CN114627089A - Defect identification method, defect identification device, computer equipment and computer readable storage medium - Google Patents
Defect identification method, defect identification device, computer equipment and computer readable storage medium Download PDFInfo
- Publication number
- CN114627089A CN114627089A CN202210281601.9A CN202210281601A CN114627089A CN 114627089 A CN114627089 A CN 114627089A CN 202210281601 A CN202210281601 A CN 202210281601A CN 114627089 A CN114627089 A CN 114627089A
- Authority
- CN
- China
- Prior art keywords
- defect
- type
- defects
- confidence
- preset
- Prior art date
- Legal status (The legal status 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 status listed.)
- Pending
Links
- 230000007547 defect Effects 0.000 title claims abstract description 587
- 238000000034 method Methods 0.000 title claims abstract description 51
- 238000003860 storage Methods 0.000 title claims abstract description 13
- 238000001514 detection method Methods 0.000 claims abstract description 6
- 238000004590 computer program Methods 0.000 claims description 9
- 238000012549 training Methods 0.000 claims description 7
- 238000013528 artificial neural network Methods 0.000 claims description 5
- 238000004364 calculation method Methods 0.000 description 8
- 230000002950 deficient Effects 0.000 description 8
- 230000008569 process Effects 0.000 description 8
- 238000004519 manufacturing process Methods 0.000 description 6
- 238000012545 processing Methods 0.000 description 6
- 238000013135 deep learning Methods 0.000 description 4
- 239000000126 substance Substances 0.000 description 4
- 238000004891 communication Methods 0.000 description 3
- 230000008878 coupling Effects 0.000 description 3
- 238000010168 coupling process Methods 0.000 description 3
- 238000005859 coupling reaction Methods 0.000 description 3
- 230000006870 function Effects 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000003062 neural network model Methods 0.000 description 2
- 238000006467 substitution reaction Methods 0.000 description 2
- 238000012935 Averaging Methods 0.000 description 1
- 238000004422 calculation algorithm Methods 0.000 description 1
- 238000010219 correlation analysis Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000001788 irregular Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000000737 periodic effect Effects 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 238000002360 preparation method Methods 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
- 238000007619 statistical method Methods 0.000 description 1
- 239000002699 waste material Substances 0.000 description 1
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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- 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
- G06T2207/20081—Training; Learning
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Physics & Mathematics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Quality & Reliability (AREA)
- Electroluminescent Light Sources (AREA)
Abstract
The application provides a defect identification method, a defect identification device, computer equipment and a computer readable storage medium, and relates to the technical field of defect detection. The method comprises the steps that an initial picture of an OLED panel to be detected is obtained, and the initial picture is input into a defect recognition model trained in advance to obtain a primary recognition result, wherein the primary recognition result comprises the type of a defect and the confidence coefficient of the defect; judging whether all defects with confidence degrees larger than confidence degree threshold values preset by corresponding defect types have at least two first defects of preset types or not; if at least two first defects exist, calculating the similarity between each first defect and the corresponding defect template, and acquiring the target defect type of the first defect, wherein the target defect type of the first defect is the defect type corresponding to the defect template with the maximum similarity to the first defect, and the target defect type is used as the output result of the initial picture. By the defect identification method provided by the embodiment, the accuracy of judging the image can be improved.
Description
Technical Field
The invention relates to the technical field of defect detection, in particular to a defect identification method, a defect identification device, computer equipment and a computer readable storage medium.
Background
Organic Light-Emitting diodes (OLEDs) have a number of excellent characteristics and are widely considered to be an ideal next-generation flat panel display technology. However, due to the complicated production process, various defects are inevitably generated in the preparation process. These defects have the characteristics of fuzzy boundaries, irregular shapes, periodic texture backgrounds, uneven overall brightness and the like. Therefore, the surface defects of the OLED display screen are detected, statistical analysis on the defects is facilitated, products are repaired or eliminated, the manufacturing quality of the products is improved through process improvement, and waste of production resources is avoided.
At present, a camera is generally used for photographing for detecting the OLED display screen, and manual graph judgment is carried out, but the manual graph judgment cost is high, and the manual graph judgment is influenced by subjective factors of graph judgment personnel and the proficiency of the graph judgment. Therefore, the deep learning technique has become a mainstream method for recognizing multiple defects. However, the defect forms generated in the production environment are different and complex, and the simple discrimination model has the problems of low recognition precision and accuracy.
Disclosure of Invention
In order to solve the above technical problem, embodiments of the present application provide a defect identification method, apparatus, computer device, and computer-readable storage medium.
In a first aspect, an embodiment of the present application provides a defect identification method, where the method includes:
acquiring an initial picture of an OLED panel to be detected, and inputting the initial picture into a defect recognition model trained in advance to obtain a primary recognition result, wherein the primary recognition result comprises the type of a defect and the confidence coefficient of the defect;
judging whether the confidence coefficient of each defect is greater than a preset confidence coefficient threshold value corresponding to the defect type;
if the confidence degrees of at least two defects are larger than the confidence degree threshold value preset by the corresponding defect type, judging whether at least two first defects of the preset type exist in all the defects of which the confidence degrees are larger than the confidence degree threshold value preset by the corresponding defect type;
if at least two first defects exist, the target defect type of each first defect is obtained by calculating the similarity between each first defect and the corresponding defect template, wherein the target defect type of each first defect is the defect type corresponding to the defect template with the maximum similarity to the first defect, and the target defect type is used as the output result of the initial picture.
