CN110567967B - Display panel detection method, system, terminal device and computer readable medium - Google Patents

Display panel detection method, system, terminal device and computer readable medium Download PDF

Info

Publication number
CN110567967B
CN110567967B CN201910770661.5A CN201910770661A CN110567967B CN 110567967 B CN110567967 B CN 110567967B CN 201910770661 A CN201910770661 A CN 201910770661A CN 110567967 B CN110567967 B CN 110567967B
Authority
CN
China
Prior art keywords
convolutional neural
neural network
loss function
display panel
loss
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.)
Active
Application number
CN201910770661.5A
Other languages
Chinese (zh)
Other versions
CN110567967A (en
Inventor
陈春煦
张胜森
郑增强
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Wuhan Jingce Electronic Group Co Ltd
Wuhan Jingli Electronic Technology Co Ltd
Original Assignee
Wuhan Jingce Electronic Group Co Ltd
Wuhan Jingli Electronic Technology Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Wuhan Jingce Electronic Group Co Ltd, Wuhan Jingli Electronic Technology Co Ltd filed Critical Wuhan Jingce Electronic Group Co Ltd
Priority to CN201910770661.5A priority Critical patent/CN110567967B/en
Publication of CN110567967A publication Critical patent/CN110567967A/en
Application granted granted Critical
Publication of CN110567967B publication Critical patent/CN110567967B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
    • G01N21/84—Systems specially adapted for particular applications
    • G01N21/88—Investigating the presence of flaws or contamination
    • G01N21/8851—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
    • G01N21/84—Systems specially adapted for particular applications
    • G01N21/88—Investigating the presence of flaws or contamination
    • G01N21/95—Investigating the presence of flaws or contamination characterised by the material or shape of the object to be examined
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/0002—Inspection of images, e.g. flaw detection
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
    • G01N21/84—Systems specially adapted for particular applications
    • G01N21/88—Investigating the presence of flaws or contamination
    • G01N21/8851—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges
    • G01N2021/8887—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges based on image processing techniques
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
    • G01N21/84—Systems specially adapted for particular applications
    • G01N21/88—Investigating the presence of flaws or contamination
    • G01N21/95—Investigating the presence of flaws or contamination characterised by the material or shape of the object to be examined
    • G01N2021/9513—Liquid crystal panels
    • 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
    • G06T2207/20—Special algorithmic details
    • G06T2207/20084—Artificial neural networks [ANN]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30168—Image quality inspection

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • General Health & Medical Sciences (AREA)
  • Computational Linguistics (AREA)
  • Chemical & Material Sciences (AREA)
  • Evolutionary Computation (AREA)
  • Biophysics (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Biomedical Technology (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Artificial Intelligence (AREA)
  • Data Mining & Analysis (AREA)
  • Analytical Chemistry (AREA)
  • Biochemistry (AREA)
  • Immunology (AREA)
  • Pathology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Quality & Reliability (AREA)
  • Signal Processing (AREA)
  • Testing Of Optical Devices Or Fibers (AREA)
  • Image Analysis (AREA)
  • Control Of Indicators Other Than Cathode Ray Tubes (AREA)

Abstract

The invention discloses a display panel detection method, a display panel detection system, terminal equipment and a computer readable medium, wherein a detection image sample set is generated by collecting detection images of display panels with various quality grades in a preset number, and the detection image sample set is divided into a training set and a verification set according to a preset proportion; calculating the probability of the detection image corresponding to each quality grade by using the characteristic data output by the convolutional neural network, and weighting by using the probability to obtain a predicted value of the quality grade of the detection image; constructing a plurality of convolutional neural network loss functions trained currently, and weighting according to a certain weight proportion to obtain a custom loss function of the convolutional neural network; and adjusting the weight parameters of the convolutional neural network by using the custom loss function trained at present, predicting the image of the display panel to be tested by using the trained convolutional neural network to obtain the quality grade predicted value of the display panel to be tested, and improving the accuracy of the quality grade predicted value of the panel to be tested.

