CN114627089B - Defect identification method, defect identification device, computer equipment and computer readable storage medium - Google Patents

Abstract

The present application provides a defect recognition method, device, computer equipment and computer-readable storage medium, which relate to the field of defect detection technology. By obtaining an initial image of the OLED panel to be detected, the initial image is input into a pre-trained defect recognition model to obtain a preliminary recognition result, which includes the type of defect and the confidence of the defect; it is determined whether there are at least two first defects of preset types among all defects whose confidence is greater than the confidence threshold preset for the corresponding defect type; if there are at least two first defects, the similarity between each first defect and the corresponding defect template is calculated, and the target defect type of the first defect is obtained. The target defect type of the first defect is the defect type corresponding to the defect template with the greatest similarity to the first defect, and the target defect type is used as the output result of the initial image. The defect recognition method provided by this embodiment can improve the accuracy of image judgment.

Description

Defect identification method, device, computer equipment and computer readable storage medium

Technical Field

The present invention relates to the field of defect detection technology, and in particular to a defect recognition method, device, computer equipment and computer-readable storage medium.

Background Art

Organic Light-Emitting Diode (OLED) has many excellent properties and is generally considered to be an ideal next-generation flat panel display technology. However, due to its complex production process, various defects are inevitable during the preparation process. These defects have the characteristics of blurred boundaries, irregular shapes, periodic texture backgrounds, and uneven overall brightness. Therefore, the detection of surface defects on OLED displays is conducive to statistical analysis of defects, while repairing or eliminating products, and improving their manufacturing quality through process improvements to avoid wasting production resources.

Currently, the inspection of OLED displays usually uses cameras to take pictures and perform manual image judgment, but the cost of manual image judgment is high and is affected by subjective factors and the proficiency of the judge. Therefore, deep learning technology has become the mainstream method for multi-defect recognition. However, due to the different forms of defects generated in the production environment, which are complex and changeable, simple discrimination models have the problem of low recognition precision and accuracy.

Summary of the invention

In order to solve the above technical problems, the 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, the method comprising:

Obtaining an initial image of the OLED panel to be inspected, and inputting the initial image into a pre-trained defect recognition model to obtain a preliminary recognition result, wherein the preliminary recognition result includes the type of defect and the confidence level of the defect;

Determining whether the confidence level of each defect is greater than a preset confidence level threshold for the corresponding defect type;

If there are at least two defects whose confidences are greater than a preset confidence threshold for the corresponding defect type, determining whether there are at least two first defects of a preset type among all defects whose confidences are greater than the preset confidence threshold for the corresponding defect type;

If there are at least two first defects, the target defect type of the first defect is obtained by calculating the similarity between each of the first defects and the corresponding defect template, wherein the target defect type of the first defect is the defect type corresponding to the defect template with the greatest similarity to the first defect, and the target defect type is used as the output result of the initial image.

In a specific embodiment, if the confidence of at least two defects is greater than the preset confidence threshold of the corresponding defect type, after the step of determining whether there are at least two first defects of the preset type among all defects with confidence greater than the preset confidence threshold of the corresponding defect type, the method further includes:

If there are at least two of the first defects, determining whether there is a second defect among all defects whose confidence is greater than a preset confidence threshold corresponding to the defect type;

If the second defect exists, a priority judgment is performed on the target defect type and the defect type of each second defect, and the defect type corresponding to the defect with the highest priority is used as the output result of the initial image. The first defect is a defect whose confidence is greater than a first preset multiple of the maximum confidence, and the second defect is a defect whose confidence is greater than a second preset multiple of the maximum confidence.

In a specific implementation, the maximum confidence is the maximum value of the confidences of the defects in the preliminary identification result.

In a specific implementation, the step of obtaining the target defect type of the first defect by calculating the similarity between each of the first defects and the corresponding defect template includes:

Acquire the center point coordinates 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 the grey correlation degree between the defect sequence of the first defect and each defect template;

The defect type corresponding to the defect template corresponding to the maximum value of the grey relational degree of each defect template is determined as the target defect type.

In a specific embodiment, the defect type of the defect template includes at least one of a foreign matter type on the OLED panel film, a foreign matter type under the film, and a dirt type.