In a specific embodiment, after the step of determining whether there are at least two first defects of a preset type in all the defects whose confidence levels are greater than the confidence level threshold preset for the corresponding defect type if there are at least two defects whose confidence levels are greater than the confidence level threshold preset for the corresponding defect type, the method further includes:
if at least two first defects exist, judging whether second defects exist in all the defects with confidence degrees larger than confidence degree threshold values preset by corresponding defect types or not;
and if the second defects exist, performing priority judgment on the target defect type and the defect types of the second defects, and taking the defect type corresponding to the defect with the highest priority as an output result of the initial picture, wherein the first defect is a defect with a confidence coefficient larger than a first preset multiple of the maximum confidence coefficient, and the second defect is a defect with a confidence coefficient larger than a second preset multiple of the maximum confidence coefficient.
In a specific embodiment, the maximum confidence is the maximum of the confidences of the defects in the preliminary identification result.
In a specific embodiment, the step of obtaining the target defect type of each first defect by calculating a similarity between each first defect and a corresponding defect template includes:
acquiring the coordinates of the center point of the first defect and the edge values of the pixel area of the first defect to form a defect sequence corresponding to the first defect;
respectively calculating gray correlation degrees of the defect sequence of the first defect and each defect template;
and determining the defect type corresponding to the defect template corresponding to the maximum value in the gray relevance degrees of the corresponding defect templates as the target defect type.
In a specific embodiment, the defect type of the defect template comprises at least one of a foreign matter type on the OLED panel film, a foreign matter type under the OLED panel film and a smudge type.
In a specific embodiment, after the step of determining whether the confidence of each defect is greater than the confidence threshold preset for the corresponding defect type, the method further includes:
if the confidence coefficient of one defect is larger than the preset confidence coefficient threshold value of the corresponding defect type, judging that the initial picture has single defect, and taking the defect type corresponding to the single defect as the output result of the initial picture.
In a specific embodiment, the defect identification model is obtained by:
collecting historical defect pictures of the OLED panel;
marking the defect characteristics in the historical defect picture to obtain a defect sample picture and defect information corresponding to the defect sample picture;
and inputting the defect sample picture and the defect information into a basic neural network, and training to obtain the defect identification model.
In a second aspect, an embodiment of the present application provides a defect identification apparatus, including:
the device comprises an acquisition module, a detection module and a display module, wherein the acquisition module is used for acquiring an initial picture of the OLED panel to be detected and inputting the initial picture into a defect recognition model trained in advance to obtain a primary recognition result, and the primary recognition result comprises the type of the defect and the confidence coefficient of the defect;
the first judgment module is used for judging whether the confidence coefficient of each defect is greater than a preset confidence coefficient threshold corresponding to the defect type;
the second judging module is used for judging whether at least two first defects of preset types exist in all the defects with the confidence degrees larger than the confidence degree threshold value preset by the corresponding defect type or not if the confidence degrees of at least two defects are larger than the confidence degree threshold value preset by the corresponding defect type;
and the determining module is used for obtaining a target defect type of the first defect by calculating the similarity between each first defect and the corresponding defect template if at least two first defects exist, wherein the target defect type of the first defect is the defect type corresponding to the defect template with the maximum similarity to the first defect, and the target defect type is used as the output result of the initial picture.
In a third aspect, an embodiment of the present application provides a computer device, where the computer device includes a memory and a processor, the memory stores a computer program, and the computer program executes the defect identification method according to the first aspect when the processor runs.
In a fourth aspect, the present application provides a computer-readable storage medium, on which a computer program is stored, where the computer program, when executed by a processor, implements the defect identification method of the first aspect.
According to the defect identification method provided by the embodiment of the application, the pre-trained defect identification model is used for identifying the OLED panel picture to be detected, the obtained primary identification result is screened by adopting the confidence threshold value, and then subsequent similarity judgment is carried out, so that the precision and accuracy of the defect identification of the OLED panel are improved.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings needed to be used in the embodiments will be briefly described below, it should be understood that the following drawings only illustrate some embodiments of the present invention and therefore should not be considered as limiting the scope, and for those skilled in the art, other related drawings can be obtained according to the drawings without inventive efforts.
FIG. 1 is a flow chart of a defect identification method according to an embodiment of the present invention;
fig. 2 shows a schematic structural diagram of a defect recognition apparatus provided in an embodiment of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments.
The components of embodiments of the present invention generally described and illustrated in the figures herein may be arranged and designed in a wide variety of different configurations. Thus, the following detailed description of the embodiments of the present invention, presented in the figures, is not intended to limit the scope of the invention, as claimed, but is merely representative of selected embodiments of the invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments of the present invention without making any creative effort, shall fall within the protection scope of the present invention.
Hereinafter, the terms "including", "having", and their derivatives, which may be used in various embodiments of the present invention, are only intended to indicate specific features, numbers, steps, operations, elements, components, or combinations of the foregoing, and should not be construed as first excluding the existence of, or adding to, one or more other features, numbers, steps, operations, elements, components, or combinations of the foregoing.
Furthermore, the terms "first," "second," "third," and the like are used solely to distinguish one from another and are not to be construed as indicating or implying relative importance.
Unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by one of ordinary skill in the art to which various embodiments of the present invention belong. The terms (such as those defined in commonly used dictionaries) should be interpreted as having a meaning that is consistent with their contextual meaning in the relevant art and will not be interpreted in an idealized or overly formal sense unless expressly so defined herein in various embodiments of the present invention.
Example 1
Referring to fig. 1, fig. 1 is a schematic flow chart of a defect identification method provided in this embodiment. As shown in fig. 1, the method includes:
step S101, an initial picture of an OLED panel to be detected is obtained, and the initial picture is input into a defect recognition model trained in advance to obtain a primary recognition result, wherein the primary recognition result comprises the type of defects and the confidence coefficient of the defects.
