model of the ADC it is usually necessary for the user to use the collected sample pictures to train the model (such as a neural network model), and after the training is completed and reaches After the online use standard, the trained model is used to automatically obtain bad pictures from a specific storage location (such as a shared disk dedicated to storing bad pictures output by AOI), and classify the bad pictures.
  • the classified pictures can be classified according to bad types (such as The presence of defective particles or defective passivation layer fracture) is marked.
  • Training a neural network model usually requires collecting enough different types of sample images, labeling the sample images by type, and then dividing the sample images into a training set and a validation set, using the training set to train the model, and using the validation set for training. After the model is verified, the user also manually judges the bad pictures in the verification set, and determines the accuracy and recall rate of the trained model based on the results of the manual judgment. After setting the value, it is determined that the model is successfully trained and can be automatically identified online. If any one does not reach the corresponding preset value, add samples for retraining and verification until both the precision and recall rate reach their respective preset values.
  • the trained model After the trained model is online, it will automatically obtain the bad pictures taken by AOI from a specific storage location for identification, and output a corresponding confidence level after automatic identification of each picture.
  • the confidence level > 90% can be determined as the bad classification judgment is correct. If the confidence level is > 90%, the bad picture can enter the bad positioning, labeling, and classification links; After extraction, these bad pictures are used to update and iterate the trained model to improve the algorithm's ability to identify bad pictures in the future.
  • the trained model is the neural network model used online in the ADC system.
  • the neural network model can be used to classify bad pictures to determine whether the bad type of bad pictures is bad particles (poor PT) or bad passivation layer fracture ( Bad PVX Open).
  • the defect level judgment condition used for the defect in the defective image can be determined, and the corresponding defect level can be set for each defect type in advance Judgment condition:
  • the defect type is particle defect (PT defect)
  • the particle defect (PT defect) judgment condition for the defective image it is determined to use the particle defect (PT defect) judgment condition for the defective image, and the value range of the set magnification is 1.3-1.7.
  • the set magnification may be 1.5.
  • the conditions for determining bad particles can be:
  • the defect level is determined to be poor and of good quality Risk; if the ratio of the defect size to the design size of the channel is greater than the set magnification, the defect level is determined to be bad without quality risk.
  • FIG. 3 is a schematic diagram of particle defects located in the channel of a thin film transistor provided by an embodiment of the present disclosure. It is assumed that the defect in FIG. 3 belongs to PT failure through ADC, the magnification is set to 1.5, and the PT in FIG. 3 has been determined.
  • the defect size of the defect is L1
  • the design size of the channel of the TFT (the U-shaped area with a diagonal line in Figure 3, that is, the vertical space corresponding to the U-shaped area above the gate of the TFT, between the source and drain) is S1
  • X1/S1 ⁇ 1.5 it can be determined that the defect level of the PT defect in Figure 3 is bad with quality risk
  • X1/S1>1.5 it can be determined that the defect level of the PT defect in Figure 3 is bad without quality risk .
  • the defect is black, and the ratio of the defect size to the design size of the intersection position is less than or equal to the set magnification, the defect is determined.
  • the grade is bad with quality risk; if the ratio of the defect size to the design size of the intersection position is greater than the set magnification, the defect grade is determined to be bad without quality risk.
  • FIG. 4 is a schematic diagram of particle defects located at intersections of scan lines and data lines according to an embodiment of the present disclosure.
  • Fig. 4 is a PT defect, and the set magnification is 1.5.
  • the defect size of the elliptical PT defect is L1
  • the defect size of the circular PT defect is L2
  • the design size of the intersection position is S2.
  • L2 ⁇ S2 (satisfies L2/S2 ⁇ 1.5)
  • L1/S2>1.5 so it can be determined
  • the defect level of the oval PT defect at the intersection in FIG. 4 is bad without quality risk.
  • the defect type is poor passivation layer fracture (PVX Open)
  • the value range of the set magnification is 0.8 to 1.2.
  • the value of the set magnification is 1.
  • the conditions for the judgment of poor passivation layer fracture can be:
  • the defect level is determined to be defective and repairable .
  • the defect level is determined to be non-repairable.
  • FIG. 5 is a schematic diagram showing that the particle defect is a poor fracture of the passivation layer provided by the embodiment of the present disclosure.
  • the defect in Figure 5 belongs to poor passivation layer fracture (PVX Open) through ADC
  • the magnification is set to 1
  • the distance between the two data lines in Figure 5 is d
  • the defect size of the oval PVX Open defect is L1
  • the defect size of the circular PVX Open defect is L2.
  • L1>d it can be determined that the defect level of the elliptical PT defect is not good and cannot be repaired; in Figure 5, because L2 ⁇ d, it can be determined that the circular The defect level of the PT defect is bad repairable.
  • the passivation layer is usually located on the upper layer of the data line, it can be understood that the passivation layer is transparent in FIG. 5 , and the circles and ellipses in FIG. 5 indicate fractures in the passivation layer.
  • the defects may also be marked with different marks for different defect levels.
  • bad pictures can be separated by defect level. For example, if it is a bad picture that cannot be repaired, the label of ADC will be used; other defect levels will be relabeled. If the defect level of the defect in the bad picture is bad and repairable, it will be marked with a red round frame.
  • an embodiment of the present disclosure provides a device for identifying the defect level of a bad picture.
  • the device includes:
  • a design size determination unit 602 configured to determine the design size of the pattern corresponding to the adjacent parts at the location of the defect according to the product model corresponding to the defective picture;
  • Defect level determination unit 603 configured to determine the defect level of the defect according to the defect type to which the defect belongs, and the size relationship between the defect size and the design size; rate of influence.
  • the defect size determination unit is used for:
  • the maximum length of the defect in the first direction and the second direction is determined as the defect size; wherein the first direction is perpendicular to the second direction.
  • the defect level determination unit 603 is used for:
  • the grade defect is determined according to the judgment result.
  • the bad type when the bad picture is a bad picture in the display panel, the bad type includes bad particles and bad breakage of the passivation layer.
  • the defect level determination unit 603 is further configured to:
  • the defect type is particle defect
  • the value range of the set magnification is 1.3-1.7;
  • the failure type is failure of passivation layer fracture
  • the conditions for determining poor particles include:
  • the defect level is determined Defect with quality risk; if the ratio of the defect size to the design size of the channel is greater than the set magnification, determine the defect level as poor without quality risk;
  • the defect level is determined to be defective with quality risk; if the ratio of the defect size to the design size of the intersection position is greater than the set magnification, the defect level is determined to be Bad quality without risk.
  • the conditions for determining the poor fracture of the passivation layer include:
  • the defect level is bad and repairable
  • the defect level is determined to be defective and irreparable.
  • the defect level determination unit 603 is further configured to:
  • the defects are marked with different marks for different defect grades.
  • the embodiments of the present disclosure provide an apparatus for automatically identifying defect levels of bad pictures, including: at least one processor, and
  • a memory connected to the at least one processor
  • the memory stores instructions executable by the at least one processor, and the at least one processor executes the above-mentioned method for identifying defect levels of bad pictures by executing the instructions stored in the memory.
  • an embodiment of the present disclosure also provides a readable storage medium, including:
  • the memory is used to store instructions, and when the instructions are executed by the processor, make the apparatus including the readable storage medium complete the above-mentioned method for identifying a defect level of a bad picture.
  • embodiments of the present disclosure may be provided as a method, system, or computer program product. Accordingly, embodiments of the present disclosure may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, embodiments of the present disclosure may take the form of a computer program product implemented on one or more computer-usable storage media having computer-usable program code embodied therein, including but not limited to disk storage, CD-ROM, optical storage, and the like.
  • Embodiments of the present disclosure are described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the present disclosure. It will be understood that each process and/or block in the flowchart illustrations and/or block diagrams, and combinations of processes and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to the processor of a general purpose computer, special purpose computer, embedded processor or other programmable data processing device to produce a machine such that the instructions executed by the processor of the computer or other programmable data processing device produce Means for implementing the functions specified in a flow or flow of a flowchart and/or a block or blocks of a block diagram.
  • These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory result in an article of manufacture comprising instruction means, the instructions
  • the apparatus implements the functions specified in the flow or flows of the flowcharts and/or the block or blocks of the block diagrams.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Quality & Reliability (AREA)
  • General Health & Medical Sciences (AREA)
  • Biochemistry (AREA)
  • Immunology (AREA)
  • Pathology (AREA)
  • Analytical Chemistry (AREA)
  • Chemical & Material Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Investigating Materials By The Use Of Optical Means Adapted For Particular Applications (AREA)