In a specific implementation, after the step of determining whether the confidence level of each defect is greater than a preset confidence level threshold corresponding to the defect type, the method further includes:

If the confidence of the existence of a defect is greater than the preset confidence threshold of the corresponding defect type, it is determined that there is a single defect in the initial image, and the defect type corresponding to the single defect is used as the output result of the initial image.

In a specific implementation, the step of acquiring the defect recognition model is:

Collect historical defect images of OLED panels;

Marking defect features in the historical defect images to obtain defect sample images and defect information corresponding to the defect sample images;

The defect sample image and the defect information are input into a basic neural network and trained to obtain the defect recognition model.

In a second aspect, an embodiment of the present application provides a defect identification device, including:

An acquisition module, used to acquire an initial image of the OLED panel to be inspected, and input the initial image into a pre-trained defect recognition model to obtain a preliminary recognition result, wherein the preliminary recognition result includes the type of defect and the confidence level of the defect;

A first judgment module is used to judge whether the confidence of each defect is greater than a preset confidence threshold of the corresponding defect type;

A second judgment module is used to judge whether there are at least two first defects of a preset type among all defects with confidences greater than the preset confidence threshold of the corresponding defect type if there are at least two defects with confidences greater than the preset confidence threshold of the corresponding defect type;

A determination module is used to obtain a target defect type of the first defect by calculating the similarity between each of the first defects and the corresponding defect template if there are at least two first defects, wherein the target defect type of the first defect is the defect type corresponding to the defect template having the greatest similarity to the first defect, and the target defect type is used as the output result of the initial image.

In a third aspect, an embodiment of the present application provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program, and when the processor runs the computer program, the defect identification method described in the first aspect is executed.

In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the defect identification method described in the first aspect is implemented.

The defect recognition method provided in the embodiment of the present application uses a pre-trained defect recognition model to identify the OLED panel image to be inspected, screens the obtained preliminary recognition results using a confidence threshold, and then performs subsequent similarity judgment, thereby improving the precision and accuracy of OLED panel defect recognition.

BRIEF DESCRIPTION OF THE DRAWINGS

In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

FIG1 is a schematic diagram showing a flow chart of a defect identification method provided by an embodiment of the present invention;

FIG. 2 shows a schematic structural diagram of a defect identification device provided in an embodiment of the present invention.

DETAILED DESCRIPTION

The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

The components of the embodiments of the present invention generally described and shown in the drawings herein may be arranged and designed in a variety of different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the claimed invention, but merely represents selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

Hereinafter, the terms "including", "having" and their cognates, which may be used in various embodiments of the present invention, are intended only to indicate specific features, numbers, steps, operations, elements, components or combinations of the foregoing items, and should not be understood as first excluding the existence of one or more other features, numbers, steps, operations, elements, components or combinations of the foregoing items or adding the possibility of one or more features, numbers, steps, operations, elements, components or combinations of the foregoing items.

Furthermore, the terms “first”, “second”, “third”, etc. are merely used for distinguishing descriptions and are not to be understood as indicating or implying relative importance.

Unless otherwise defined, all terms (including technical terms and scientific terms) used herein have the same meanings as those generally understood by those skilled in the art to which the various embodiments of the present invention belong. The terms (such as those defined in generally used dictionaries) will be interpreted as having the same meanings as the contextual meanings in the relevant technical field and will not be interpreted as having idealized meanings or overly formal meanings unless clearly defined in the various embodiments of the present invention.

Example 1

Referring to Figure 1, Figure 1 is a schematic flow chart of a defect identification method provided by this embodiment. As shown in Figure 1, the method includes:

Step S101, obtaining an initial image of the OLED panel to be inspected, and inputting the initial image into a pre-trained defect recognition model to obtain a preliminary recognition result, wherein the preliminary recognition result includes the type of defect and the confidence level of the defect.

In specific implementation, the production process of OLED panels is prone to various product defects, such as color mixing, color loss, dirt, etc., due to the excessive and complicated process flow, which will affect the normal use of OLED panels. It is necessary to detect possible defects in OLED panels before leaving the factory. At present, deep learning technology is mostly used to identify defects in OLED panels, but due to the different production environments of OLED panels, the defects generated are of different forms and complex and changeable, and the recognition precision and accuracy of conventional deep learning recognition models are low.