In specific implementation, as the process flow is too many and complicated in the production process of the OLED panel, various product defects such as color mixing, color shortage, dirt and the like are easily generated, and the defects may affect the normal use of the OLED panel and need to be detected before the OLED panel leaves a factory. At present, the deep learning technology is mostly adopted to carry out defect identification work on the OLED panel, but the defect forms generated are different, complex and changeable due to different production environments of the OLED panel, and the identification precision and accuracy of the conventional deep learning identification model are lower.
The embodiment provides a method combining deep learning with subsequent processing to identify defects of an OLED panel, that is, for an OLED panel to be detected, a corresponding picture is collected, defect detection is performed by using the collected picture, and the picture to be detected can be defined as an initial picture. The initial picture may be obtained by directly shooting and collecting the OLED panel to be detected, or by obtaining pictures of the OLED panel to be detected from other terminals through a network.
Specifically, a neural network with target detection and positioning functions is trained in advance, a defect identification model is defined, the collected initial pictures to be detected are input into the neural network model, a defect information table corresponding to each initial picture can be obtained, a preliminary identification result is defined, and the defect information table comprises information such as types, confidence degrees, positions and the like of estimated defects. The estimated defects are all estimated results of the defects existing on the initial picture by the defect identification model, each estimated result corresponds to one defect type, for example, only one defect exists on the initial picture, but the defect identification model may give a plurality of estimated results after identifying the initial picture, that is, a defect information table obtained after identifying includes a plurality of defect information.
Step S102, determining whether the confidence of each defect is greater than a confidence threshold preset for the corresponding defect type.
In specific implementation, before each defect in the defect information table is processed, a confidence threshold needs to be set for each type of defect that may exist on the OLED panel in advance, the confidence of the defect is compared with the preset confidence threshold of the defect, and the processing mode of the defect is determined according to the size relationship.
Specifically, the confidence threshold may be set based on the historical data of the defect information table, that is, the historical data of the defect information table is collected, the historical data is subjected to certain preprocessing, and the preprocessed data is summarized and analyzed, so that on the basis, a suitable confidence threshold is set for each type of defect that may exist on the OLED panel.
Specifically, in the defect information table, the defect with the confidence coefficient smaller than the confidence coefficient threshold value is assigned to the other, and subsequent processing is not performed any more, if the assignment of each defect in the defect information table is the other, that is, the confidence coefficient of each defect is smaller than the confidence coefficient threshold value, the other is output as the recognition result of the initial picture, in addition, the defect with the confidence coefficient larger than the confidence coefficient threshold value continues to perform subsequent processing operation, the defect needing to be subjected to subsequent operation can be directly screened out through the clamping control of the confidence coefficient threshold value, and the accuracy of defect recognition is improved to a certain extent.
Step S103, if the confidence degrees of at least two defects are larger than the confidence degree threshold value preset by the corresponding defect type, judging whether at least two first defects of the preset type exist in all the defects of which the confidence degrees are larger than the confidence degree threshold value preset by the corresponding defect type;
in a specific embodiment, the maximum confidence is the maximum of the confidences of the defects in the preliminary identification result.
In specific implementation, defects needing to be subjected to subsequent treatment are screened out according to confidence threshold values, the confidence degrees of the defects need to be larger than the confidence threshold values preset by the defects, among all the defects meeting the conditions, the defects with the confidence degrees larger than a times of the maximum confidence degree are defined as first defects, the value of a can be 0.8, the value of a can be adjusted according to actual conditions, and the maximum confidence degree refers to the maximum value of the confidence degrees of the defects in the defect information table.
Specifically, when two or more first defects exist, the number of the defects of which the defect types are the preset types in all the first defects needs to be determined, the preset types include the foreign matter on the OLED film, the foreign matter under the OLED film, and the dirt, because the three types of defects are very similar, the defect identification model is particularly easy to be confused when identifying the three types of defects, for example, the defect of the dirt exists on the initial picture, and a defect information table obtained after the defect identification model identifies may include two or three types of the three types of defects of the foreign matter on the film, the foreign matter under the film, and the dirt, so that the specific defect types cannot be accurately determined.
Specifically, in this embodiment, a subsequent processing method is provided for the three defects, that is, when at least two first defects exist in the recognition result and the types of the first defects include at least two defect types of the foreign matter on the film, the foreign matter under the film, and the dirt, the subsequent determination needs to be performed on the first defects including the at least two defect types, so as to improve the recognition accuracy of the three defects.
Step S104, if at least two first defects exist, acquiring a target defect type of the first defects by calculating the similarity between each first defect and a corresponding defect template, wherein the target defect type of the first defects is the defect type corresponding to the defect template with the maximum similarity to the first defects, and the target defect type is used as an output result of the initial picture;
in a specific embodiment, the step of obtaining the target defect type of the first defect by calculating a similarity between each first defect and the corresponding defect template includes:
acquiring the coordinates of the center point of the first defect and the edge values of the pixel area of the first defect to form a defect sequence corresponding to the first defect;
respectively calculating gray correlation degrees of the defect sequence of the first defect and each defect template;
determining the defect type corresponding to the defect template corresponding to the maximum value in the gray correlation degrees of the defect templates as the target defect type;
in a specific embodiment, the defect type of the defect template comprises at least one of a foreign matter type on the OLED panel film, a foreign matter type under the OLED panel film and a smudge type.