Description

Display panel detection method, system, terminal device and computer readable medium
Technical Field
The invention belongs to the field of display panel detection, and particularly relates to a display panel detection method, a display panel detection system, terminal equipment and a computer readable medium.
Background
With the popularization and rapid update of mobile phones and consumer electronics, the liquid crystal screen and the OLED screen of an industrial production line have great output requirements, and the manufacturing process and the detection technology of the display panel are also rapidly developed. Although the manufacturing process of the panel is promoted year by year, the appearance of the visual defect cannot be completely avoided due to the complicated structure and the various manufacturing processes. According to the defect shape of the panel, three defects can be roughly classified: point defects, line defects, and area defects. Among the three defects, a point defect is the most common defect, which is divided into a bright point and a dark point, and the size of the defect is one pixel; the line defects are also divided into bright lines and dark lines, the size of the defects is a line segment with the width of a whole pixel, and the point defects and the line defects are both electrical defects. The Mura defect is a common surface defect, belongs to a non-electrical defect and is one of the most difficult defects to detect in all panel display defects.
The term "Mura" is from Japanese and means macula and smudge. Mura defects are macroscopic display defects having characteristics of indefinite size, indefinite shape, uneven gray scale distribution, blurred edges, and the like. Generally, the Mura defect is caused by a combination of material overlapping, material unevenness, processing environment variation, and poor quality of the material itself. In the actual production process of the panel, the Mura defect may be generated due to many factors of the manufacturing process including the workshop environment and the like. In the actual production process of the panel, panel manufacturers can sell the panel in a grading way according to the severity of the Mura defect, but at present, the panel grade is judged by training a batch of detection personnel, and the detection personnel perform visual judgment and grading according to experience. On one hand, human eye detection has subjectivity, so that complete consistency of detection standards cannot be guaranteed among different detection personnel; on the other hand, the detection mode has low efficiency and high cost, the time for detecting a 2.4-inch LCD image generally takes even forty seconds, and the detection personnel can have visual fatigue after carrying out long-term detection work and even can hurt the eyes of people. Therefore, a more efficient and more standardized panel quality evaluation method is pursued.
Deep learning, particularly a convolutional neural network in deep learning, has been successful in the fields of image classification, target detection, image semantic segmentation, instance segmentation and the like. At present, the evaluation of the image quality by using the neural network panel is a new field, the image quality grades are distinguished according to the categories by adopting a plurality of modes, but the multi-classification problem cannot describe the association between different categories, and if the panel quality is 1-10 grades, because the association does not exist between the less categories, the 8-grade panel is possibly wrongly divided into 2 grades instead of 7 grades, so that the accuracy of the convolutional neural network-based display panel detection method needs to be improved.
Disclosure of Invention
Aiming at the defects or improvement requirements of the prior art, the invention provides a display panel detection method, a system, terminal equipment and a computer readable medium, wherein the probability of each quality grade corresponding to a detected image is obtained through calculation of a convolutional neural network, the predicted value of the quality grade of the detected image is obtained through weighting, a plurality of loss functions are used for obtaining a custom loss function through weighting, and the weight parameters of the convolutional neural network are adjusted through the custom loss function, so that a trained convolutional neural network is obtained, and the accuracy of the predicted value of the quality grade of the panel to be detected is improved.
To achieve the above object, according to one aspect of the present invention, there is provided a display panel inspection method including the steps of:
calculating the probability of each quality grade corresponding to the detected image sample by using the characteristic data output by the convolutional neural network so as to obtain the quality grade predicted value of the detected image sample;
constructing a plurality of convolutional neural network loss functions by using the quality grade predicted values, weighting according to a preset weight proportion to obtain custom loss functions of the convolutional neural networks, and training the convolutional neural networks by using the custom loss functions;
and acquiring the image of the display panel to be tested, and obtaining the quality grade predicted value of the image of the display panel to be tested by utilizing the trained convolutional neural network.
The method is further improved by acquiring detection images of display panels with various quality grades in a preset number, generating a detection image sample set, and dividing the detection image sample set into a training set and a verification set; and training the convolutional neural network by using the training set, and verifying the convolutional neural network by using the verification set.