This embodiment provides a method of combining deep learning with subsequent processing to identify defects in OLED panels, that is, for the OLED panel to be detected, the corresponding picture is collected, and the collected picture is used for defect detection, and the picture to be detected can be defined as the initial picture. The initial picture can be obtained by directly photographing the OLED panel to be detected, or by obtaining a picture of the OLED panel to be detected sent by other terminals from the network.

Specifically, a neural network with target detection and positioning functions is pre-trained, which is defined as a defect recognition model. The collected initial images to be detected are input into the neural network model, and a defect information table corresponding to each initial image can be obtained, which is defined as a preliminary recognition result. The defect information table includes information such as the type, confidence and location of each estimated defect. Among them, the estimated defects are all the estimated results of the defect recognition model on the defects existing in the initial image. Each estimated result corresponds to a defect type. For example, there is only one defect in the initial image, but the defect recognition model may give multiple estimated results after recognizing the initial image, that is, the defect information table obtained after recognition includes multiple defect information. Based on this, the defect recognition model often cannot give a very accurate recognition result when recognizing the defects existing in the initial image. Therefore, it is necessary to correct the recognition result of the defect recognition model, that is, it is necessary to perform subsequent processing on each defect in the defect information table to improve the accuracy of defect recognition.

Step S102, determining whether the confidence level of each defect is greater than a preset confidence level threshold corresponding to the defect type.

In specific implementation, before processing each defect in the defect information table, it is necessary to set a confidence threshold for each defect that may exist on the OLED panel, compare the confidence of the defect with the preset confidence threshold of the defect, and determine the processing method for the defect based on the size relationship.

Specifically, the confidence threshold can 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 preprocessed, and the preprocessed data is summarized and analyzed. On this basis, a suitable confidence threshold is set for each defect that may exist on the OLED panel.

Specifically, in the defect information table, defects with confidence values less than the confidence threshold are assigned the value of other, and no subsequent processing is performed. If the values of all defects in the defect information table are other, that is, the confidence values of all defects are less than the confidence threshold, then the other class is output as the recognition result of the initial image. In addition, defects with confidence values greater than the confidence threshold continue to undergo subsequent processing operations. Through the card control of the confidence threshold, defects that require subsequent operations can be directly screened out, thereby improving the accuracy of defect recognition to a certain extent.

Step S103, if there are at least two defects whose confidences are greater than the preset confidence threshold of the corresponding defect type, then determine whether there are at least two first defects of the preset type among all defects whose confidences are greater than the preset confidence threshold of the corresponding defect type;

In a specific implementation, the maximum confidence is the maximum value of the confidences of the defects in the preliminary identification result.

In specific implementation, defects that require subsequent processing are screened out based on the confidence threshold. The confidence of these defects needs to be greater than their preset confidence threshold. Among all defects that meet the above conditions, the defect with a confidence greater than a times the maximum confidence is defined as the first defect. The value of a can be 0.8 and can be adjusted according to actual conditions. The maximum confidence refers to the maximum value of the confidence of each defect in the defect information table.

Specifically, when there are two or more first defects, it is necessary to determine the number of defects of a preset type among all the first defects. The preset types include foreign matter on the OLED film, foreign matter under the film, and dirt. Because these three defects are very similar, the defect recognition model is particularly easy to be confused when identifying these three defects. For example, there is a defect of dirt in the initial picture. The defect information table obtained after the defect recognition model identifies it may include two or three of the three defects of foreign matter on the film, foreign matter under the film, and dirt, resulting in the inability to accurately determine the specific defect type.

Specifically, in the present embodiment, a subsequent processing method is set for these three defects, that is, when there are at least two first defects in the identification results, and the types of the first defects include at least two defect types of foreign matter on the membrane, foreign matter under the membrane and dirt, it is necessary to perform subsequent judgment on the first defects including at least two defect types, so as to improve the recognition accuracy of these three defects.