In specific implementation, because the three defects of the foreign matter on the OLED panel film, the foreign matter under the OLED panel film and the dirt are easy to be confused, and the defect identification model is difficult to accurately identify, the embodiment provides a similarity judgment mode to distinguish the three defects of the foreign matter on the OLED panel film, the foreign matter under the OLED film and the dirt, that is, by respectively calculating the gray correlation degrees of the defect sequences of the three defects existing on the initial picture and the defect templates corresponding to the three defects, the defect type corresponding to the maximum gray correlation degree is defined as the target defect type, and the target defect type is defined as the defect identification result of the initial picture, and the defect identification result is defined as the output result.
Specifically, the defect sequence is composed of a central point coordinate of the defect and an edge value of the pixel region of the defect, the central point coordinate of the defect is the central point coordinate of the pixel region of the defect, the edge value of the pixel region of the defect is the maximum value of the abscissa and the ordinate of the minimum circumscribed rectangular frame of the pixel region of the defect, and the central point coordinate of the pixel region of the defect is calculated according to the maximum value of the abscissa and the ordinate of the minimum circumscribed rectangular frame of the pixel region of the defect.
Specifically, an average value is calculated according to the maximum value and the minimum value of the abscissa of the minimum circumscribed rectangular frame of the defective pixel region to obtain the abscissa of the center point of the defective pixel region, an average value is calculated according to the maximum value and the minimum value of the ordinate of the minimum circumscribed rectangular frame of the defective pixel region to obtain the ordinate of the center point of the defective pixel region, and the maximum value of the ordinate of the minimum circumscribed rectangular frame of the defective pixel region can be obtained according to the position information of each defect in the defect information table.
Specifically, according to the above calculation process, the defect sequence [ o ] corresponding to the three defects existing in the initial picture is obtainedxi,oyi,xmaxi,xmini,ymaxi,ymini]Wherein [ o ]xi,oyi]Is the center point coordinate of the defective pixel region, [ xmaxi,xmini,ymaxi,ymini]I represents any one of foreign matter on the OLED panel film, foreign matter under the film, and dirt.
Specifically, when a defect identification model is trained, the sample picture needs to be marked, characteristics and types of various sample pictures can be conveniently induced and learned, data of an external rectangular frame of a pixel region corresponding to each defect can be obtained in the process of marking the sample picture, data of the external rectangular frame of the pixel region corresponding to the three defects, namely foreign matters on an OLED panel film, foreign matters under the film and dirt, are selected, the coordinate maximum value of each external rectangular frame is obtained, the coordinates of the central point of the pixel region corresponding to the three defects are respectively calculated according to the coordinate maximum value and the coordinate minimum value of each external rectangular frame, namely the abscissa of the central point of the pixel region corresponding to the defect is obtained according to the calculated average value of the maximum value and the minimum value of the abscissa of the external rectangular frame, and the ordinate of the central point of the pixel region corresponding to the defect is obtained according to the calculated average value of the maximum value and the minimum value of the ordinate of the external rectangular frame.
Specifically, the defect sequences respectively corresponding to the three defects are formed according to the coordinates of the center point of the pixel region of the three defects and the maximum values of the abscissa and the ordinate of the circumscribed rectangular frame of the pixel region, because the three defects may appear many times in the historical defect picture, the defect sequences corresponding to the three defects may have a plurality of numbers, the average value of the defect sequences corresponding to each of the three defects is calculated respectively to obtain the average defect sequence corresponding to each of the three defects, the average defect sequence may be defined as a defect template, taking the calculation of the defect template of the dirty defect as an example, the defect sequences corresponding to the dirty defect are calculated firstly, then the average value of the maximum values of the abscissa of all the circumscribed rectangular frames in the defect sequences is calculated to obtain the maximum value of the abscissa of the circumscribed rectangular frame in the average sequence, the calculation of the remaining values in the average defect sequence is consistent with the calculation mode of the maximum value of the abscissa of the circumscribed rectangular frame, and the defect template with the smudged defect is obtained through the calculation process.
Specifically, if the types of defects existing on the initial picture include at least two types of defects among a foreign substance on the OLED panel film, a foreign substance under the film, and dirt, and the confidence degrees of the defects corresponding to the at least two types of defects are both greater than a times of the maximum confidence degree, for example, if two types of defects including a foreign substance on the film and a foreign substance under the film exist on the initial picture, and the confidence degrees of the two types of defects are both greater than a times of the maximum confidence degree, similarity judgment needs to be performed on the two types of defects, and the similarity judgment step includes: firstly, determining the defect sequences of the two defects, then respectively calculating the gray correlation degrees of the defect sequences of the two defects and the corresponding defect templates, and according to the calculation result, if the gray correlation degree corresponding to the foreign matter on the film is greater than the gray correlation degree corresponding to the foreign matter under the film, the result of the similarity judgment is the foreign matter on the film, and at the moment, the foreign matter on the film is taken as the defect identification result of the initial picture, and the calculation process of the gray correlation degree comprises the following steps: calculating gray correlation coefficients of the defect sequence and the corresponding defect template, and averaging the gray correlation coefficients to obtain a gray correlation degree, wherein a specific calculation formula is as follows:
in the above formula, γ (X)0,Xi) Indicates a defect sequence XiCorresponding defect template X0Gray correlation of (a), gamma (x)0(k),xi(k) ) represents a defect sequence XiCorresponding defect template X0The k-th gray correlation coefficient of (c),indicates to the defect sequence XiCorresponding defect template X0Is averaged over the individual gray correlation coefficients of (1), where xi(k) And x0(k) Respectively represent defect sequences XiAnd corresponding defect template X0The kth value in (k) is 1,2, …,6, n represents the number of gray correlation coefficients, n is 6, and ρ ∈ (0,1) is a resolution coefficient.
xi(k)dtCan represent a defect sequence XiThe k-th value of the initial value image can also represent the defect sequence XiAny one of the mean image, the inverted image and the inverted image of the kth value can be selected according to actual needs, and x in this embodiment isi(k)dtIndicates a defect sequence XiThe initial value of the k-th value is like the example, as described above, x in the present embodiment0(k)dtTemplate X for indicating defects0The k-th value is the initial value of t, which corresponds to 1,2,3,4 in turn to the defect sequence XiAnd defect template X0The initial value image, the mean value image, the inverted image and the inverted image of each value.