Abstract

一种不良图片缺陷等级识别的方法、装置及存储介质，用以解决不能快速、准确的判断出对产品良率有实质性影响的不良图片的技术问题，该方法包括：从不良图片中确定出缺陷的缺陷尺寸(101)；根据不良图片对应的产品型号，确定缺陷所在位置相邻部件对应的图案的设计尺寸(102)；根据缺陷所属的不良类型，以及缺陷尺寸与设计尺寸大小关系，确定缺陷的缺陷等级；其中，缺陷等级为缺陷对产品良率的影响程度(103)。

Description

不良图片缺陷等级识别的方法、装置及存储介质 技术领域
本公开涉及图像识别领域，尤其是涉及不良图片缺陷等级识别的方法、装置及存储介质。
背景技术
在显示面板的背板生产过程中，通常包括薄膜沉积、曝光(photo)和显影刻蚀等工序。
在生产的过程中，由于制造设备、制造工艺、工艺环境、工艺材料、像素布局设计等因素的影响，极易造成显示面板在制造的过程中产生缺陷。而这些缺陷又会大幅影响产品的良率，进而造成生产成本的提升。
目前，在显示面板的背板生产过程中，可以通过自动光学检测(Auto Optical Inspection，AOI)检出产品在各个工序中出现的缺陷，并拍摄包含该缺陷的不良图片。
之后，采用人工的方式从每类不良图片中，判断出对产品良率有实质性影响的不良图片，常常会出现遗漏、误判等问题，且人工判图还存在耗时长、效率低、无法及时进行不良预警、降低关键工序的维修等问题。
鉴于此，如何快速、准确的判断出对产品良率有实质性影响的不良图片，成为一个亟待解决的技术问题。
发明内容
本公开提供一种不良图片缺陷等级识别的方法、装置及存储介质，用以解决现有技术中存在的不能快速、准确的判断出对产品良率有实质性影响的不良图片的技术问题。
第一方面，为解决上述技术问题，本公开实施例提供的一种不良图片缺陷等级识别的方法的技术方案如下：
从不良图片中确定出缺陷的缺陷尺寸；
根据所述不良图片对应的产品型号，确定所述缺陷所在位置相邻部件对应的图案的设计尺寸；
根据所述缺陷所属的不良类型，以及所述缺陷尺寸与所述设计尺寸大小关系，确定所述缺陷的缺陷等级；其中，所述缺陷等级为缺陷对产品良率的影响程度。
一种可能的实施方式，从不良图片中确定出缺陷的缺陷尺寸，包括：
将所述缺陷在第一方向和第二方向中的最大长度，确定为所述缺陷尺寸；其中，所述第一方向垂直于所述第二方向。
一种可能的实施方式，根据所述缺陷所属的不良类型，以及所述缺陷尺寸与所述设计尺寸的大小关系，确定所述缺陷的缺陷等级，包括：
过缺陷自动化分类系统中的神经网络模型判断出所述缺陷所属的不良类型，并确定对所述缺陷使用的缺陷等级判定条件；
用所述判定条件，对所述缺陷尺寸与所述设计尺寸的比值和设定倍率的大小关系进行判定，获得判定结果；
根据所述判定结果确定所述等级缺陷。
一种可能的实施方式，所述不良图片为显示面板中的不良图片时，所述不良类型包括颗粒不良、钝化层断裂不良。
一种可能的实施方式，通过缺陷自动化分类系统中的神经网络模型判断出所述缺陷所属的不良类型，确定对所述缺陷使用的缺陷等级判定条件，包括：
当所述不良类型为颗粒不良时，确定对所述不良图片使用颗粒不良判定条件，所述设定倍率的取值范围为1.3～1.7；
当所述不良类型为钝化层断裂不良时，确定对所述不良图片使用钝化层断裂不良判定条件，所述设定倍率的取值范围为0.8～1.2。
一种可能的实施方式，所述颗粒不良判定条件，包括：
当所述缺陷所在位置周围的图案为薄膜晶体管的沟道时，若所述缺陷尺 寸与所述沟道的设计尺寸的比值小于或等于所述设定倍率，则确定所述缺陷等级为不良有品质风险；若所述缺陷尺寸与所述沟道的设计尺寸的比值大于所述设定倍率，则确定所述缺陷等级为不良无品质风险；
当所述缺陷所在位置周围的图案为所述显示面板的扫描线和数据线的交叉位置对应的图案时，若所述缺陷为黑色，且所述缺陷尺寸与所述交叉位置的设计尺寸的比值小于或等于所述设定倍率，则确定所述缺陷等级为不良有品质风险；若所述缺陷尺寸与所述交叉位置的设计尺寸的比值大于所述设定倍率，则确定所述缺陷等级为不良无品质风险。
一种可能的实施方式，所述钝化层断裂不良判定条件，包括：
当所述缺陷所在位置周围的图案为所述显示面板中数据线对应的图案时，若所述缺陷尺寸与相邻两个数据线之间的设计间距小于或等于所述设定倍率，则确定所述缺陷等级为不良可维修；
若所述缺陷尺寸与所述相邻两个数据线之间的设计间距大于所述设定倍率，则确定所述缺陷等级为不良不可维修。
一种可能的实施方式，确定所述不良图片的缺陷等级后，还包括：
对不同缺陷等级用不同的标记对所述缺陷进行标注。
第二方面，本公开实施例提供了一种不良图片缺陷等级识别的装置，包括：
缺陷尺寸确定单元，用于从不良图片中确定出缺陷的缺陷尺寸；
设计尺寸确定单元，用于根据所述不良图片对应的产品型号，确定所述缺陷所在位置相邻部件对应的图案的设计尺寸；
缺陷等级确定单元，用于根据所述缺陷所属的不良类型，以及所述缺陷尺寸与所述设计尺寸的大小关系，确定所述缺陷的缺陷等级；其中，所述缺陷等级为缺陷对产品良率的影响程度。
一种可能的实施方式，所述缺陷尺寸确定单元用于：
将所述缺陷在第一方向和第二方向中的最大长度，确定为所述缺陷尺寸；其中，所述第一方向垂直于所述第二方向。
一种可能的实施方式，所述缺陷等级确定单元用于：
通过缺陷自动化分类系统中的神经网络模型判断出所述缺陷所属的不良类型，并确定对所述缺陷使用的缺陷等级判定条件；
用所述判定条件，对所述缺陷尺寸与所述设计尺寸的比值和设定倍率的大小关系进行判定，获得判定结果；
根据所述判定结果确定所述等级缺陷。
一种可能的实施方式，所述不良图片为显示面板中的不良图片时，所述不良类型包括颗粒不良、钝化层断裂不良。
一种可能的实施方式，所述缺陷等级确定单元还用于：
当所述不良类型为颗粒不良时，确定对所述不良图片使用颗粒不良判定条件，所述设定倍率的取值范围为1.3～1.7；
当所述不良类型为钝化层断裂不良时，确定对所述不良图片使用钝化层断裂不良判定条件，所述设定倍率的取值范围为0.8～1.2。
一种可能的实施方式，所述颗粒不良判定条件，包括：
当所述缺陷所在位置周围的图案为位于薄膜晶体管的沟道对应的图案时，若所述缺陷尺寸与所述沟道的设计尺寸的比值小于或等于所述设定倍率，则确定所述缺陷等级为不良有品质风险；若所述缺陷尺寸与所述沟道的比值大于所述设定倍率，则确定所述缺陷等级为不良无品质风险；
当所述缺陷所在位置周围的图案为所述显示面板的扫描线和数据线的交叉位置对应的图案时，若所述缺陷为黑色，且所述缺陷尺寸与所述交叉位置的设计尺寸的比值小于或等于所述设定倍率，则确定所述缺陷等级为不良有品质风险；若所述缺陷尺寸与所述交叉位置的设计尺寸的比值大于所述设定倍率，则确定所述缺陷等级为不良无品质风险。
一种可能的实施方式，所述PVX Open不良判定条件，包括：
当所述缺陷所在位置周围的图案为所述显示面板中数据线对应的图案时，若所述缺陷尺寸与相邻两个数据线之间的设计间距小于或等于所述设定倍率，则确定所述缺陷等级为不良可维修；
若所述缺陷尺寸与所述相邻两个数据线之间的设计间距大于所述设定倍率，则确定所述缺陷等级为不良不可维修。
一种可能的实施方式，所述缺陷等级确定单元还用于：
对不同缺陷等级用不同的标记对所述缺陷进行标注。
基于同一发明构思，本公开实施例中提供了一种不良图片缺陷等级自动识别的装置，包括：至少一个处理器，以及
与所述至少一个处理器连接的存储器；
其中，所述存储器存储有可被所述至少一个处理器执行的指令，所述至少一个处理器通过执行所述存储器存储的指令，执行如上所述的不良图片缺陷等级自动识别方法。
第三方面，本公开实施例还提供一种不良图片缺陷等级自动识别的装置，包括：
至少一个处理器，以及
与所述至少一个处理器连接的存储器；
其中，所述存储器存储有可被所述至少一个处理器执行的指令，所述至少一个处理器通过执行所述存储器存储的指令，执行如上述第一方面所述的方法。
第四方面，本公开实施例还提供一种可读存储介质，包括：
存储器，
所述存储器用于存储指令，当所述指令被处理器执行时，使得包括所述可读存储介质的装置完成如上述第一方面所述的方法。
附图说明
图1为本公开实施例提供的一种不良图片缺陷等级识别的流程图；
图2为本公开实施例提供的从不良图片中确定缺陷的缺陷尺寸的示意图；
图3为本公开实施例提供的颗粒缺陷位于薄膜晶体管的沟道的示意图；
图4为本公开实施例提供的颗粒缺陷位于扫描线和数据线的交叉位置的 示意图；
图5为本公开实施例提供的颗粒缺陷为钝化层断裂不良的示意图；
图6为本公开实施例提供的一种不良图片缺陷等级识别的结构示意图。
具体实施方式
本公开实施列提供一种不良图片缺陷等级自动识别的方法、装置及存储介质，用以解决现有技术中存在的不能快速、准确的判断出对产品良率有实质性影响的不良图片的技术问题。
为了更好的理解上述技术方案，下面通过附图以及具体实施例对本公开技术方案做详细的说明，应当理解本公开实施例以及实施例中的具体特征是对本公开技术方案的详细的说明，而不是对本公开技术方案的限定，在不冲突的情况下，本公开实施例以及实施例中的技术特征可以相互组合。
请参考图1，本公开实施例提供一种不良图片缺陷等级自动识别的方法，该方法的处理过程如下。
步骤101：从不良图片中确定出缺陷的缺陷尺寸。
在获得不良图片后，可以将缺陷在第一方向和第二方向中的最大长度，确定为缺陷尺寸；其中，第一方向垂直于第二方向。
请参见图2为本公开实施例提供的从不良图片中确定缺陷的缺陷尺寸的示意图。图2中的缺陷为颗粒(Particle，PT)缺陷如图2中虚线区域内的黑色连续区域所示，在图2所示的第一方向(X)确定出PT缺陷的最大长度为X1，第二方向(Y)确定出PT缺陷的最大长度为Y1，取X1、Y1中的较大者作为PT缺陷的缺陷尺寸。
在确定出缺陷的缺陷长度之后，便可执行步骤101。
步骤102：根据不良图片对应的产品型号确定缺陷所在位置相邻部件对应的图案的设计尺寸。
缺陷所在位置相邻部件对应的图案例如可以是被检测产品(如显示面板、电路板、芯片)中位于缺陷周围的器件、走线、膜层等部件对应的图案。
例如，在被检测产品为显示面板时，缺陷所在位置相邻部件对应的图案可以为薄膜晶体管(Thin Film Transistor，TFT)、数据线(Data Line)、扫描线(Gate Line)，根据不良图片对应的产品型号可以获取TFT、数据线、扫描线的相关设计尺寸，如TFT中沟道(Channel)的设计尺寸、数据线的宽度、设计间距等。
在确定缺陷所在位置相邻部件对应的图案的设计尺寸后，便可执行步骤103。
步骤103：根据缺陷所属的不良类型，以及缺陷尺寸与设计尺寸大小关系，确定缺陷的缺陷等级；其中，缺陷等级为缺陷对产品良率的影响程度。
不良图片为显示面板中的不良图片时，不良图片中缺陷的不良类型可以包括颗粒不良(PT不良)、钝化层断裂不良(Passivation Open，PVX Open)等。
缺陷等级例如可以是不良无品质风险、不良有品质风险、不良可维修、不良不可维修等。当然，也可以将这些缺陷等级数值化，如用0代表不良无品质风险，用3代表不良可维修，用8代表不良不可维修、用9代表不良有品质风险；或等级化，如用高、中、低、无等风险等级，用无代表不良无品质风险，用低代表不良可维修，用高代表不良不可维修、用高代表不良有品质风险等。可以根据实际需要设计缺陷等级，而不限于上述举例。
在本公开提供的实施例中，根据不良图片中缺陷所属的不良类型，以及缺陷尺寸与设计尺寸的大小关系，确定不良图片中缺陷的缺陷等级，可以通过下列方式实现：
通过缺陷自动化分类系统(Automatic Defect Classification，ADC)中的神经网络模型判断出不良图片中缺陷所属的不良类型，确定对不良图片中缺陷使用的缺陷等级判定条件；用判定条件，对缺陷尺寸与设计尺寸的比值和设定倍率的大小关系进行判定，获得判定结果；根据判定结果确定等级缺陷。
在本公开提供的实施例中，通过ADC的神经网络模型判断出不良图片中缺陷所属的不良类型，通常需要用户用收集的样本图片进行模型(如神经网 络模型)训练，并在训练完成并达到上线使用标准后，采用训练好的模型自动从特定的存储位置(如专用于存放AOI输出不良图片的共享盘)获取不良图片，并将不良图片进行分类，分类后的图片可以按不良类型(如存在颗粒不良还是钝化层断裂不良)进行标注。
对神经网络模型进行训练通常需要收集足够的不同种类的样本图片，并将样本图片按类型进行标注好，然后把样本图片分为训练集和验证集，用训练集训练模型，用验证集对训练后的模型进行验证，同时用户也对验证集中的不良图片进行人工判断，以人工判断的结果为基准，确定训练后的模型的准确率和召回率，当准确率和召回率都达到各自的预设值之后，确定模型训练成功，可以上线自动识别。若任一个未达到对应的预设值则增加样本进行重新训练和验证，直至准确率和召回率都达到各自的预设值。
训练好的模型上线后，将自动从特定的存储位置中获取AOI拍摄的不良图片进行识别，在对每张图片进行自动识别后输出一个对应的置信度。可以将置信度>90％确定为不良分类判断正确，如果置信度>90％，不良图片就可以进入不良定位及标注、分类环节；如果置信度<90％，不良图片通过人工进行不良特征判定并提取后，用这些不良图片再对训练好的模型进行更新迭代，以提高算法后续对不良图片的识别能力。该训练好的模型即为ADC系统中上线使用的神经网络模型，用该神经网络模型可以对不良图片进行分类，以确定不良图片的不良类型是颗粒不良(PT不良)还是钝化层断裂不良(PVX Open不良)。
在本公开提供的实施例中，在确定不良图片中的缺陷所属的不良类型后，便可确定对不良图片中缺陷使用的缺陷等级判定条件，可以预先为每种不良类型设定对应的缺陷等级判定条件：
在被检测产品为显示面板时，当不良类型为颗粒不良(PT不良)时，确定对不良图片使用颗粒不良(PT不良)判定条件，设定倍率的取值范围为1.3～1.7。优选地，设定倍率可以为1.5。
颗粒不良判定条件，可以是：
当缺陷所在位置周围的图案为薄膜晶体管(TFT)的沟道(Channel)对应的图案时，若缺陷尺寸与沟道的设计尺寸的比值小于或等于设定倍率，则确定缺陷等级为不良有品质风险；若缺陷尺寸与沟道的设计尺寸的比值大于设定倍率，则确定缺陷等级为不良无品质风险。
请参见图3为本公开实施例提供的颗粒缺陷位于薄膜晶体管的沟道的示意图，假设通过ADC确定图3中的缺陷属于PT不良，设定倍率为1.5，并且已确定出图3中的PT缺陷的缺陷尺寸为L1，TFT的沟道(图3中具有斜线的U形区域，即TFT的栅极之上、源极和漏极之间的U形区域对应的垂直空间)的设计尺寸为S1，当X1/S1≤1.5时可以确定图3中的PT缺陷的缺陷等级为不良有品质风险，当X1/S1>1.5时可以确定图3中的PT缺陷的缺陷等级为不良无品质风险。
当缺陷所在位置周围的图案为显示面板的扫描线和数据线的交叉位置对应的图案时，若缺陷为黑色，且缺陷尺寸与交叉位置的设计尺寸的比值小于或等于设定倍率，则确定缺陷等级为不良有品质风险；若缺陷尺寸与交叉位置的设计尺寸的比值大于设定倍率，则确定缺陷等级为不良无品质风险。
例如，请参见图4为本公开实施例提供的颗粒缺陷位于扫描线和数据线的交叉位置的示意图。假设通过ADC确定图4属于PT不良，设定倍率为1.5，在图4中椭圆形PT缺陷的缺陷尺寸为L1、圆形PT缺陷的缺陷尺寸为L2，交叉位置的设计尺寸为S2，在图4中L2<S2(满足L2/S2≤1.5)，因此可以确定图4中的交叉位置的圆形PT缺陷的缺陷等级为不良有品质风险；在图4中L1/S2>1.5，因此可以确定图4中的交叉位置的椭圆形PT缺陷的缺陷等级为不良无品质风险。
当不良类型为钝化层断裂(PVX Open)不良时，确定对不良图片中缺陷使用钝化层断裂不良判定条件，设定倍率的取值范围为0.8～1.2。优选地，设定倍率的取值为1。
钝化层断裂不良判定条件，可以是：
当缺陷所在位置周围的图案为显示面板中数据线对应的图案时，若缺陷 尺寸与相邻两个数据线之间的设计间距的比值小于或等于设定倍率，则确定缺陷等级为不良可维修。
若缺陷尺寸与相邻两个数据线之间的设计间距的比值大于设定倍率，则确定缺陷等级为不良不可维修。
例如，请参见图5为本公开实施例提供的颗粒缺陷为钝化层断裂不良的示意图。假设通过ADC确定图5中缺陷属于钝化层断裂(PVX Open)不良，设定倍率为1，在图5中两个数据线之间的间距为d，椭圆形PVX Open缺陷的缺陷尺寸为L1、圆形PVX Open缺陷的缺陷尺寸为L2，在图5中由于L1>d，因此可以确定椭圆形PT缺陷的缺陷等级为不良不可维修；在图5中由于L2<d，因此可以确定圆形PT缺陷的缺陷等级为不良可维修。
需要说明的是，由于钝化层通常位于数据线上层，因此在图5中可以理解为钝化层为透明的，图5中的圆形和椭圆形示意钝化层中存在的断裂。
在本公开提供的实施例中，在确定不良图片的缺陷等级后，还可以对不同缺陷等级用不同的标记对缺陷进行标注。此外，还可以按缺陷等级将不良图片分开放置。例如，若为不良不可维修图片，则沿用ADC的标注；对其它缺陷等级进行重新标注，如不良图片中缺陷的缺陷等级为不良可维修，则用红色圆形框标注。
需要理解的是，本公开提供的上述方案，也可以应用于其它技术领域，如对PCB电路板中的缺陷等级进行自动识别，具体可以通过改变缺陷等级判定条件来实现。
基于同一发明构思，本公开一实施例中提供一种不良图片缺陷等级识别的装置，该装置的不良图片缺陷等级自动识别方法的具体实施方式可参见方法实施例部分的描述，重复之处不再赘述，请参见图6，该装置包括：
缺陷尺寸确定单元601，用于从不良图片中确定出缺陷的缺陷尺寸；
设计尺寸确定单元602，用于根据所述不良图片对应的产品型号，确定所述缺陷所在位置相邻部件对应的图案的设计尺寸；
缺陷等级确定单元603，用于根据所述缺陷所属的不良类型，以及所述缺 陷尺寸与所述设计尺寸的大小关系，确定所述缺陷的缺陷等级；其中，所述缺陷等级为缺陷对产品良率的影响程度。
一种可能的实施方式，所述缺陷尺寸确定单元用于：
将所述缺陷在第一方向和第二方向中的最大长度，确定为所述缺陷尺寸；其中，所述第一方向垂直于所述第二方向。
一种可能的实施方式，所述缺陷等级确定单元603用于：
通过缺陷自动化分类系统中的神经网络模型判断出所述缺陷所属的不良类型，并确定对所述缺陷使用的缺陷等级判定条件；
用所述判定条件，对所述缺陷尺寸与所述设计尺寸的比值和设定倍率的大小关系进行判定，获得判定结果；
根据所述判定结果确定所述等级缺陷。
一种可能的实施方式，所述不良图片为显示面板中的不良图片时，所述不良类型包括颗粒不良、钝化层断裂不良。
一种可能的实施方式，所述缺陷等级确定单元603还用于：
当所述不良类型为颗粒不良时，确定对所述不良图片使用颗粒不良判定条件，所述设定倍率的取值范围为1.3～1.7；
当所述不良类型为钝化层断裂不良时，确定对所述不良图片使用钝化层断裂不良判定条件，所述设定倍率的取值范围为0.8～1.2。
一种可能的实施方式，所述颗粒不良判定条件，包括：
当所述缺陷所在位置周围的图案为薄膜晶体管的沟道对应的图案时，若所述缺陷尺寸与所述沟道的设计尺寸的比值小于或等于所述设定倍率，则确定所述缺陷等级为不良有品质风险；若所述缺陷尺寸与所述沟道的设计尺寸的比值大于所述设定倍率，则确定所述缺陷等级为不良无品质风险；
当所述缺陷所在位置周围的图案为所述显示面板的扫描线和数据线的交叉位置对应的图案时，若所述缺陷为黑色，且所述缺陷尺寸与所述交叉位置的设计尺寸的比值小于或等于所述设定倍率，则确定所述缺陷等级为不良有品质风险；若所述缺陷尺寸与所述交叉位置的设计尺寸的比值大于所述设定 倍率，则确定所述缺陷等级为不良无品质风险。
一种可能的实施方式，所述钝化层断裂不良判定条件，包括：
当所述缺陷所在位置周围的图案为所述显示面板中数据线对应的图案时，若所述缺陷尺寸与相邻两个数据线之间的设计间距的比值小于或等于所述设定倍率，则确定所述缺陷等级为不良可维修；
若所述缺陷尺寸与所述相邻两个数据线之间的设计间距的比值大于所述设定倍率，则确定所述缺陷等级为不良不可维修。
一种可能的实施方式，所述缺陷等级确定单元603还用于：
对不同缺陷等级用不同的标记对所述缺陷进行标注。
基于同一发明构思，本公开实施例中提供了一种不良图片缺陷等级自动识别的装置，包括：至少一个处理器，以及
与所述至少一个处理器连接的存储器；
其中，所述存储器存储有可被所述至少一个处理器执行的指令，所述至少一个处理器通过执行所述存储器存储的指令，执行如上所述的不良图片缺陷等级识别方法。
基于同一发明构思，本公开实施例还提一种可读存储介质，包括：
存储器，
所述存储器用于存储指令，当所述指令被处理器执行时，使得包括所述可读存储介质的装置完成如上所述的不良图片缺陷等级识别方法。
本领域内的技术人员应明白，本公开实施例可提供为方法、系统、或计算机程序产品。因此，本公开实施例可采用完全硬件实施例、完全软件实施例、或结合软件和硬件方面的实施例的形式。而且，本公开实施例可采用在一个或多个其中包含有计算机可用程序代码的计算机可用存储介质(包括但不限于磁盘存储器、CD-ROM、光学存储器等)上实施的计算机程序产品的形式。
本公开实施例是参照根据本公开实施例的方法、设备(系统)、和计算机程序产品的流程图和/或方框图来描述的。应理解可由计算机程序指令实现 流程图和/或方框图中的每一流程和/或方框、以及流程图和/或方框图中的流程和/或方框的结合。可提供这些计算机程序指令到通用计算机、专用计算机、嵌入式处理机或其他可编程数据处理设备的处理器以产生一个机器，使得通过计算机或其他可编程数据处理设备的处理器执行的指令产生用于实现在流程图一个流程或多个流程和/或方框图一个方框或多个方框中指定的功能的装置。
这些计算机程序指令也可存储在能引导计算机或其他可编程数据处理设备以特定方式工作的计算机可读存储器中，使得存储在该计算机可读存储器中的指令产生包括指令装置的制造品，该指令装置实现在流程图一个流程或多个流程和/或方框图一个方框或多个方框中指定的功能。
这些计算机程序指令也可装载到计算机或其他可编程数据处理设备上，使得在计算机或其他可编程设备上执行一系列操作步骤以产生计算机实现的处理，从而在计算机或其他可编程设备上执行的指令提供用于实现在流程图一个流程或多个流程和/或方框图一个方框或多个方框中指定的功能的步骤。
显然，本领域的技术人员可以对本公开进行各种改动和变型而不脱离本公开的精神和范围。这样，倘若本公开的这些修改和变型属于本公开权利要求及其等同技术的范围之内，则本公开也意图包含这些改动和变型在内。

Claims (11)

  1. 一种不良图片缺陷等级识别的方法，其中，包括：
    从不良图片中确定出缺陷的缺陷尺寸；
    根据所述不良图片对应的产品型号，确定所述缺陷所在位置相邻部件对应的图案的设计尺寸；
    根据所述缺陷所属的不良类型，以及所述缺陷尺寸与所述设计尺寸大小关系，确定所述缺陷的缺陷等级；其中，所述缺陷等级为缺陷对产品良率的影响程度。
  2. 如权利要求1所述的方法，其中，从不良图片中确定出缺陷的缺陷尺寸，包括：
    将所述缺陷在第一方向和第二方向中的最大长度，确定为所述缺陷尺寸；其中，所述第一方向垂直于所述第二方向。
  3. 如权利要求1所述的方法，其中，根据所述缺陷所属的不良类型，以及所述缺陷尺寸与所述设计尺寸的大小关系，确定所述缺陷的缺陷等级，包括：
    通过缺陷自动化分类系统中的神经网络模型判断出所述缺陷所属的不良类型，并确定对所述缺陷使用的缺陷等级判定条件；
    用所述判定条件，对所述缺陷尺寸与所述设计尺寸的比值和设定倍率的大小关系进行判定，获得判定结果；
    根据所述判定结果确定所述等级缺陷。
  4. 如权利要求1-3任一项所述的方法，其特征在于，所述不良图片为显示面板中的不良图片时，所述不良类型包括颗粒不良、钝化层断裂不良。
  5. 如权利要求4所述的方法，其中，通过缺陷自动化分类系统中的神经网络模型判断出所述缺陷所属的所述不良类型，并确定对所述缺陷使用的缺陷等级判定条件，包括：
    当所述不良类型为颗粒不良时，确定对所述不良图片使用颗粒不良判定 条件，所述设定倍率的取值范围为1.3～1.7；
    当所述不良类型为钝化层断裂不良时，确定对所述不良图片使用钝化层断裂不良判定条件，所述设定倍率的取值范围为0.8～1.2。
  6. 如权利要求5所述的方法，其中，所述颗粒不良判定条件，包括：
    当所述缺陷所在位置周围的图案为薄膜晶体管的沟道对应的图案时，若所述缺陷尺寸与所述沟道的设计尺寸的比值小于或等于所述设定倍率，则确定所述缺陷等级为不良有品质风险；若所述缺陷尺寸与所述沟道的比值大于所述设定倍率，则确定所述缺陷等级为不良无品质风险；
    当所述缺陷所在位置周围的图案为所述显示面板的扫描线和数据线的交叉位置对应的图案时，若所述缺陷为黑色，且所述缺陷尺寸与所述交叉位置的设计尺寸的比值小于或等于所述设定倍率，则确定所述缺陷等级为不良有品质风险；若所述缺陷尺寸与所述交叉位置的设计尺寸的比值大于所述设定倍率，则确定所述缺陷等级为不良无品质风险。
  7. 如权利要求5所述的方法，其中，所述钝化层断裂不良判定条件，包括：
    当所述缺陷所在位置周围的图案为所述显示面板中数据线对应的图案时，若所述缺陷尺寸与相邻两个数据线之间的设计间距的比值小于或等于所述设定倍率，则确定所述缺陷等级为不良可维修；
    若所述缺陷尺寸与所述相邻两个数据线之间的设计间距的比值大于所述设定倍率，则确定所述缺陷等级为不良不可维修。
  8. 如权利要求1-3任一项所述的方法，其中，确定所述不良图片的缺陷等级后，还包括：
    对不同缺陷等级用不同的标记对所述缺陷进行标注。
  9. 一种不良图片缺陷等级识别的装置，其中，包括：
    缺陷尺寸确定单元，用于从不良图片中确定出缺陷的缺陷尺寸；
    设计尺寸确定单元，用于根据所述不良图片对应的产品型号，确定所述缺陷所在位置相邻部件对应的图案的设计尺寸；
    缺陷等级确定单元，用于根据所述缺陷所属的不良类型，以及所述缺陷尺寸与所述设计尺寸的大小关系，确定所述缺陷的缺陷等级；其中，所述缺陷等级为缺陷对产品良率的影响程度。
  10. 一种不良图片缺陷等级自动识别的装置，其中，包括：
    至少一个处理器，以及
    与所述至少一个处理器连接的存储器；
    其中，所述存储器存储有可被所述至少一个处理器执行的指令，所述至少一个处理器通过执行所述存储器存储的指令，执行如权利要求1-8任一项所述的方法。
  11. 一种可读存储介质，其中，包括存储器，
    所述存储器用于存储指令，当所述指令被处理器执行时，使得包括所述可读存储介质的装置完成如权利要求1～8中任一项所述的方法。
PCT/CN2020/131918 2020-11-26 2020-11-26 不良图片缺陷等级识别的方法、装置及存储介质 WO2022109949A1 (zh)

Priority Applications (3)

Application Number Priority Date Filing Date Title
US18/038,197 US20230419466A1 (en) 2020-11-26 2020-11-26 Method and apparatus for identifying defect grade of bad picture, and storage medium
PCT/CN2020/131918 WO2022109949A1 (zh) 2020-11-26 2020-11-26 不良图片缺陷等级识别的方法、装置及存储介质
CN202080003018.0A CN115023603A (zh) 2020-11-26 2020-11-26 不良图片缺陷等级识别的方法、装置及存储介质

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/CN2020/131918 WO2022109949A1 (zh) 2020-11-26 2020-11-26 不良图片缺陷等级识别的方法、装置及存储介质

Publications (1)

Publication Number Publication Date
WO2022109949A1 true WO2022109949A1 (zh) 2022-06-02

Family

ID=81755083

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2020/131918 WO2022109949A1 (zh) 2020-11-26 2020-11-26 不良图片缺陷等级识别的方法、装置及存储介质

Country Status (3)

Country Link
US (1) US20230419466A1 (zh)
CN (1) CN115023603A (zh)
WO (1) WO2022109949A1 (zh)

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107941812A (zh) * 2017-12-20 2018-04-20 联想(北京)有限公司 信息处理方法及电子设备
US20190086340A1 (en) * 2017-09-18 2019-03-21 Elite Semiconductor Inc. Smart defect calibration system and the method thereof
CN109741324A (zh) * 2019-01-10 2019-05-10 惠科股份有限公司 一种检测方法、检测装置及终端设备
CN110690133A (zh) * 2019-09-05 2020-01-14 长江存储科技有限责任公司 半导体结构的检测方法及其检测装置

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20190086340A1 (en) * 2017-09-18 2019-03-21 Elite Semiconductor Inc. Smart defect calibration system and the method thereof
CN107941812A (zh) * 2017-12-20 2018-04-20 联想(北京)有限公司 信息处理方法及电子设备
CN109741324A (zh) * 2019-01-10 2019-05-10 惠科股份有限公司 一种检测方法、检测装置及终端设备
CN110690133A (zh) * 2019-09-05 2020-01-14 长江存储科技有限责任公司 半导体结构的检测方法及其检测装置

Also Published As

Publication number Publication date
CN115023603A (zh) 2022-09-06
US20230419466A1 (en) 2023-12-28

Similar Documents

Publication Publication Date Title
CN109683360B (zh) 液晶面板缺陷检测方法及装置
WO2021143343A1 (zh) 一种产品质量检测方法及装置
WO2017215241A1 (zh) 电路板元件漏件检测方法和系统
TWI525317B (zh) 整合影像分析與資料挖礦之自動光學檢測缺陷影像分類方法
CN111709948B (zh) 容器瑕疵检测方法和装置
TWI798627B (zh) 探測半導體晶片中缺陷的方法及半導體晶片缺陷探測系統
KR101982347B1 (ko) 오토 리페어 시스템의 불량 검출장치 및 방법
CN109616426A (zh) 智能型的缺陷校正系统与其实施方法
US20220076404A1 (en) Defect management apparatus, method and non-transitory computer readable medium
CN111199543A (zh) 一种基于卷积神经网络的冰柜表面缺陷检测
CN115661160B (zh) 一种面板缺陷检测方法及系统及装置及介质
CN115880520A (zh) 一种基于模板匹配和缺陷分割的缺陷等级分类方法及系统
CN108445020A (zh) 一种玻璃基板缺陷聚集识别方法
US20150324664A1 (en) Automatic calibration sample selection for die-to-database photomask inspection
CN115063357A (zh) 表面缺陷的检测方法、系统、电子设备及存储介质
TWI707299B (zh) 光學檢測二次圖像分類方法
CN103245667A (zh) 自动检测机械性划痕的方法及系统
CN116385380A (zh) 基于深度特征的缺陷检测方法、系统、设备及存储介质
CN115937175A (zh) 一种电路板表面缺陷检测方法、系统、设备及存储介质
TW202401294A (zh) Pcb缺陷檢測系統及pcb缺陷檢測方法
CN117392042A (zh) 缺陷检测方法、缺陷检测设备及存储介质
CN117372424B (zh) 一种缺陷检测方法、装置、设备及存储介质
WO2014103617A1 (ja) 位置合せ装置、欠陥検査装置、位置合せ方法、及び制御プログラム
WO2022109949A1 (zh) 不良图片缺陷等级识别的方法、装置及存储介质
CN114299040A (zh) 瓷砖瑕疵检测方法、装置和电子设备

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 20962840

Country of ref document: EP

Kind code of ref document: A1

WWE Wipo information: entry into national phase

Ref document number: 18038197

Country of ref document: US

NENP Non-entry into the national phase

Ref country code: DE

32PN Ep: public notification in the ep bulletin as address of the adressee cannot be established

Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205A DATED 21.09.2023)

122 Ep: pct application non-entry in european phase

Ref document number: 20962840

Country of ref document: EP

Kind code of ref document: A1