Step S104: if there are at least two first defects, a target defect type of the first defect is obtained by calculating the similarity between each of the first defects and the corresponding defect template, wherein the target defect type of the first defect is the defect type corresponding to the defect template having the greatest similarity to the first defect, and the target defect type is used as the output result of the initial image;

In a specific implementation, 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 includes:

Acquire the center point coordinates 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 the grey correlation degree between the defect sequence of the first defect and each defect template;

Determine the defect type corresponding to the defect template corresponding to the maximum value of the grey correlation degree of each defect template as the target defect type;

In a specific embodiment, the defect type of the defect template includes at least one of a foreign matter type on the OLED panel film, a foreign matter type under the film, and a dirt type.

In specific implementation, because the three defects of foreign matter on the OLED panel film, foreign matter under the film and dirt are easy to be confused, the defect recognition model is difficult to accurately identify. This embodiment provides a similarity judgment method to distinguish the three defects of foreign matter on the OLED panel film, foreign matter under the film and dirt, that is, by respectively calculating the gray correlation between the defect sequence of the three defects existing in the initial image and the defect templates corresponding to the three defects, the defect type corresponding to the maximum gray correlation is defined as the target defect type, and the target defect type is used as the defect recognition result of the initial image, and the defect recognition result is defined as the output result.

Specifically, the defect sequence is composed of the center point coordinates of the defect and the edge values of the pixel area of the defect. The center point coordinates of the defect are the center point coordinates of the pixel area of the defect, and the edge values of the pixel area of the defect are the maximum values of the horizontal and vertical coordinates of the minimum circumscribed rectangular box of the pixel area of the defect. The center point coordinates of the pixel area of the defect are calculated based on the maximum values of the horizontal and vertical coordinates of the minimum circumscribed rectangular box of the pixel area of the defect.

Specifically, the average value is calculated based on the maximum and minimum values of the horizontal coordinates of the minimum circumscribed rectangular box of the defective pixel area to obtain the horizontal coordinate of the center point of the defective pixel area. The average value is calculated based on the maximum and minimum values of the vertical coordinates of the minimum circumscribed rectangular box of the defective pixel area to obtain the vertical coordinate of the center point of the defective pixel area. The maximum values of the horizontal and vertical coordinates of the minimum circumscribed rectangular box of the defective pixel area can be obtained based on the position information of each defect in the defect information table.

Specifically, according to the above calculation process, the defect sequence [o _xi , o _yi , xmax _i , xmin _i , ymax _i , ymin _i ] corresponding to the three defects in the initial image is obtained, where [o _xi , o _yi ] are the center point coordinates of the defective pixel area, [xmax _i , xmin _i , ymax _i , ymin _i ] are the set of edge values of the defective pixel area, and i represents any one of foreign matter on the OLED panel film, foreign matter under the film, and dirt.

Specifically, when training a defect recognition model, it is necessary to mark sample images to facilitate the summary and learning of the characteristics and types of various sample images. In the process of marking the sample images, the data of the circumscribed rectangular boxes of the pixel areas corresponding to each defect are obtained, and the data of the circumscribed rectangular boxes of the pixel areas corresponding to the three defects of foreign matter on the OLED panel film, foreign matter under the film, and dirt are selected to obtain the maximum coordinate values of each circumscribed rectangular box, and the coordinates of the center points of the pixel areas corresponding to the three defects are calculated according to the maximum coordinate values of each circumscribed rectangular box, that is, the horizontal coordinate of the center point of the defective pixel area is obtained by calculating the average value according to the maximum and minimum values of the horizontal coordinate of the circumscribed rectangular box, and the vertical coordinate of the center point of the defective pixel area is obtained by calculating the average value according to the maximum and minimum values of the vertical coordinate of the circumscribed rectangular box.

Specifically, defect sequences corresponding to the three defects are formed according to the center point coordinates of the pixel areas of the three defects and the maximum values of the horizontal and vertical coordinates of the circumscribed rectangular boxes of the pixel areas. Because the three defects may appear multiple times in historical defect images, there may be multiple defect sequences corresponding to the three defects. The average value of the multiple defect sequences of each of the three defects is calculated to obtain the average defect sequence corresponding to each of the three defects. The average defect sequence can be defined as a defect template. Taking the calculation of the defect template of the dirt defect as an example, first, multiple defect sequences corresponding to the dirt defect are calculated, and then the average value of the maximum values of the horizontal coordinates of all the circumscribed rectangular boxes in the multiple defect sequences is calculated to obtain the maximum value of the horizontal coordinate of the circumscribed rectangular box in the average sequence. The calculation of the remaining values in the average defect sequence is consistent with the calculation method of the maximum value of the horizontal coordinate of the circumscribed rectangular box. Through the above calculation process, the defect template of the dirt defect is obtained. For the calculation of the defect template of foreign matter on the OLED panel film and foreign matter under the film, please refer to the calculation process of the defect template of the dirt defect. To avoid repetition, it will not be repeated here.