When t is 1, | x0(k)dt-xi(k)dtI denotes the Defect template X0And defect sequence XiThe absolute value of the difference between the initial images of the kth value of (a);indicating to defective template X0And defect sequence XiTraversing the absolute value of the difference between the initial value images of each corresponding value in the image data, and selecting the minimum value;indicating to defective template X0And defect sequence XiAnd traversing the absolute value of the difference between the initial value images of each corresponding value, and selecting the maximum value.
Because the data in each defect sequence and defect template may not be convenient for comparison due to different dimensions or it is difficult to obtain a correct conclusion during comparison, in order to ensure the reliability of the result, when performing gray correlation analysis, the data is first subjected to dimensionless processing. According to the actual situation, any one of the methods is selected to process the data.
In a specific embodiment, after the step of determining whether there are at least two first defects of a preset type in all the defects whose confidences are greater than the confidence threshold preset for the corresponding defect type if there are at least two defects whose confidences are greater than the confidence threshold preset for the corresponding defect type, the method further includes:
if at least two first defects exist, judging whether second defects exist in all the defects with confidence degrees larger than confidence degree threshold values preset by corresponding defect types or not;
and if the second defects exist, performing priority judgment on the target defect type and the defect types of the second defects, and taking the defect type corresponding to the defect with the highest priority as an output result of the initial picture, wherein the first defect is a defect with a confidence coefficient larger than a first preset multiple of the maximum confidence coefficient, and the second defect is a defect with a confidence coefficient larger than a second preset multiple of the maximum confidence coefficient.
In specific implementation, a defect with a confidence greater than b times the maximum confidence can be defined as a second defect, the value of b is 0.8, the value of b can be correspondingly adjusted according to the identification result, the priority judgment of the defect refers to sorting the defects according to the priority, the defect with the maximum priority is used as the defect identification result of the initial picture, the priority sorting can be based on the influence degree of the defects, the defect with the larger influence degree has higher priority, and certainly, the defect can be sorted according to other factors.
Specifically, after the similarity is judged, whether a second defect exists on the initial picture needs to be judged, if the second defect exists, priority judgment is performed on each second defect and the target defect type existing on the initial picture, and the defect with the highest priority is used as a defect identification result of the initial picture.
Specifically, when the similarity is not determined, if at least two second defects exist on the initial picture, priority determination needs to be performed on each second defect, and the defect with the highest priority is used as the defect identification result of the initial picture.
In a specific embodiment, after the step of determining whether the confidence of each defect is greater than the confidence threshold preset for the corresponding defect type, the method further includes:
if the confidence coefficient of one defect is larger than the preset confidence coefficient threshold value of the corresponding defect type, judging that the initial picture has single defect, and taking the defect type corresponding to the single defect as the output result of the initial picture.
In specific implementation, after the initial picture is detected by the defect identification model, if the confidence of only one defect is greater than the confidence threshold of the defect, it is indicated that only one defect exists in the initial picture, and the defect is directly used as the defect identification result of the initial picture.
In a specific embodiment, the defect identification model is obtained by:
collecting historical defect pictures of the OLED panel;
marking the defect characteristics in the historical defect picture to obtain a defect sample picture and defect information corresponding to the defect sample picture;
and inputting the defect sample picture and the defect information into a basic neural network, and training to obtain the defect identification model.
In specific implementation, a defect data set for training needs to be prepared before training a defect identification model, the defect data set includes historical pictures of various defects existing in the OLED panel, the pictures can be defined as historical defect pictures, and the historical defect pictures can be obtained by directly shooting and collecting the OLED panel including various defects or by obtaining pictures of the OLED panel including various defects from other terminals acquired through a network.
Specifically, a collected historical defect picture of the OLED panel is cut into a defect sample picture with a proper size, the defect sample picture is classified according to the defect type, a Labellmg marking tool is used for marking the cut defect sample picture, the position and the defect type of the defect are marked, the marking tool can generate an XML file which is marked with the same name as the defect sample picture, each XML file is converted to finally obtain a json file, the json file and each defect sample picture are input into a neural network model together as a data set, and a Faster RCNN algorithm is used for training to obtain a defect identification model.
According to the defect identification method provided by the embodiment, the pre-trained defect identification model is used for identifying the OLED panel picture to be detected, the obtained primary identification result is screened by adopting the confidence threshold, and then subsequent similarity judgment and priority judgment are carried out, so that the precision and accuracy of the defect identification of the OLED panel are improved.
Example 2
Referring to fig. 2, the present embodiment further provides a defect identifying apparatus 200, including:
an obtaining module 201, configured to obtain an initial picture of an OLED panel to be detected, and input the initial picture into a defect recognition model trained in advance to obtain a preliminary recognition result, where the preliminary recognition result includes actual confidence levels of defects in the initial picture;
a first determining module 202, configured to determine whether an actual confidence of each defect is within a confidence threshold range corresponding to a preset plurality of types of defects;
a second determining module 203, configured to determine whether there are at least two first defects of a preset type in all the defects whose confidence degrees are greater than the confidence degree threshold preset for the corresponding defect type if there are at least two defects whose confidence degrees are greater than the confidence degree threshold preset for the corresponding defect type;
the determining module 204 is configured to, if there are at least two first defects, obtain a target defect type of the first defect by calculating a similarity between each first defect and a corresponding defect template, where the target defect type of the first defect is a defect type corresponding to the defect template with the largest similarity to the first defect, and use the target defect type as an output result of the initial picture.