As a further improvement of the present invention, the plurality of convolutional neural network loss functions includes a mean loss function, a variance loss function, and a classification loss function.
As a further improvement of the present invention, the probability of detecting each quality level of the image sample is: the feature data for each quality class is divided by the sum of the feature data for all quality classes.
As a further improvement of the present invention, the quality level prediction values of the detected image samples are: the probability of each quality level is summed with the product of its one-to-one corresponding quality level value.
To achieve the above object, according to another aspect of the present invention, there is provided a terminal device comprising at least one processing unit, and at least one memory unit, wherein the memory unit stores a computer program which, when executed by the processing unit, causes the processing unit to perform the steps of the above method.
To achieve the above object, according to another aspect of the present invention, there is provided a computer-readable medium storing a computer program executable by a terminal device, the program, when executed on the terminal device, causing the terminal device to perform the steps of the above method.
To achieve the above object, according to another aspect of the present invention, there is provided a display panel inspection system, which includes an image capturing device, a convolutional neural network module, and a custom loss function obtaining module connected in sequence,
the image capturing device is used for collecting a detection image of the display panel and sending the detection image to the convolutional neural network module;
the image capturing device is used for collecting a detection image of the display panel and sending the detection image to the convolutional neural network module;
the convolutional neural network module is used for calculating the probability of each quality grade corresponding to the detected image sample by utilizing the characteristic data output by the convolutional neural network so as to obtain the quality grade predicted value of the detected image sample; the quality grade prediction value of the display panel image to be tested is obtained by utilizing the trained convolutional neural network;
the custom loss function acquisition module is used for constructing a plurality of convolutional neural network loss functions by using the quality grade predicted values, weighting according to a preset weight proportion to obtain the custom loss functions of the convolutional neural networks, and training the convolutional neural networks by using the custom loss functions.
As a further improvement of the invention, the generation process of the detection image sample set comprises the following steps: and collecting detection images of display panels with various quality grades in a preset number to generate a detection image sample set.
As a further improvement of the present invention, the plurality of convolutional neural network loss functions includes a mean loss function, a variance loss function, and a classification loss function.
Generally, compared with the prior art, the above technical solution conceived by the present invention has the following beneficial effects:
according to the display panel detection method, the system, the terminal device and the computer readable medium, the probability of each quality grade corresponding to the detected image is obtained through the calculation of the convolutional neural network, the predicted value of the quality grade of the detected image is obtained through weighting, the user-defined loss function is obtained through weighting of the loss functions, the weight parameters of the convolutional neural network are adjusted through the user-defined loss function, the trained convolutional neural network is obtained, and the accuracy of the predicted value of the quality grade of the panel to be detected is improved.
According to the display panel detection method, the display panel detection system, the terminal device and the computer readable medium, the custom loss function constructed by the display panel detection method, the custom loss function constructed by the display panel detection system comprises a traditional classification loss function, a mean loss function and a variance loss function, wherein the mean loss function enables the prediction grade of the image quality scoring system to be close to the grade of a real label, the variance loss function penalizes the discreteness of the distribution of the prediction grade, a convolutional neural network is facilitated to obtain a grade prediction value with a small confidence interval and high confidence, the convolutional neural network is guided to be trained by the weighted summation of the three loss functions, and the accuracy of the quality grade prediction value of the panel to be detected can be further improved.
According to the display panel detection method, the display panel detection system, the terminal device and the computer readable medium, the characteristic data of the detection image corresponding to each quality grade is calculated, the probability of each quality grade is calculated, the predicted value of the quality grade is obtained in a weighting mode, the convolutional neural network is facilitated to obtain the grade predicted value with a small confidence interval but high confidence, and the accuracy of the quality grade predicted value of the panel to be detected can be further improved.
Drawings
Fig. 1 is a schematic diagram of a display panel detection method according to an embodiment of the invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
In addition, the technical features involved in the embodiments of the present invention described below may be combined with each other as long as they do not conflict with each other. The present invention will be described in further detail with reference to specific embodiments.
Fig. 1 is a schematic diagram of a display panel detection method according to an embodiment of the invention. As shown in fig. 1, a method for detecting a display panel includes the following steps:
dividing a detection image sample set into a training set and a verification set; training a convolutional neural network by using a training set, calculating the probability of the detection image corresponding to each quality grade by using the characteristic data output by the convolutional neural network, and weighting by using the probability to obtain the predicted value of the quality grade of the detection image;
as a preferred example, acquiring a preset number of detection images of display panels with various quality grades, generating a detection image sample set, and dividing the detection image sample set into a training set and a verification set according to a preset proportion;
specifically, the input of the convolutional neural network is the detection image of the training set, the output is the characteristic data of the detection image corresponding to each quality grade,
the probability of detecting each quality level of the image sample is as follows: dividing the feature data of each quality grade by the sum of the feature data of all quality grades;
as an example, a total number of quality classes of k may be defined,
Figure BDA0002173459600000041
characteristic data of the ith detection image corresponding to the jth quality grade are obtained, j is a natural number, and j is more than or equal to 1 and less than or equal to k;
probability p of ith detection image sample corresponding to jth quality gradeijComprises the following steps:
Figure BDA0002173459600000042
the quality grade prediction value of the measured image sample is as follows: the product of the probability of each quality level and the corresponding quality level value, specifically, the predicted value of the quality level of the i-th detected image is:
Figure BDA0002173459600000051
constructing a plurality of convolutional neural network loss functions trained currently, and weighting according to a certain weight proportion to obtain a custom loss function of the convolutional neural network; adjusting the weight parameters of the convolutional neural network by using the custom loss function trained at present, and verifying by using a verification set to obtain the trained convolutional neural network;
further, the number of samples of the convolutional neural network which is currently trained is defined as N,
obtaining a first loss function of the current training by using a first calculation mode of a convolutional neural network loss function as follows:
Figure BDA0002173459600000052
wherein, yiQuality of the image for the ith test, etcThe true value of the stage;
the first loss function is a mean loss function, and the prediction grade of the picture quality scoring system is close to the grade of the real label by calculating the grade prediction value obtained by prediction probability regression and the mean loss of the real label;
obtaining a second loss function of the current training by using a second calculation mode of the convolutional neural network loss function, wherein the second loss function is as follows:
Figure BDA0002173459600000053
the second loss function is a variance loss function which penalizes the discreteness of prediction grade distribution and is beneficial to obtaining a grade prediction value with small confidence interval but high confidence.
Obtaining a third loss function of the current training by using a third calculation mode of the convolutional neural network loss function, wherein the third loss function is as follows:
Figure BDA0002173459600000054
the third loss function is a classification loss function, which makes the probability value of the predicted true level as large as possible, but the classification loss function cannot describe the association between different classes (for example, the quality of the panel is 1-10 levels, and the classification of panel errors of 8 levels into 7 levels is obviously more acceptable than the classification of panel errors into 2 levels), and the classification loss function treats the panel errors of 8 levels into 7 levels and the classification of panel errors into 2 levels.
Of course, the first to third calculation manners of the neural network loss function may be adjusted according to the calculation requirements, for example, the third loss function may be a metric learning loss function (e.g., centrloss, etc.), and may be used. On the basis of the third loss function, the three loss functions are weighted and summed to guide the neural network to train through the designed mean loss function and variance loss function, and the method for evaluating the quality of the panel picture with high accuracy and end-to-end is realized.
Obtaining a custom loss function of the current training by using the first to third loss functions of the current training:
Figure BDA0002173459600000061
wherein the content of the first and second substances,
Figure BDA0002173459600000062
and
Figure BDA0002173459600000063
weighting coefficients for the first loss function and the second loss function respectively,
Figure BDA0002173459600000064
and
Figure BDA0002173459600000065
the specific value of (2) is an empirical value, and corresponding adjustment can be carried out according to the calculation requirement of the convolutional neural network;
and collecting the image of the display panel to be tested, and predicting the image of the display panel to be tested by using the trained convolutional neural network to obtain a quality grade predicted value of the display panel to be tested.
Specifically, the serial number of the display panel to be tested is defined as s,
Figure BDA0002173459600000066
the characteristic data of the jth quality grade corresponding to the ith display panel to be tested,
probability p that the s-th display panel to be tested is the j-th quality gradesjComprises the following steps:
Figure BDA0002173459600000067
the predicted value of the quality grade of the s-th display panel to be tested is as follows:
Figure BDA0002173459600000068
a terminal device comprising at least one processing unit and at least one memory unit, wherein the memory unit stores a computer program which, when executed by the processing unit, causes the processing unit to carry out the steps of the above-mentioned method.
A computer-readable medium, in which a computer program executable by a terminal device is stored, causes the terminal device to perform the steps of the above-mentioned method when the program is run on the terminal device.
A display panel detection system comprises an image capture device, a convolutional neural network module and a custom loss function acquisition module which are connected in sequence,
the image capturing device is used for collecting a detection image of the display panel and sending the detection image to the convolutional neural network module;
the convolutional neural network module is used for generating a detection image sample set from detection images of display panels with various quality grades in preset quantity, and dividing the detection image sample set into a training set and a verification set according to a preset proportion; training a convolutional neural network by using a training set, calculating the probability of the detection image corresponding to each quality grade by using the characteristic data output by the convolutional neural network, and obtaining a predicted value of the quality grade of the detection image by using probability weighting; adjusting the weight parameters of the convolutional neural network by using the feedback value of the custom loss function acquisition module, and verifying by using a verification set to obtain a trained convolutional neural network; predicting the image of the display panel to be tested by using the trained convolutional neural network to obtain a quality grade predicted value of the display panel to be tested;
the custom loss function acquisition module is used for constructing a plurality of currently trained convolutional neural network loss functions and obtaining custom loss functions of the convolutional neural networks by weighting according to a certain weight proportion; and obtaining the adjustment value of the weight parameter of the convolutional neural network by using the custom loss function of the current training.
The probability of detecting each quality level of the image sample is as follows: dividing the characteristic data of each quality grade by the sum of the characteristic data of all the quality grades;
as an example, it is possible to define the total number of quality levels as k, and the feature data of the ith detection image corresponding to each quality level as Zi＝(zi1,…zij,…,zik)，zijCharacteristic data of the ith detection image corresponding to the jth quality grade are obtained, j is a natural number, and j is more than or equal to 1 and less than or equal to k;
probability p that the ith detected image is the jth quality levelijComprises the following steps:
Figure BDA0002173459600000071
the quality grade prediction value of the measured image sample is as follows: the product of the probability of each quality level and the corresponding quality level value, specifically, the predicted value of the quality level of the i-th detected image is:
Figure BDA0002173459600000072
as a preferred embodiment, the custom loss function of the current training is specifically:
defining the number of samples of the convolutional neural network of the current training as N, and obtaining a first loss function of the current training by using a first calculation mode of the convolutional neural network loss function as follows:
Figure BDA0002173459600000073
wherein, yiTrue value, m, of quality level for the ith test imageiA predicted value of a quality level for an i-th detected image;
obtaining a second loss function of the current training by using a second calculation mode of the convolutional neural network loss function, wherein the second loss function is as follows:
Figure BDA0002173459600000074
obtaining a third loss function of the current training by using a third calculation mode of the convolutional neural network loss function, wherein the third loss function is as follows:
Figure BDA0002173459600000075
obtaining a custom loss function of the current training by using the first to third loss functions of the current training:
Figure BDA0002173459600000076
wherein p isijIs the probability that the ith detected image is the jth quality level, k is the total number of quality levels,
Figure BDA0002173459600000077
and
Figure BDA0002173459600000078
respectively, the weighting coefficients of the first loss function and the second loss function.
As a preferred embodiment, the calculation process of the quality grade prediction value of the display panel to be tested specifically includes:
defining the serial number of the display panel to be tested as s, and the characteristic data corresponding to each quality grade as Zs＝(zs1,…zsj,…,zsk)，zsjThe characteristic data of the jth quality grade corresponding to the ith display panel to be tested is j, which is a natural number and is more than or equal to 1 and less than or equal to k;
probability p of jth quality grade of s to-be-tested display panelsjComprises the following steps:
Figure BDA0002173459600000081
the predicted value of the quality grade of the s-th display panel to be tested is as follows:
Figure BDA0002173459600000082
it will be understood by those skilled in the art that the foregoing is only an exemplary embodiment of the present invention, and is not intended to limit the invention to the particular forms disclosed, since various modifications, substitutions and improvements within the spirit and scope of the invention are possible and within the scope of the appended claims.

Claims (8)

1. A display panel detection method is characterized by comprising the following steps:
calculating the probability of the detection image sample corresponding to each quality grade by using the characteristic data output by the convolutional neural network, and further obtaining the quality grade predicted value of the detection image sample;
constructing a plurality of convolutional neural network loss functions by using the quality grade predicted values, weighting according to a preset weight proportion to obtain a custom loss function of the convolutional neural network, and training the convolutional neural network by using the custom loss function;
acquiring a display panel image to be tested, and obtaining a quality grade predicted value of the display panel image to be tested by using a trained convolutional neural network;
the loss functions of the convolutional neural networks comprise a mean loss function, a variance loss function and a classification loss function, and the custom loss function obtained by weighting according to a preset weight proportion is as follows:
Figure FDA0003587218570000011
wherein L is1As a function of said mean loss, L2Is the loss of variance function, L3In order to be a function of the classification loss,
Figure FDA0003587218570000012
and
Figure FDA0003587218570000013
are respectively the mean loss functionAnd a weighting factor for the variance loss function.
2. The method according to claim 1, wherein a preset number of detection images of display panels of various quality grades are collected to generate a detection image sample set, and the detection image sample set is divided into a training set and a verification set; and training the convolutional neural network by using the training set, and verifying the convolutional neural network by using the verification set.
3. The method according to any one of claims 1-2, wherein the probability of detecting the image sample corresponding to each quality level is: the feature data for each quality class is divided by the sum of the feature data for all quality classes.
4. The method according to claim 3, wherein the quality level prediction value of the detected image sample is: and summing the products of the probability of each quality grade and the quality grade value corresponding to each quality grade.
5. A terminal device, comprising at least one processing unit and at least one memory unit, wherein the memory unit stores a computer program which, when executed by the processing unit, causes the processing unit to carry out the steps of the method according to any one of claims 1 to 4.
6. A computer-readable medium, in which a computer program is stored which is executable by a terminal device, and which, when run on the terminal device, causes the terminal device to carry out the steps of the method as claimed in any one of claims 1 to 4.
7. A display panel detection system comprises an image capturing device, a convolutional neural network module and a custom loss function acquisition module which are connected in sequence,
the image capturing device is used for collecting a detection image of the display panel and sending the detection image to the convolutional neural network module;
the convolutional neural network module is used for calculating the probability of each quality grade corresponding to the detected image sample by utilizing the characteristic data output by the convolutional neural network so as to obtain the quality grade predicted value of the detected image sample; the quality grade prediction value of the image of the display panel to be tested is obtained by utilizing the trained convolutional neural network;
the user-defined loss function acquisition module is used for constructing a plurality of convolutional neural network loss functions by using the quality grade predicted values, obtaining the user-defined loss functions of the convolutional neural networks by weighting according to a preset weight proportion, and training the convolutional neural networks by using the user-defined loss functions;
the loss functions of the convolutional neural networks comprise a mean loss function, a variance loss function and a classification loss function, and the custom loss function obtained by weighting according to a preset weight proportion is as follows:
Figure FDA0003587218570000021
wherein L is1As a function of said mean loss, L2Is the loss of variance function, L3In order to be a function of the classification loss,
Figure FDA0003587218570000022
and
Figure FDA0003587218570000023
the weighting coefficients of the mean loss function and the variance loss function are respectively.
8. The display panel inspection system of claim 7, wherein the set of inspection image samples is generated by: and acquiring detection images of display panels with various quality grades in a preset number to generate a detection image sample set.
CN201910770661.5A 2019-08-20 2019-08-20 Display panel detection method, system, terminal device and computer readable medium Active CN110567967B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201910770661.5A CN110567967B (en) 2019-08-20 2019-08-20 Display panel detection method, system, terminal device and computer readable medium

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201910770661.5A CN110567967B (en) 2019-08-20 2019-08-20 Display panel detection method, system, terminal device and computer readable medium

Publications (2)

Publication Number Publication Date
CN110567967A CN110567967A (en) 2019-12-13
CN110567967B true CN110567967B (en) 2022-06-17

Family

ID=68774130

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201910770661.5A Active CN110567967B (en) 2019-08-20 2019-08-20 Display panel detection method, system, terminal device and computer readable medium

Country Status (1)

Country Link
CN (1) CN110567967B (en)

Families Citing this family (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111553439A (en) * 2020-05-08 2020-08-18 视睿（杭州）信息科技有限公司 Agricultural product quality classification method and device, computer equipment and storage medium
CN112070762A (en) * 2020-09-18 2020-12-11 惠州高视科技有限公司 Mura defect detection method and device for liquid crystal panel, storage medium and terminal
CN116823015A (en) * 2021-10-30 2023-09-29 深圳曦华科技有限公司 Display function test method of display screen, related device, storage medium and program

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108171707A (en) * 2018-01-23 2018-06-15 武汉精测电子集团股份有限公司 A kind of Mura defects level evaluation method and device based on deep learning
CN108596226A (en) * 2018-04-12 2018-09-28 武汉精测电子集团股份有限公司 A kind of defects of display panel training method and system based on deep learning
CN109376786A (en) * 2018-10-31 2019-02-22 中国科学院深圳先进技术研究院 A kind of image classification method, device, terminal device and readable storage medium storing program for executing
CN109583496A (en) * 2018-11-28 2019-04-05 武汉精立电子技术有限公司 A kind of network model and method for the classification of display panel large area defect
CN110033035A (en) * 2019-04-04 2019-07-19 武汉精立电子技术有限公司 A kind of AOI defect classification method and device based on intensified learning

Family Cites Families (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109034281A (en) * 2018-07-18 2018-12-18 中国科学院半导体研究所 The Chinese handwritten body based on convolutional neural networks is accelerated to know method for distinguishing
CN109902722A (en) * 2019-01-28 2019-06-18 北京奇艺世纪科技有限公司 Classifier, neural network model training method, data processing equipment and medium

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108171707A (en) * 2018-01-23 2018-06-15 武汉精测电子集团股份有限公司 A kind of Mura defects level evaluation method and device based on deep learning
CN108596226A (en) * 2018-04-12 2018-09-28 武汉精测电子集团股份有限公司 A kind of defects of display panel training method and system based on deep learning
CN109376786A (en) * 2018-10-31 2019-02-22 中国科学院深圳先进技术研究院 A kind of image classification method, device, terminal device and readable storage medium storing program for executing
CN109583496A (en) * 2018-11-28 2019-04-05 武汉精立电子技术有限公司 A kind of network model and method for the classification of display panel large area defect
CN110033035A (en) * 2019-04-04 2019-07-19 武汉精立电子技术有限公司 A kind of AOI defect classification method and device based on intensified learning

Also Published As

Publication number Publication date
CN110567967A (en) 2019-12-13

Similar Documents

Publication Publication Date Title
CN110567967B (en) Display panel detection method, system, terminal device and computer readable medium
CN106875373B (en) Mobile phone screen MURA defect detection method based on convolutional neural network pruning algorithm
CN108665460B (en) Image quality evaluation method based on combined neural network and classified neural network
KR20190075707A (en) Method for sorting products using deep learning
CN108664840A (en) Image-recognizing method and device
CN111242123A (en) Power equipment fault diagnosis method based on infrared image
CN104252056B (en) The detection method and device of a kind of substrate
CN109142367A (en) A kind of lithium battery pole ear rosin joint detection method and tab welding detection system
CN109035226B (en) Mura defect detection method based on LSTM model
CN110020691B (en) Liquid crystal screen defect detection method based on convolutional neural network impedance type training
CN102456142A (en) Analysis method for smoke blackness based on computer vision
CN113869162A (en) Violation identification method and system based on artificial intelligence
CN109142368A (en) Lithium battery pole slice is crumpled detection method and tab welding detection system
CN113592828A (en) Nondestructive testing method and system based on industrial endoscope
CN115131283A (en) Defect detection and model training method, device, equipment and medium for target object
CN114066857A (en) Infrared image quality evaluation method and device, electronic equipment and readable storage medium
CN110751170A (en) Panel quality detection method, system, terminal device and computer readable medium
CN111896549A (en) Building crack monitoring system and method based on machine learning
CN112070762A (en) Mura defect detection method and device for liquid crystal panel, storage medium and terminal
CN117392042A (en) Defect detection method, defect detection apparatus, and storage medium
CN112653884B (en) Method for evaluating quality of screen
CN113902710A (en) Method and system for detecting surface defects of industrial parts based on anomaly detection algorithm
KR102189951B1 (en) System and method for inspection of ship painting condition using image analysis
CN115984158A (en) Defect analysis method and device, electronic equipment and computer readable storage medium
CN109187547A (en) Lithium battery pole ear solder joint welds broken detection method and tab welding detection system

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
GR01 Patent grant
GR01 Patent grant