Specifically, if the defect types existing in the initial image include at least two of foreign matter on the OLED panel film, foreign matter under the film, and dirt, and the confidences corresponding to at least two types of defects are both greater than a times the maximum confidence, for example, there are two defects, foreign matter on the film and foreign matter under the film, in the initial image, and the confidences corresponding to the two defects are both greater than a times the maximum confidence, then it is necessary to perform similarity judgment on the two defects, and the steps of similarity judgment include: first, obtaining the defect sequence of the two defects, and then respectively calculating the gray correlation between the defect sequence of the two defects and the corresponding defect template. According to the calculation result, if the gray correlation corresponding to the foreign matter on the film is greater than the gray correlation corresponding to the foreign matter under the film, then the result of the similarity judgment is the foreign matter on the film. At this time, the foreign matter on the film is used as the defect recognition result of the initial image. The gray correlation calculation process includes: calculating the gray correlation coefficient between the defect sequence and the corresponding defect template, and averaging the gray correlation coefficients to obtain the gray correlation. The specific calculation formula is as follows:

In the above formula, γ(X ₀ ,X _i ) represents the grey correlation degree between the defect sequence _Xi and the corresponding defect template X ₀ , γ(x ₀ (k), _xi (k)) represents the kth grey correlation coefficient between the defect sequence _Xi and the corresponding defect template X ₀ , It means to find the average value of each grey relational coefficient of defect sequence _Xi and corresponding defect template _X0 , where _Xi (k) and _x0 (k) represent the kth value in defect sequence _Xi and corresponding defect template _X0 respectively, k＝1,2,…,6, n represents the number of grey relational coefficients, n＝6, ρ∈(0,1) is the resolution coefficient.

x _i (k)d _t can represent the initial value image of the kth value in the defect sequence _Xi , and can also represent any one of the mean image, inverse image and reciprocal image of the kth value in the defect sequence _Xi , and can be selected according to actual needs. In this embodiment, x _i (k)d _t represents an example of the initial value image of the kth value in the defect sequence _Xi . As described above, x ₀ (k)d _t in this embodiment represents an example of the initial value image of the kth value in the defect template X _0, and t=1, 2, 3, 4 correspond to the initial value image, mean image, inverse image and reciprocal image of each value in the defect sequence _Xi and the defect template X _0, respectively.

When t=1, |x ₀ (k)d _{t -xi} (k)d _t | represents the absolute value of the difference between the initial image of the defect template X ₀ and the kth value in the defect sequence _Xi ; It means to traverse the absolute value of the difference between the initial value image of each corresponding value in the defect template _X0 and the defect sequence Xi _, and select the minimum value; It means to traverse the absolute values of the differences between the initial images of the corresponding values in the defect template _X0 and the defect sequence _Xi , and select the maximum value.

Because the data in each defect sequence and defect template may be inconvenient to compare or difficult to obtain correct conclusions due to different dimensions, in order to ensure the reliability of the results, when performing grey correlation analysis, the data is first dimensionally processed. The methods used in this embodiment include initialization, averaging, inversion and reciprocalization. Each defect sequence and defect template is initialized to obtain the initial image corresponding to each defect sequence and defect template. For other methods, please refer to the description of the initialization process above. To avoid repetition, it will not be repeated here. According to the actual situation, any of the above methods can be selected to process the data.