In an embodiment, the obtaining module 201 is specifically configured to: collecting historical defect pictures of the OLED panel;
marking the defect characteristics in the historical defect picture to obtain a defect sample picture and defect information corresponding to the defect sample picture;
and inputting the defect sample picture and the defect information into a basic neural network, and training to obtain the defect identification model.
In an embodiment, the first determining module 202 is specifically configured to: if the confidence coefficient of one defect is larger than the preset confidence coefficient threshold value of the corresponding defect type, judging that the initial picture has single defect, and taking the defect type corresponding to the single defect as the output result of the initial picture.
In an embodiment, the second determining module 203 is specifically configured to: the maximum confidence coefficient is the maximum value of the confidence coefficients of all the defects in the preliminary identification result.
In an embodiment, the determining module 204 is specifically configured to: if at least two first defects exist, judging whether second defects exist in all the defects with confidence degrees larger than confidence degree threshold values preset by corresponding defect types or not;
and if the second defects exist, performing priority judgment on the target defect type and the defect types of the second defects, and taking the defect type corresponding to the defect with the highest priority as an output result of the initial picture, wherein the first defect is a defect with a confidence coefficient larger than a first preset multiple of the maximum confidence coefficient, and the second defect is a defect with a confidence coefficient larger than a second preset multiple of the maximum confidence coefficient.
In an embodiment, the determining module 204 is further specifically configured to: acquiring the coordinates of the center point of the first defect and the edge values of the pixel area of the first defect to form a defect sequence corresponding to the first defect;
respectively calculating gray correlation degrees of the defect sequence of the first defect and each defect template;
and determining the defect type corresponding to the defect template corresponding to the maximum value in the gray correlation degrees of the defect templates as the target defect type.
In an embodiment, the determining module 204 is further specifically configured to: the defect type of the defect template comprises at least one of a foreign matter type on the OLED panel film, a foreign matter type under the OLED panel film and a dirt type.
For specific functions of the defect identification apparatus 200 provided in this embodiment, reference may be made to a specific implementation process of the defect identification method in embodiment 1, which is not described in detail herein.
The defect identification device provided by the embodiment can identify the OLED panel picture to be detected through the pre-trained defect identification model, screen the obtained primary identification result by adopting a confidence threshold value, and then perform subsequent similarity judgment and priority judgment, so that the precision and accuracy of the defect identification of the OLED panel are improved.
Example 3
The present embodiment provides a computer device, which includes a memory and a processor, wherein the memory stores a computer program, and the computer program executes the defect identification method according to embodiment 1 when the processor runs.
The computer device provided in this embodiment may implement the defect identification method described in embodiment 1, and is not described herein again to avoid repetition.
Example 4
The present embodiment provides a computer-readable storage medium having stored thereon a computer program which, when executed by a processor, implements the defect identifying method of embodiment 1.
The computer-readable storage medium provided in this embodiment may implement the defect identification method described in embodiment 1, and is not described herein again to avoid repetition.
In the embodiments provided in the present invention, it should be understood that the disclosed apparatus and method may be implemented in other ways. The above-described embodiments of the apparatus are merely illustrative, and for example, the division of the units is only one logical division, and there may be other divisions when actually implemented, and for example, a plurality of units or components may be combined or integrated into another system, or some features may be omitted, or not executed. In addition, the shown or discussed mutual coupling or direct coupling or communication connection may be an indirect coupling or communication connection of devices or units through some communication interfaces, and may be in an electrical, mechanical or other form.
The units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of the embodiment.
In addition, functional units in the embodiments provided by the present invention may be integrated into one processing unit, or each unit may exist alone physically, or two or more units are integrated into one unit.
The functions, if implemented in the form of software functional units and sold or used as a stand-alone product, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present invention may be embodied in the form of a software product, which is stored in a storage medium and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: various media capable of storing program codes, such as a usb disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk, or an optical disk.
It should be noted that: like reference numbers and letters refer to like items in the following figures, and thus once an item is defined in one figure, it need not be further defined and explained in subsequent figures, and moreover, the terms "first", "second", "third", etc. are used merely to distinguish one description from another and are not to be construed as indicating or implying relative importance.
Finally, it should be noted that: the above-mentioned embodiments are only specific embodiments of the present invention, which are used for illustrating the technical solutions of the present invention and not for limiting the same, and the protection scope of the present invention is not limited thereto, although the present invention is described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: those skilled in the art can still make modifications or changes to the embodiments described in the foregoing embodiments, or make equivalent substitutions for some features, within the scope of the disclosure; such modifications, changes or substitutions do not depart from the spirit and scope of the present invention in its spirit and scope. Are intended to be covered by the scope of the present invention.
Claims (10)
1. A defect identification method for identifying defects of an OLED panel, the method comprising:
acquiring an initial picture of an OLED panel to be detected, and inputting the initial picture into a defect recognition model trained in advance to obtain a primary recognition result, wherein the primary recognition result comprises the type of a defect and the confidence coefficient of the defect;
judging whether the confidence coefficient of each defect is greater than a preset confidence coefficient threshold value corresponding to the defect type;
if the confidence degrees of at least two defects are larger than the confidence degree threshold value preset by the corresponding defect type, judging whether at least two first defects of the preset type exist in all the defects of which the confidence degrees are larger than the confidence degree threshold value preset by the corresponding defect type;
if at least two first defects exist, the target defect type of each first defect is obtained by calculating the similarity between each first defect and the corresponding defect template, wherein the target defect type of each first defect is the defect type corresponding to the defect template with the maximum similarity to the first defect, and the target defect type is used as the output result of the initial picture.
2. The method of claim 1, wherein after the step of determining whether there are at least two first defects of a predetermined type in all the defects with the confidence levels greater than the confidence level threshold predetermined for the corresponding defect type if there are at least two defects with the confidence levels greater than the confidence level threshold predetermined for the corresponding defect type, the method further comprises:
if at least two first defects exist, judging whether second defects exist in all the defects with confidence degrees larger than confidence degree threshold values preset by corresponding defect types or not;
and if the second defects exist, performing priority judgment on the target defect type and the defect types of the second defects, and taking the defect type corresponding to the defect with the highest priority as an output result of the initial picture, wherein the first defect is a defect with a confidence coefficient larger than a first preset multiple of the maximum confidence coefficient, and the second defect is a defect with a confidence coefficient larger than a second preset multiple of the maximum confidence coefficient.
3. The defect identification method according to claim 2, wherein the maximum confidence is a maximum value among the confidence of the defects in the preliminary identification result.
4. The method of claim 1, wherein the step of obtaining the target defect type of the first defect by calculating the similarity between each first defect and the corresponding defect template comprises:
acquiring the coordinates of the center point of the first defect and the edge values of the pixel area of the first defect to form a defect sequence corresponding to the first defect;
respectively calculating gray correlation degrees of the defect sequence of the first defect and each defect template;
and determining the defect type corresponding to the defect template corresponding to the maximum value in the gray correlation degrees of the defect templates as the target defect type.
5. The defect identification method according to claim 1, wherein the defect type of the defect template comprises at least one of a foreign matter type on an OLED panel film, a foreign matter type under the OLED panel film and a smudge type.
6. The method of claim 1, wherein after the step of determining whether the confidence level of each defect is greater than the confidence level threshold preset for the corresponding defect type, the method further comprises:
if the confidence coefficient of one defect is larger than the preset confidence coefficient threshold value of the corresponding defect type, judging that the initial picture has single defect, and taking the defect type corresponding to the single defect as the output result of the initial picture.
7. The defect identification method of claim 1, wherein the defect identification model is obtained by:
collecting historical defect pictures of the OLED panel;
marking the defect characteristics in the historical defect picture to obtain a defect sample picture and defect information corresponding to the defect sample picture;
and inputting the defect sample picture and the defect information into a basic neural network, and training to obtain the defect identification model.
8. A defect identifying apparatus for identifying a defect of an OLED panel, the apparatus comprising:
the device comprises an acquisition module, a detection module and a display module, wherein the acquisition module is used for acquiring an initial picture of the OLED panel to be detected and inputting the initial picture into a defect recognition model trained in advance to obtain a primary recognition result, and the primary recognition result comprises the type of the defect and the confidence coefficient of the defect;
the first judgment module is used for judging whether the confidence coefficient of each defect is greater than a preset confidence coefficient threshold corresponding to the defect type;
the second judging module is used for judging whether at least two first defects of preset types exist in all the defects with the confidence degrees larger than the confidence degree threshold value preset by the corresponding defect type or not if the confidence degrees of at least two defects are larger than the confidence degree threshold value preset by the corresponding defect type;
and the determining module is used for obtaining a target defect type of the first defect by calculating the similarity between each first defect and the corresponding defect template if at least two first defects exist, wherein the target defect type of the first defect is the defect type corresponding to the defect template with the maximum similarity to the first defect, and the target defect type is used as the output result of the initial picture.
9. A computer device, characterized in that the computer device comprises a memory and a processor, the memory storing a computer program which, when run by the processor, performs the defect identification method of any one of claims 1-7.