In a specific embodiment, if the confidence of at least two defects is greater than the preset confidence threshold of the corresponding defect type, after the step of determining whether there are at least two first defects of the preset type among all defects with confidence greater than the preset confidence threshold of the corresponding defect type, the method further includes:

If there are at least two of the first defects, determining whether there is a second defect among all defects whose confidence is greater than a preset confidence threshold corresponding to the defect type;

If the second defect exists, a priority judgment is performed on the target defect type and the defect type of each second defect, and the defect type corresponding to the defect with the highest priority is used as the output result of the initial image. The first defect is a defect whose confidence is greater than a first preset multiple of the maximum confidence, and the second defect is a defect whose confidence is greater than a second preset multiple of the maximum confidence.

In specific implementation, a defect with a confidence level greater than b times the maximum confidence level can be defined as a second defect, and the value of b is 0.8. The value of b can be adjusted accordingly based on the recognition result. The priority judgment of the defect refers to sorting the defects according to the priority, and taking the defect with the highest priority as the defect recognition result of the initial image. The basis for priority sorting can be the degree of influence of each defect. The greater the degree of influence, the higher the priority of the defect. Of course, the defects can also be prioritized according to other factors.

Specifically, after making a similarity judgment, it is also necessary to determine whether a second defect exists on the initial image. If a second defect exists, a priority judgment is made on each second defect and target defect type existing on the initial image, and the defect with the highest priority is used as the defect recognition result of the initial image.

Specifically, when no similarity judgment is performed, if there are at least two second defects on the initial image, it is necessary to perform a priority judgment on each second defect, and use the defect with the highest priority as the defect recognition result of the initial image.

In a specific implementation, after the step of determining whether the confidence level of each defect is greater than a preset confidence level threshold corresponding to the defect type, the method further includes:

If the confidence level of a defect is greater than a preset confidence threshold of the corresponding defect type, it is determined that a single defect exists in the initial image, and the defect type corresponding to the single defect is used as the output result of the initial image.

In specific implementation, after the initial image is detected by the defect recognition model, if there is only one defect whose confidence is greater than the confidence threshold of the defect, it means that there is only one defect in the initial image, and the defect is directly used as the defect recognition result of the initial image.

In a specific implementation, the step of acquiring the defect recognition model is:

Collect historical defect images of OLED panels;

Marking defect features in the historical defect images to obtain defect sample images and defect information corresponding to the defect sample images;

The defect sample image and the defect information are input into a basic neural network and trained to obtain the defect recognition model.

During specific implementation, a defect data set for training needs to be prepared before training the defect recognition model. The defect data set includes historical pictures of various defects existing in the OLED panel. These pictures can be defined as historical defect pictures. The historical defect pictures can be obtained by directly photographing and collecting the OLED panels including various defects, or by obtaining pictures of OLED panels including various defects sent by other terminals from the network.

Specifically, the collected historical defect images of OLED panels are cropped into defect sample images of appropriate sizes, and the defect sample images are classified according to the defect type. The cropped defect sample images are annotated with the Labelmg annotation tool to mark the location and type of the defects. The annotation tool will generate an XML file with the same name and annotations as the defect sample image. Each XML file is converted to finally obtain a json file. The json file and each defect sample image are input into the neural network model as a data set, and the Faster RCNN algorithm is used for training to obtain a defect recognition model.

The defect recognition method provided in this embodiment uses a pre-trained defect recognition model to identify the OLED panel image to be inspected, and the preliminary recognition results are screened using a confidence threshold, and then subsequent similarity judgments and priority judgments are performed, thereby improving the precision and accuracy of OLED panel defect recognition.

Example 2

2 , this embodiment further provides a defect identification device 200, including:

An acquisition module 201 is used to acquire an initial image of the OLED panel to be inspected, and input the initial image into a pre-trained defect recognition model to obtain a preliminary recognition result, wherein the preliminary recognition result includes an actual confidence level of each defect existing in the initial image;

The first judgment module 202 is used to judge whether the actual confidence level of each defect is within the confidence level threshold range corresponding to the preset multiple types of defects;

The second judgment module 203 is used to judge whether there are at least two first defects of a preset type among all defects with confidences greater than the preset confidence threshold of the corresponding defect type if there are at least two defects with confidences greater than the preset confidence threshold of the corresponding defect type;

Determination module 204 is used to obtain the target defect type of the first defect by calculating the similarity between each of the first defects and the corresponding defect template if there are at least two first defects, wherein the target defect type of the first defect is the defect type corresponding to the defect template having the greatest similarity with the first defect, and the target defect type is used as the output result of the initial image.

In one implementation, the acquisition module 201 is specifically used to: collect historical defect images of OLED panels;

Marking defect features in the historical defect images to obtain defect sample images and defect information corresponding to the defect sample images;

The defect sample image and the defect information are input into a basic neural network and trained to obtain the defect recognition model.

In one embodiment, the first judgment module 202 is specifically used to: if the confidence of the existence of a defect is greater than the preset confidence threshold of the corresponding defect type, it is determined that there is a single defect in the initial image, and the defect type corresponding to the single defect is used as the output result of the initial image.

In one implementation, the second judgment module 203 is specifically configured to: the maximum confidence is the maximum value among the confidences of the defects in the preliminary identification result.

In one embodiment, the determination module 204 is specifically configured to: if there are at least two of the first defects, determine whether there is a second defect among all defects whose confidence is greater than a preset confidence threshold corresponding to the defect type;

If the second defect exists, a priority judgment is performed on the target defect type and the defect type of each second defect, and the defect type corresponding to the defect with the highest priority is used as the output result of the initial image. The first defect is a defect whose confidence is greater than a first preset multiple of the maximum confidence, and the second defect is a defect whose confidence is greater than a second preset multiple of the maximum confidence.

In one embodiment, the determining module 204 is further configured to: obtain the coordinates of the center point of the first defect and the edge value of the pixel area of the first defect to form a defect sequence corresponding to the first defect;

respectively calculating the grey correlation degree between the defect sequence of the first defect and each defect template;

The defect type corresponding to the defect template corresponding to the maximum value of the grey relational degree of each defect template is determined as the target defect type.

In one embodiment, the determination module 204 is further specifically configured to: the defect type of the defect template includes at least one of a foreign matter type on the OLED panel film, a foreign matter type under the film, and a dirt type.

The specific functions of the defect identification device 200 provided in this embodiment can be found in the specific implementation process of the defect identification method in Example 1, and will not be described in detail here.

The defect recognition device provided in this embodiment can identify the OLED panel image to be inspected through a pre-trained defect recognition model, and filter the preliminary recognition results obtained using a confidence threshold, and then perform subsequent similarity judgment and priority judgment, thereby improving the precision and accuracy of OLED panel defect recognition.

Example 3

This embodiment provides a computer device, which includes a memory and a processor. The memory stores a computer program. When the processor runs the computer program, the defect identification method described in Embodiment 1 is executed.

The computer device provided in this embodiment can implement the defect identification method described in Example 1, and to avoid repetition, it will not be described here.

Example 4

This embodiment provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the defect identification method described in Embodiment 1 is implemented.

The computer-readable storage medium provided in this embodiment can implement the defect identification method described in Example 1, and will not be described again here to avoid repetition.

In the embodiments provided by the present invention, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some communication interfaces, and the indirect coupling or communication connection of devices or units can be electrical, mechanical or other forms.

The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

In addition, each functional unit in the embodiment provided by the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in each embodiment of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

It should be noted that similar numbers and letters represent similar items in the following figures. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. In addition, the terms "first", "second", "third", etc. are only used to distinguish the description and are not to be understood as indicating or implying relative importance.

Finally, it should be noted that the above-mentioned embodiments are only specific implementation methods of the present invention, which are used to illustrate the technical solutions of the present invention, rather than to limit them. The protection scope of the present invention is not limited thereto. Although the present invention is described in detail with reference to the above-mentioned embodiments, ordinary technicians in the field should understand that any technician familiar with the technical field can still modify the technical solutions recorded in the above-mentioned embodiments within the technical scope disclosed by the present invention, or can easily think of changes, or make equivalent replacements for some of the technical features therein; and these modifications, changes or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention. They should all be covered within the protection scope of the present invention.

Claims (9)

1. A defect identification method, characterized in that it is used to identify defects of an OLED panel, the method comprising: Obtaining an initial image of the OLED panel to be inspected, and inputting the initial image into a pre-trained defect recognition model to obtain a preliminary recognition result, wherein the preliminary recognition result includes the type of defect and the confidence level of the defect; Determining whether the confidence level of each defect is greater than a preset confidence level threshold for the corresponding defect type; If there are at least two defects whose confidences are greater than a preset confidence threshold for the corresponding defect type, determining whether there are at least two first defects of a preset type among all defects whose confidences are greater than the preset confidence threshold for the corresponding defect type; If there are at least two first defects, a target defect type of the first defect is obtained by calculating the similarity between each of the first defects and the corresponding defect template, wherein the target defect type of the first defect is the defect type corresponding to the defect template having the greatest similarity to the first defect, and the target defect type is used as the output result of the initial image; Determine whether there is a second defect among all defects whose confidence is greater than a preset confidence threshold of the corresponding defect type; If the second defect exists, a priority judgment is performed on the target defect type and the defect types of each of the second defects, and the defect type corresponding to the defect with the highest priority is used as the output result of the initial image. The first defect is a defect whose confidence is greater than a first preset multiple of the maximum confidence, and the second defect is a defect whose confidence is greater than a second preset multiple of the maximum confidence. The priority is the degree of influence of each defect on the OLED panel. 2. The defect identification method according to claim 1 is characterized in that the maximum confidence is the maximum value among the confidences of each defect in the preliminary identification result. 3. The defect identification method according to claim 1, characterized in that 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: Acquire the center point coordinates 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 the grey correlation degree between the defect sequence of the first defect and each defect template; The defect type corresponding to the defect template corresponding to the maximum value of the grey relational degree of each defect template is determined as the target defect type. 4. The defect identification method according to claim 1 is characterized in that the defect type of the defect template includes at least one of a foreign matter type on the OLED panel film, a foreign matter type under the film, and a dirt type. 5. The defect identification method according to claim 1, characterized in that after the step of determining whether the confidence of each defect is greater than a preset confidence threshold corresponding to the defect type, the method further comprises: If the confidence level of a defect is greater than a preset confidence threshold of the corresponding defect type, it is determined that a single defect exists in the initial image, and the defect type corresponding to the single defect is used as the output result of the initial image. 6. The defect recognition method according to claim 1, characterized in that the step of acquiring the defect recognition model is: Collect historical defect images of OLED panels; Marking defect features in the historical defect images to obtain defect sample images and defect information corresponding to the defect sample images; The defect sample image and the defect information are input into a basic neural network and trained to obtain the defect recognition model. 7. A defect recognition device, characterized in that it is used to identify defects of an OLED panel, the device comprising: An acquisition module, used to acquire an initial image of the OLED panel to be inspected, and input the initial image into a pre-trained defect recognition model to obtain a preliminary recognition result, wherein the preliminary recognition result includes the type of defect and the confidence level of the defect; A first judgment module is used to judge whether the confidence of each defect is greater than a preset confidence threshold of the corresponding defect type; A second judgment module is used to judge whether there are at least two first defects of a preset type among all defects with confidences greater than the preset confidence threshold of the corresponding defect type if there are at least two defects with confidences greater than the preset confidence threshold of the corresponding defect type; A determination module, used for, if there are at least two first defects, obtaining a target defect type of the first defect by calculating a similarity between each of the first defects and a corresponding defect template, wherein the target defect type of the first defect is a defect type corresponding to a defect template having the greatest similarity to the first defect, and taking the target defect type as an output result of the initial image; judging whether a second defect exists among all defects having a confidence greater than a confidence threshold preset for the corresponding defect type; if the second defect exists, making a priority judgment on the target defect type and the defect types of each of the second defects, and taking the defect type corresponding to the defect with the highest priority as the output result of the initial image, wherein the first defect is a defect having a confidence greater than a first preset multiple of a maximum confidence, and the second defect is a defect having a confidence greater than a second preset multiple of the maximum confidence, and the priority is a degree of influence of each defect on the OLED panel. 8. A computer device, characterized in that the computer device comprises a memory and a processor, the memory stores a computer program, and the computer program executes the defect identification method according to any one of claims 1 to 6 when the processor runs. 9. A computer-readable storage medium, characterized in that a computer program is stored on the computer-readable storage medium, and when the computer program is executed by a processor, the defect identification method according to any one of claims 1 to 6 is implemented.