10. A computer-readable storage medium, characterized in that the computer-readable storage medium has stored thereon a computer program which, when being executed by a processor, carries out the method for defect identification according to any one of claims 1-7.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210281601.9A CN114627089A (en) | 2022-03-21 | 2022-03-21 | Defect identification method, defect identification device, computer equipment and computer readable storage medium |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210281601.9A CN114627089A (en) | 2022-03-21 | 2022-03-21 | Defect identification method, defect identification device, computer equipment and computer readable storage medium |
Publications (1)
Publication Number | Publication Date |
---|---|
CN114627089A true CN114627089A (en) | 2022-06-14 |
Family
ID=81903931
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202210281601.9A Pending CN114627089A (en) | 2022-03-21 | 2022-03-21 | Defect identification method, defect identification device, computer equipment and computer readable storage medium |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN114627089A (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN115616082A (en) * | 2022-12-14 | 2023-01-17 | 杭州兆华电子股份有限公司 | Keyboard defect analysis method based on noise detection |
CN116258721A (en) * | 2023-05-16 | 2023-06-13 | 成都数之联科技股份有限公司 | OLED panel defect judging method, device, equipment and medium |
CN116645362A (en) * | 2023-06-29 | 2023-08-25 | 日照鲁光电子科技有限公司 | Intelligent quality detection method and system for silicon carbide wafer |
Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20170206431A1 (en) * | 2016-01-20 | 2017-07-20 | Microsoft Technology Licensing, Llc | Object detection and classification in images |
CN106991661A (en) * | 2017-03-31 | 2017-07-28 | 重庆大学 | Merge KL conversion and the non-local mean denoising method of grey relational grade |
CN107239740A (en) * | 2017-05-05 | 2017-10-10 | 电子科技大学 | A kind of SAR image automatic target recognition method of multi-source Fusion Features |
CN111242124A (en) * | 2020-01-13 | 2020-06-05 | 支付宝实验室(新加坡)有限公司 | Certificate classification method, device and equipment |
CN112419401A (en) * | 2020-11-23 | 2021-02-26 | 上海交通大学 | Aircraft surface defect detection system based on cloud edge cooperation and deep learning |
WO2021217857A1 (en) * | 2020-04-27 | 2021-11-04 | 平安科技(深圳)有限公司 | Slice defect detection method and apparatus, and electronic device and readable storage medium |
WO2021232613A1 (en) * | 2020-05-22 | 2021-11-25 | 五邑大学 | Liquor bottle surface defect inspection method, electronic device, and storage medium |
CN114021964A (en) * | 2021-11-03 | 2022-02-08 | 福州大学 | Industrial enterprise equipment working condition judging and environment-friendly condition monitoring method |
CN114140625A (en) * | 2022-01-30 | 2022-03-04 | 深圳市博铭维技术股份有限公司 | Pipeline defect detection method, system, equipment and computer readable storage medium |
-
2022
- 2022-03-21 CN CN202210281601.9A patent/CN114627089A/en active Pending
Patent Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20170206431A1 (en) * | 2016-01-20 | 2017-07-20 | Microsoft Technology Licensing, Llc | Object detection and classification in images |
CN106991661A (en) * | 2017-03-31 | 2017-07-28 | 重庆大学 | Merge KL conversion and the non-local mean denoising method of grey relational grade |
CN107239740A (en) * | 2017-05-05 | 2017-10-10 | 电子科技大学 | A kind of SAR image automatic target recognition method of multi-source Fusion Features |
CN111242124A (en) * | 2020-01-13 | 2020-06-05 | 支付宝实验室(新加坡)有限公司 | Certificate classification method, device and equipment |
WO2021217857A1 (en) * | 2020-04-27 | 2021-11-04 | 平安科技(深圳)有限公司 | Slice defect detection method and apparatus, and electronic device and readable storage medium |
WO2021232613A1 (en) * | 2020-05-22 | 2021-11-25 | 五邑大学 | Liquor bottle surface defect inspection method, electronic device, and storage medium |
CN112419401A (en) * | 2020-11-23 | 2021-02-26 | 上海交通大学 | Aircraft surface defect detection system based on cloud edge cooperation and deep learning |
CN114021964A (en) * | 2021-11-03 | 2022-02-08 | 福州大学 | Industrial enterprise equipment working condition judging and environment-friendly condition monitoring method |
CN114140625A (en) * | 2022-01-30 | 2022-03-04 | 深圳市博铭维技术股份有限公司 | Pipeline defect detection method, system, equipment and computer readable storage medium |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN115616082A (en) * | 2022-12-14 | 2023-01-17 | 杭州兆华电子股份有限公司 | Keyboard defect analysis method based on noise detection |
CN115616082B (en) * | 2022-12-14 | 2023-04-28 | 杭州兆华电子股份有限公司 | Keyboard defect analysis method based on noise detection |
CN116258721A (en) * | 2023-05-16 | 2023-06-13 | 成都数之联科技股份有限公司 | OLED panel defect judging method, device, equipment and medium |
CN116645362A (en) * | 2023-06-29 | 2023-08-25 | 日照鲁光电子科技有限公司 | Intelligent quality detection method and system for silicon carbide wafer |
CN116645362B (en) * | 2023-06-29 | 2024-01-23 | 日照鲁光电子科技有限公司 | Intelligent quality detection method and system for silicon carbide wafer |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN112949564B (en) | Pointer type instrument automatic reading method based on deep learning | |
CN110060237B (en) | Fault detection method, device, equipment and system | |
CN114627089A (en) | Defect identification method, defect identification device, computer equipment and computer readable storage medium | |
CN111161243B (en) | Industrial product surface defect detection method based on sample enhancement | |
CN111915704A (en) | Apple hierarchical identification method based on deep learning | |
CN108827969A (en) | Metal parts surface defects detection and recognition methods and device | |
CN109858480A (en) | A kind of digital instrument recognition methods | |
CN111161295B (en) | Dish image background stripping method | |
CN114612469A (en) | Product defect detection method, device and equipment and readable storage medium | |
CN111062934A (en) | Real-time detection method for fabric image defects | |
CN114119591A (en) | Display screen picture quality detection method | |
CN114549441A (en) | Sucker defect detection method based on image processing | |
CN114266764A (en) | Character integrity detection method and device for printed label | |
CN110781913A (en) | Zipper cloth belt defect detection method | |
CN116228741A (en) | PCBA (printed circuit board assembly) component defect detection method and device | |
CN112070762A (en) | Mura defect detection method and device for liquid crystal panel, storage medium and terminal | |
CN112381751A (en) | Online intelligent detection system and method based on image processing algorithm | |
CN114612399A (en) | Picture identification system and method for mobile phone appearance mark | |
CN111028209B (en) | Telephone silk-screen quality detection method and system | |
CN107545565A (en) | A kind of solar energy half tone detection method | |
CN115330794B (en) | LED backlight foreign matter defect detection method based on computer vision | |
CN116843618A (en) | Method for detecting shallow apparent appearance defects of metal parts | |
CN116664540A (en) | Rubber sealing ring surface defect detection method based on Gaussian line detection | |
CN116503340A (en) | Micro oled panel defect detection method, device and equipment | |
US20230386023A1 (en) | Method for detecting medical images, electronic device, and storage medium |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination |