CN117455841A - Panel detection judging method and system - Google Patents
Panel detection judging method and system Download PDFInfo
- Publication number
- CN117455841A CN117455841A CN202311264046.XA CN202311264046A CN117455841A CN 117455841 A CN117455841 A CN 117455841A CN 202311264046 A CN202311264046 A CN 202311264046A CN 117455841 A CN117455841 A CN 117455841A
- Authority
- CN
- China
- Prior art keywords
- picture
- preset
- display panel
- defect type
- judging
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000001514 detection method Methods 0.000 title claims abstract description 52
- 238000000034 method Methods 0.000 title claims abstract description 40
- 230000007547 defect Effects 0.000 claims abstract description 148
- 238000004364 calculation method Methods 0.000 claims abstract description 40
- 238000012216 screening Methods 0.000 claims abstract description 32
- 238000012545 processing Methods 0.000 claims description 33
- 238000000605 extraction Methods 0.000 claims description 19
- 238000001914 filtration Methods 0.000 claims description 15
- 230000003321 amplification Effects 0.000 claims description 8
- 238000003199 nucleic acid amplification method Methods 0.000 claims description 8
- 238000004891 communication Methods 0.000 claims description 3
- 238000004519 manufacturing process Methods 0.000 abstract description 10
- 230000002159 abnormal effect Effects 0.000 description 14
- 238000012790 confirmation Methods 0.000 description 5
- 238000010586 diagram Methods 0.000 description 5
- 238000007689 inspection Methods 0.000 description 3
- 230000003287 optical effect Effects 0.000 description 2
- 238000012552 review Methods 0.000 description 2
- 238000012549 training Methods 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 230000001934 delay Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0004—Industrial image inspection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- 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
- G06N3/0464—Convolutional networks [CNN, ConvNet]
-
- 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
- G06T7/60—Analysis of geometric attributes
- G06T7/64—Analysis of geometric attributes of convexity or concavity
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/25—Determination of region of interest [ROI] or a volume of interest [VOI]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/764—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30108—Industrial image inspection
- G06T2207/30121—CRT, LCD or plasma display
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Computing Systems (AREA)
- General Health & Medical Sciences (AREA)
- Multimedia (AREA)
- Biophysics (AREA)
- Databases & Information Systems (AREA)
- Medical Informatics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biomedical Technology (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Molecular Biology (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Geometry (AREA)
- Quality & Reliability (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a panel detection judging method and a system, which are used for detecting and judging defects of a display panel, wherein the method comprises the steps of detecting the display panel, establishing a picture folder, wherein at least one first picture in the picture folder is an image corresponding to the position of the defect on the display panel, and each first picture is provided with a first identification frame for selecting the position of the corresponding defect; the method comprises the steps of obtaining a picture folder entirely, judging whether a first picture accords with a primary screening condition, wherein the primary screening condition relates to the defect type and the defect quantity; cutting the first picture into a corresponding second picture by taking the first identification frame as the center; and removing the first identification frame in the second picture, extracting the characteristics, obtaining a predicted probability value aiming at the preset defect type from the calculation model, comparing the predicted probability value with a preset probability threshold, and judging that the preset defect type exists in the display panel if the predicted probability value is larger than the probability threshold. The invention can improve the yield and the production efficiency of products.
Description
Technical Field
The invention relates to a panel detection judging method and a system.
Background
With the development of technology, display devices are widely used in many electronic products, such as mobile phones, tablet computers, notebook computers, desktop displays, watches, automobiles, etc.
In the process of manufacturing a display device, it is necessary to detect the display panel used. There may be a Dimple (simple) or a bump (sample) anomaly on the surface of the display panel, if the Dimple or bump is of a size that is within a certain specification, it may be used normally, but when the display panel is detected, for example, using Automated Optical Inspection (AOI), the AOI machine classifies the Dimple or bump as another type of defect (for example, white spot defect) that is not the foregoing Dimple or bump, and the proportion of the other type of defect is significantly increased, because the Dimple/bump affects the optical transmission path where the display panel is located, which results in the anomaly of the detection result of the display panel. To avoid misjudgment of the AOI machine, it is necessary to additionally assign a worker on the production line to perform secondary confirmation for the display panel with the defect (especially, the defect of the foregoing type). However, this requires additional cost for training and manpower arrangement to perform secondary confirmation of the display panel, and also has a high failure rate due to human error.
Disclosure of Invention
In view of the problems in the prior art, the present invention provides a method and a system for detecting and judging a panel.
In order to achieve the above object, the present invention provides a panel detecting and judging method for detecting and judging defects of a display panel, the panel detecting and judging method comprises,
step A, detecting the display panel by utilizing an AOI detection module, and establishing a picture folder corresponding to the display panel, wherein the picture folder is provided with at least one first picture, the at least one first picture is an image corresponding to the position of the defect on the display panel, the position of the corresponding defect is selected by a first identification frame in each first picture, and each first picture is provided with a file name which comprises a defect type;
b, the picture folder is obtained entirely, whether the at least one first picture accords with a primary screening condition is judged, the primary screening condition relates to the defect type and the defect quantity contained in the at least one first picture, and if yes, the step C is executed;
c, cutting each first picture into at least one corresponding second picture by taking the corresponding first identification frame as the center, judging whether the at least one second picture meets a re-screening condition, wherein the re-screening condition relates to the similarity of images in the first identification frame in the at least one second picture, and if so, executing the step D; and
and D, removing the corresponding first identification frame in each second picture, extracting the characteristics, obtaining a predicted probability value aiming at a preset defect type by using a preset calculation model, comparing the predicted probability value with a preset probability threshold, and judging that the preset defect type exists in the display panel if the predicted probability value is larger than the probability threshold.
As an optional technical solution, the file name of each first picture includes a code corresponding to the detected defect type, and in step B, the primary screen condition includes a code that the file name of the at least one first picture does not include other defect types except the preset defect type, and the number of the at least one first picture in the picture folder is not greater than the preset number.
As an alternative technical scheme, on each second picture, at least one second identification frame is arranged around the first identification frame, and a position adjacent to the first identification frame is selected, and the step C further comprises the steps of calculating the distance between the first identification frame and each second identification frame in each second picture to obtain at least one corresponding frame distance, sequentially taking the corresponding second identification frames with the distance larger than a preset distance in the at least one frame distance as comparison frames, judging whether images in each comparison frame and the corresponding first identification frame are similar or not, accumulating the similar times, and the re-screening condition further comprises that the similar times of each second picture do not exceed the preset times.
As an alternative, step D comprises,
step D1, carrying out mean value filtering on each second picture to remove a corresponding first identification frame in each second picture;
step D2, carrying out image amplification on each second picture to obtain at least one corresponding third picture;
step D3, sending the at least one third picture to a residual network for feature extraction, and obtaining a first prediction probability value of the preset defect type from the preset calculation model; and
and D4, comparing the first predicted probability value with the probability threshold value, judging whether the first predicted probability value is larger than the probability threshold value, and if so, judging that the preset defect type exists in the display panel.
As an alternative, step D comprises,
step D1, carrying out mean value filtering on each second picture to remove a corresponding first identification frame in each second picture;
step D2, carrying out image amplification on each second picture to obtain at least one corresponding third picture;
step D3, sending the at least one third picture to a residual network for feature extraction, and obtaining a first prediction probability value of the preset defect type from the preset calculation model;
step D4, comparing the first predicted probability value with the probability threshold value, judging whether the first predicted probability value is larger than the probability threshold value, and if so, executing step D5;
step D5, cutting the at least one first picture into at least one fourth picture with the corresponding first identification frame as the center, wherein each fourth picture is smaller than each second picture;
step D6, sending the at least one fourth picture into the residual network for network feature extraction, and obtaining a second predicted probability value of the preset defect type from the preset calculation model; and
and D7, comparing the second predicted probability value with the probability threshold value, judging whether the second predicted probability value is larger than the probability threshold value, and if so, judging that the preset defect type exists in the display panel.
As an alternative technical scheme, the panel detection judging method further comprises the steps of,
step E, performing IOU 3 Calculating to obtain a calculation result, judging whether the calculation result is larger than a preset result, and if so, judging that the preset defect type exists in the display panel.
In addition, the invention also provides a panel detection and judgment system for detecting and judging the defects of the display panel, wherein the panel detection and judgment system comprises an AOI detection module and a processing module. The AOI detection module is used for detecting the display panel and establishing a picture folder corresponding to the display panel, wherein the picture folder is provided with at least one first picture, the at least one first picture is an image corresponding to the position of the defect on the display panel, the position of the corresponding defect is selected by a first identification frame in each first picture, and each first picture is provided with a file name which comprises a defect type; the processing module is in communication connection with the AOI detection module, and is used for wholly acquiring the picture folder and judging whether the at least one first picture accords with a primary screening condition, wherein the primary screening condition relates to the defect type and the defect quantity contained in the at least one first picture; if yes, cutting the at least one first picture into at least one corresponding second picture by taking each first identification frame as a center, and judging whether the at least one second picture meets a re-screening condition, wherein the re-screening condition relates to the similarity of images in the first identification frames in the at least one second picture; if the predicted probability value is more than the probability threshold, judging that the preset defect type exists in the display panel.
As an optional technical solution, the file name of each first picture includes a code of the detected defect type, the primary screening condition includes a code of the at least one first picture file name that does not include other defect types except the preset defect type, and the number of the at least one first picture in the picture folder is not greater than the preset number.
As an optional technical solution, on each second picture, at least one second identification frame is provided around the first identification frame, and a position adjacent to the first identification frame is selected, the processing module is further configured to calculate a distance between the first identification frame and each second identification frame in each second picture to obtain at least one inter-frame distance, and sequentially use the corresponding second identification frame in the at least one inter-frame distance that is greater than a preset distance as a comparison frame, determine whether images in each comparison frame and the corresponding first identification frame are similar and accumulate the number of similarity times, where the re-screening condition further includes that the number of similarity times of each second picture does not exceed the preset number of similarity times.
As an optional technical scheme, the processing module is further configured to perform mean filtering on each second picture to remove a corresponding first identifier frame in each second picture; amplifying the image of each second picture to obtain at least one corresponding third picture; sending the at least one third picture into a residual network for feature extraction, and obtaining a first prediction probability value of the preset defect type from the preset calculation model; and comparing the first predicted probability value with the probability threshold value, judging whether the first predicted probability value is larger than the probability threshold value, and if so, judging that the display panel has the preset defect type.
As an optional technical scheme, the processing module is further configured to perform mean filtering on each second picture to remove a corresponding first identifier frame in each second picture; amplifying the image of each second picture to obtain at least one corresponding third picture; cutting the at least one first picture into at least one corresponding fourth picture by taking the corresponding first identification frame as the center, wherein each fourth picture is smaller than each second picture; respectively sending the at least one third picture and the at least one fourth picture into a residual network for feature extraction, and respectively obtaining a first prediction probability value and a second prediction probability value corresponding to the preset defect type from the preset calculation model; and comparing the first predicted probability value and the second predicted probability value with the probability threshold respectively, and judging that the preset defect type exists in the display panel if the first predicted probability value and the second predicted probability value are both larger than the probability threshold.
As an optional technical solution, if the predicted probability value is greater than the probability threshold, the processing module is further configured to perform an IOU 3 Calculating to obtain a calculation result, judging whether the calculation result is larger than a preset result, and if so, judging that the preset defect type exists in the display panel.
According to the panel detection judging method and system, after the defect picture of the display panel is screened, the feature extraction is carried out, the predicted probability value aiming at the preset defect type is obtained from the preset calculation model, the predicted probability value is compared with the probability threshold, if the predicted probability value is larger than the probability threshold, the preset defect type of the display panel is judged, the display panel can normally follow current to the subsequent production process, the display panel with the preset defect type is prevented from being misjudged to be abnormal, and the product yield is improved. The detection judging method can automatically finish the filtering of the preset defect types, reduces the labor force for secondary confirmation additionally arranged in the prior art, and improves the production efficiency.
The advantages and spirit of the present invention will be further understood from the following detailed description of the invention and the accompanying drawings.
Drawings
The following describes the embodiments of the present invention in further detail with reference to the drawings.
FIG. 1 is a flow chart of a panel detection and judgment method of the present invention;
FIG. 2 is a schematic diagram of a partial flow of a panel detection and judgment method according to the present invention;
FIG. 3 is a schematic diagram of another partial flow chart of the panel detection and judgment method of the present invention;
FIG. 4 is a schematic diagram of a partial flow chart of a panel detection and judgment method according to the present invention;
fig. 5 is a block diagram of a panel detection and judgment system according to the present invention.
Detailed Description
In order to more clearly illustrate the present invention, the present invention will be further described with reference to preferred embodiments and the accompanying drawings. Like parts in the drawings are denoted by the same reference numerals. It is to be understood by persons skilled in the art that the following detailed description is illustrative and not restrictive, and that this invention is not limited to the details given herein.
Please refer to fig. 1 to 4. Fig. 1 is a flow chart of a panel detection and determination method according to the present invention, fig. 2 is a partial flow chart of the panel detection and determination method according to the present invention, fig. 3 is another partial flow chart of the panel detection and determination method according to the present invention, fig. 4 is a further partial flow chart of the panel detection and determination method according to the present invention, and fig. 5 is a block diagram of a panel detection and determination system according to the present invention.
As shown in fig. 1, the present invention provides a panel detection and judgment method for detecting and judging defects of a display panel, the panel detection and judgment method includes:
step A (S110), detecting a display panel by using an AOI detection module 100, if the display panel is detected to have a defect, establishing a picture folder corresponding to the display panel, wherein the picture folder has at least one first picture, the at least one first picture is an image corresponding to the position of the defect on the display panel, and each first picture is provided with a first identification frame for selecting the position of the corresponding defect, and further, each first picture is provided with a file name, and the file name comprises a defect type;
step B (S120), the processing module 200 obtains the whole picture folder, judges whether the at least one first picture in the picture folder meets the primary screening condition, if so, executes step C, wherein the primary screening condition relates to the defect type and the defect quantity contained in the at least one first picture;
c (S130), cutting each first picture into at least one corresponding second picture by taking the corresponding first identification frame as the center, judging whether the at least one second picture meets a re-screening condition or not, wherein the re-screening condition relates to the similarity of images in the first identification frame in the at least one second picture, and if so, executing the step D; and
and D (S140) removing the corresponding first identification frame in each second picture, extracting the characteristics, obtaining a predicted probability value aiming at a preset defect type from a preset calculation model, comparing the predicted probability value with a preset probability threshold, and judging that the preset defect type exists in the display panel if the predicted probability value is larger than the probability threshold.
In one embodiment, the predetermined defect type is shallow, but not limited to, shallow, or other defect types. In practice, if the size of the dimple or dimple is within a certain specification range, the AOI detection module 100 classifies the dimple/dimple as some other type of defect (such as white point defect) when detecting the display panel. According to the method, after the defect picture of the display panel is subjected to preliminary screening, feature extraction is carried out, a predicted probability value aiming at a preset defect type is obtained from a preset calculation model, the predicted probability value is compared with a probability threshold value, and if the predicted probability value is larger than the probability threshold value, the preset defect type of the display panel is judged. Therefore, the display panel can normally follow current to the subsequent production process, the display panel with the preset defect type is prevented from being misjudged to be abnormal, and the product yield is improved. The detection judging method can automatically finish the filtering of the preset defect types, reduces the labor force for secondary confirmation additionally arranged in the prior art, and improves the production efficiency.
In one embodiment, the predetermined calculation model may be a training classifier or the like stored in the processing module 200 in advance. After removing the corresponding first identification frame in each second picture to extract the characteristics, the second picture enters a preset calculation model, and the pictures/images/characteristics and the like related to the preset defect types can be stored in the preset calculation model in advance, so that the prediction probability value is obtained through comparison analysis.
In actual operation, the AOI detection module 100 is used to detect the display panel to be detected, and in this process, if the AOI detection module 100 detects that a defect exists on the display panel, the AOI detection module photographs the position with the defect on the display panel to form a picture. After the AOI detection module 100 completely saves all the pictures, the processing module 200 performs the process of obtaining the pictures. In one embodiment, the time for the AOI detection module 100 to detect a display panel and save a picture is a preset time (e.g. 12 seconds), and when the processing module 200 detects that a new picture folder is created, the processing module delays the preset time (determining that all the pictures of the AOI detection module 100 have been saved at this time) and then drags the whole picture folder. The picture folder may be associated with the number of the detected display panel or the like, and the picture corresponding to the one-piece product is saved as one picture folder.
In an embodiment, each first picture in the picture folder has a file name, the file name includes a code corresponding to the detected defect type, before the processing module 200 drags the entire picture folder, it is first determined whether there are pictures including the preset defect type in all the first pictures in the entire picture folder, and if so, the entire picture folder is copied.
In an embodiment, the file name of each first picture includes a code corresponding to the detected defect type, and in step B, the primary screen condition includes a code that the file name of the at least one first picture does not include other defect types except the predetermined defect type, and the number of the at least one first picture in the picture folder is not greater than the predetermined number. After the processing module 200 copies the entire picture folder, the pictures of the entire picture folder are pre-processed (i.e., "prescreened"). If the processing module 200 determines that the file name of at least one first picture in the picture folder contains codes of other defect types except the preset defect type, which indicates that the display panel has other defects except the preset defect type, the display panel is determined to be abnormal, and automatic determination processes such as subsequent model determination are not performed, so that manual or/and reinspection of other machines is required. Similarly, if the processing module 200 determines that the number of the first pictures in the picture folder is greater than the preset number (for example, but not limited to, 5 pictures), which indicates that the defect positions on the display panel are more, the display panel is determined to be abnormal, and the automatic determination process such as the subsequent model determination is not performed, but the manual review or the review of other machines is required.
In other embodiments, after the processing module 200 obtains a picture folder corresponding to a display panel, it is again confirmed at a certain time interval whether the corresponding picture folder established by the AOI detecting module 100 has the first picture newly added. If it is determined that there is a newly added first picture in the corresponding picture folder established by the AOI detection module 100 again at a certain interval, the display panel is determined to be abnormal. In another embodiment, if the processing module 200 detects that at least one first picture is damaged (e.g. the picture is incomplete) after obtaining a picture folder corresponding to a display panel, which indicates that the AOI detection module 100 is abnormal when storing the picture, the display panel is determined to be abnormal for reminding, and a worker can check the AOI detection module 100 or the current display panel according to the reminding.
In one embodiment, as shown in fig. 2, step C includes:
and C1, cutting each first picture into at least one corresponding second picture by taking the corresponding first identification frame as the center. Each second picture may have a fixed size. At this time, at least one second picture remains with the first identification frame in the corresponding first picture. Further, each second picture may further have at least one second identification frame around the first identification frame to select a position adjacent to the first identification frame without defects, in addition to the position of the first identification frame to select a corresponding defect.
In an embodiment, the first identification frame has a first color, the second identification frame has a second color, and the first color is different from the second color, for example, the first color is red, and the second color is yellow, but not limited thereto.
And C2, calculating the distance between the first identification frame and each second identification frame in each second picture to obtain at least one corresponding frame-to-frame distance, sequentially taking the corresponding second identification frames which are larger than the preset distance in the at least one frame-to-frame distance as comparison frames, namely judging whether the at least one frame-to-frame distance is larger than the preset distance or not, and taking the corresponding second identification frames as the comparison frames if the judgment result is yes.
In this embodiment, the second identification frame corresponding to the distance between frames greater than the preset distance is selected to perform similarity comparison with the first identification frame. Because the distance between the frames is larger than the preset distance, the false judgment that the first identification frame is similar to the corresponding image in the second identification frame due to smaller distance between the frames can be avoided.
And C3, judging whether the images in each comparison frame and the corresponding first identification frame are similar or not, and accumulating the similar times, wherein the multiple screen conditions comprise that the similar times do not exceed the preset times.
In one embodiment, the Pasteur coefficients may be used to perform similar image matching in image processing. The step C3 includes the steps of,
step C31, counting the Babbitt coefficient for the first identification frame in each second picture and the second identification frame selected as the comparison frame, judging whether the Babbitt coefficient is smaller than a set value, if so, indicating that the images in the second identification frame and the first identification frame selected as the comparison frame are similar, increasing the number of similar times by 1, and the initial value of the similar times can be 0 times; and
and step C32, judging whether the accumulated similar times exceeds the preset times (for example, 2 times) after the corresponding second identification frames with the distances larger than the preset distances among all the frames in each second picture are compared with the first identification frame as comparison frames, and judging that the corresponding second picture is abnormal and does not meet the re-screening condition if the accumulated similar times exceed the preset times (for example, 2 times), wherein the corresponding display panel is abnormal. If the accumulated similar times do not exceed the preset times, the corresponding second picture is judged to be normal, and the subsequent steps can be carried out.
In an embodiment, each first picture has a first identifier, and the step C is further performed before detecting coordinates (e.g., absolute coordinates) of the first identifier on each first picture. If the processing module 200 detects that there are no first identification frames or the number of the first identification frames is excessive in a first picture through coordinate calculation, the first picture is determined to be abnormal, and the corresponding display panel is determined to be abnormal.
In one embodiment, as shown in fig. 3, step D may specifically include,
step D1, carrying out mean value filtering on each second picture to remove a corresponding first identification frame in each second picture;
and D2, carrying out image amplification on each second picture to obtain at least one corresponding third picture. In an embodiment, the image amplification conforming to the production line scene is performed by horizontally turning over, rotating a certain angle, etc. each second picture, so that the obtained at least one third picture is convenient for subsequent comparison, prediction, etc. after the AI model is performed;
and D3, sending the at least one third picture into a residual network for feature extraction, and obtaining a first prediction probability value of the preset defect type from the preset calculation model. The residual network is an artificial neural network model, can reach the depth which is not reached by a general model, and has better model performance effect; and
and D4, comparing the first predicted probability value with a probability threshold value, judging whether the first predicted probability value is larger than the probability threshold value, and if so, judging that the display panel has the preset defect type.
In the above embodiment, the step D performs a model determination, but the method is not limited thereto. In another embodiment, as shown in fig. 4, step D may specifically include,
step D1, carrying out mean value filtering on each second picture to remove a corresponding first identification frame in each second picture;
step D2, carrying out image amplification on each second picture to obtain at least one corresponding third picture;
step D3, sending the at least one third picture to a residual network for feature extraction, and obtaining a first prediction probability value of the preset defect type from the preset calculation model;
step D4, comparing the first predicted probability value with the probability threshold value, judging whether the first predicted probability value is larger than the probability threshold value, and if so, executing step D5; in the present embodiment, the steps D1 to D4 are the first model determination, and the following steps D5 to D7 are the second model determination.
And D5, cutting the at least one second picture (or at least one first picture) into at least one corresponding fourth picture by taking the corresponding first identification frame as the center, wherein the size of the fourth picture is smaller than that of the second picture. The at least one fourth image may also be subjected to image amplification in a similar step D2.
Step D6, sending the at least one fourth picture into the residual error network for network feature extraction, and obtaining a second prediction probability value of the preset defect type from the preset calculation model; and
and D7, comparing the second predicted probability value with the probability threshold value, judging whether the second predicted probability value is larger than the probability threshold value, and if so, judging that the display panel has the preset defect type.
In this embodiment, at least one third picture is sent to a residual network to perform feature extraction, a predicted probability value of a preset defect type is obtained from a preset calculation model, and the predicted probability value is compared with a threshold probability (i.e., first model judgment); if the probability is greater than the threshold probability, cutting out at least one fourth picture with smaller size by taking the first identification frame as a center, introducing the residual network again to perform feature extraction, obtaining a predicted probability value of the preset defect type from a preset calculation model, and comparing the predicted probability value with the threshold probability (namely, judging the model for the second time). And judging that the preset defect type exists in the display panel after the models are judged to be larger than the threshold probability in the two times.
In one embodiment, the method for detecting and judging a panel of the present invention further comprises,
step E (S150), IOU is performed 3 Calculating to obtain a calculation result, judging whether the calculation result is larger than a preset result, and if so, judging that the display panel has the preset defect type. In this embodiment, to further ensure the accuracy of the determination, the IOU is performed after one or two models are undergone to determine that the display panel has the preset defect type 3 Computing, IOU 3 The method is a commonly used index in target detection, and is commonly used for measuring the accuracy degree of the position information of the predicted result in a target detection task. For example, if the IOU is set to 0.05 3 And if the value is more than or equal to 0.05, judging that the information of the preset defect type exists in the first identification frame, and judging that the display panel is abnormal in the preset defect type (for example, simple or sample). If IOU 3 If the value is smaller than 0.05, the first identification frame is considered to have no information of the preset defect type, and the display panel is judged to be abnormal, and manual or/and rechecking of other machines is needed.
As shown in fig. 4, the present invention further provides a panel detection and judgment system 10 for detecting and judging defects of a display panel (not shown). The panel inspection and judgment system 10 includes an AOI inspection module 100 and a processing module 200. The AOI detecting module 100 is configured to detect the display panel, and establish a picture folder corresponding to the display panel, where the picture folder has at least one first picture, the at least one first picture is an image corresponding to a position on the display panel having a defect, and each first picture has a first identification frame, where the position of the corresponding defect is selected, and each first picture has a file name, where the file name includes a defect type. The processing module 200 is in communication connection with the AOI detection module 100, and the processing module 200 is configured to entirely obtain the aforementioned picture folder, determine whether the at least one first picture meets a primary screening condition, where the primary screening condition includes a type and a number of defects included in the at least one first picture; if yes, cutting the at least one first picture into at least one corresponding second picture by taking each first identification frame as a center, and judging whether the at least one second picture meets a double-screen condition or not, wherein the double-screen condition relates to the similarity of images in the first identification frames in the at least one second picture; and if the predicted probability value is more than the probability threshold, judging that the preset defect type exists in the display panel.
The file name of each first picture contains the code of the detected defect type, and the processing module 200 may store the preliminary screening condition in advance. The primary screen condition comprises codes of other defect types except the preset defect type not contained in the file name of at least one first picture, and the number of at least one first picture in the picture folder is not more than the preset number.
In an embodiment, the processing module 200 is further configured to cut at least one first picture into at least one second picture with a fixed size with the corresponding first identification frame as a center; and on each second picture, at least one second identification frame is arranged around the first identification frame, the frame is selected to be adjacent to the position of the first identification frame without defects, the processing module 200 is further used for calculating the distance between the first identification frame and each second identification frame in each second picture to obtain at least one inter-frame distance, the corresponding second identification frames which are larger than the preset distance in the at least one inter-frame distance are sequentially used as comparison frames, whether the images in each comparison frame and the corresponding first identification frame are similar or not is judged, the similarity times are accumulated, and the re-screening condition further comprises that the similarity times of each second picture are not more than the preset times.
In an embodiment, the processing module 200 is further configured to perform mean filtering on each second picture to remove the corresponding first identifier frame in each second picture; amplifying the image of each second picture to obtain at least one corresponding third picture; at least one third picture is sent to a residual network for feature extraction, and a first prediction probability value of a preset defect type is obtained from a preset calculation model; and comparing the first predicted probability value with a probability threshold value, judging whether the first predicted probability value is larger than the probability threshold value, and if so, judging that the display panel has the preset defect type.
In another embodiment, the processing module 200 is further configured to perform mean filtering on each of the second pictures to remove the corresponding first identifier frame in each of the second pictures; amplifying the image of each second picture to obtain at least one corresponding third picture; cutting at least one second picture (or at least one first picture) into at least one corresponding fourth picture by taking the corresponding first identification frame as a center, wherein the size of the fourth picture is smaller than that of the second picture; respectively sending at least one third picture and at least one fourth picture into a residual network for feature extraction, and respectively obtaining a first prediction probability value and a second prediction probability value corresponding to a preset defect type from a preset calculation model; and respectively comparing the first predicted probability value and the second predicted probability value with probability threshold values, and judging that the display panel has preset defect types if the first predicted probability value and the second predicted probability value are both larger than the probability threshold values.
In one embodiment, if the predicted probability value is greater than the probability threshold, the processing module 200 is further configured to perform the IOU 3 Calculating to obtain a calculation result, judging whether the calculation result is larger than a preset result, if so, judging that the display panel has the preset defect type.
According to the panel detection judging method/system, the AI algorithm is used for judging whether the preset defect type exists on the display panel or not instead of human eyes. After screening the defect pictures of the display panel, extracting features, obtaining a predicted probability value aiming at a preset defect type from a preset calculation model, comparing the predicted probability value with a probability threshold, and judging that the preset defect type exists in the display panel if the predicted probability value is larger than the probability threshold. Therefore, the display panel can normally follow current to the subsequent production process, the display panel with the preset defect type is prevented from being misjudged to be abnormal, and the product yield is improved. The detection judging method can automatically finish the filtering of the preset defect types, reduces the labor force for secondary confirmation additionally arranged in the prior art, and improves the production efficiency.
It should be understood that the foregoing examples of the present invention are provided merely for clearly illustrating the present invention and are not intended to limit the embodiments of the present invention, and that various other changes and modifications may be made therein by one skilled in the art without departing from the spirit and scope of the present invention as defined by the appended claims.
Claims (10)
1. A panel detection judging method for detecting and judging defects of a display panel is characterized in that the panel detection judging method comprises the following steps,
step A, detecting the display panel by utilizing an AOI detection module, and establishing a picture folder corresponding to the display panel, wherein the picture folder is provided with at least one first picture, the at least one first picture is an image corresponding to the position of the defect on the display panel, the position of the corresponding defect is selected by a first identification frame in each first picture, and each first picture is provided with a file name which comprises a defect type;
b, the picture folder is obtained entirely, whether the at least one first picture accords with a primary screening condition is judged, the primary screening condition relates to the defect type and the defect quantity contained in the at least one first picture, and if yes, the step C is executed;
c, cutting each first picture into at least one corresponding second picture by taking the corresponding first identification frame as the center, judging whether the at least one second picture meets a re-screening condition, wherein the re-screening condition relates to the similarity of images in the first identification frame in the at least one second picture, and if so, executing the step D; and
and D, removing the corresponding first identification frame in each second picture, extracting the characteristics, obtaining a predicted probability value aiming at a preset defect type by using a preset calculation model, comparing the predicted probability value with a preset probability threshold, and judging that the preset defect type exists in the display panel if the predicted probability value is larger than the probability threshold.
2. The method according to claim 1, wherein the file name of each first picture includes a code corresponding to the detected defect type, in the step B, the primary screen condition includes a code not including other defect types except the predetermined defect type in the file name of the at least one first picture, and the number of the at least one first picture in the picture folder is not greater than the predetermined number.
3. The method according to claim 1 or 2, wherein at least one second identification frame is provided around each second picture, and the frame is selected from the positions adjacent to the first identification frame, step C further comprises calculating a distance between the first identification frame and each second identification frame in each second picture to obtain at least one corresponding frame distance, sequentially using the corresponding second identification frames greater than a predetermined distance in the at least one frame distance as comparison frames, determining whether the images in each comparison frame and the corresponding first identification frame are similar and accumulating the number of similar times, and the multiple screen condition includes that the number of similar times of each second picture does not exceed a predetermined number.
4. The method according to claim 1, wherein the step D comprises,
step D1, carrying out mean value filtering on each second picture to remove a corresponding first identification frame in each second picture;
step D2, carrying out image amplification on each second picture to obtain at least one corresponding third picture;
step D3, sending the at least one third picture to a residual network for feature extraction, and obtaining a first prediction probability value of the preset defect type from the preset calculation model; and
step D4, comparing the first predicted probability value with the probability threshold value, judging whether the first predicted probability value is larger than the probability threshold value, and if yes, judging that the preset defect type exists in the display panel; or,
the step D includes the steps of,
step D1, carrying out mean value filtering on each second picture to remove a corresponding first identification frame in each second picture;
step D2, carrying out image amplification on each second picture to obtain at least one corresponding third picture;
step D3, sending the at least one third picture to a residual network for feature extraction, and obtaining a first prediction probability value of the preset defect type from the preset calculation model;
step D4, comparing the first predicted probability value with the probability threshold value, judging whether the first predicted probability value is larger than the probability threshold value, and if so, executing step D5;
step D5, cutting the at least one first picture into at least one fourth picture with the corresponding first identification frame as the center, wherein each fourth picture is smaller than each second picture;
step D6, sending the at least one fourth picture into the residual network for network feature extraction, and obtaining a second predicted probability value of the preset defect type from the preset calculation model; and
and D7, comparing the second predicted probability value with the probability threshold value, judging whether the second predicted probability value is larger than the probability threshold value, and if so, judging that the preset defect type exists in the display panel.
5. The method for detecting and judging a panel according to claim 1, further comprising,
step E, performing IOU 3 Calculating to obtain a calculation result, judging whether the calculation result is larger than a preset result, and if so, judging that the preset defect type exists in the display panel.
6. A panel detection and judgment system for detecting and judging defects of a display panel is characterized in that the panel detection and judgment system comprises,
the AOI detection module is used for detecting the display panel and establishing a picture folder corresponding to the display panel, wherein the picture folder is provided with at least one first picture, the at least one first picture is an image corresponding to the position with the defect on the display panel, the position with the corresponding defect is selected by a first identification frame in each first picture, and each first picture is provided with a file name which comprises a defect type; and
the processing module is in communication connection with the AOI detection module and is used for wholly acquiring the picture folder and judging whether the at least one first picture accords with a primary screening condition, wherein the primary screening condition relates to the defect type and the defect quantity contained in the at least one first picture; if yes, cutting the at least one first picture into at least one corresponding second picture by taking each first identification frame as a center, and judging whether the at least one second picture meets a re-screening condition, wherein the re-screening condition relates to the similarity of images in the first identification frames in the at least one second picture; if the predicted probability value is more than the probability threshold, judging that the preset defect type exists in the display panel.
7. The system according to claim 6, wherein the file name of each first picture includes a code of the detected defect type, the primary screen condition includes a code of the at least one first picture file name that does not include other defect types than the predetermined defect type, and the number of the at least one first picture in the picture folder is not greater than the predetermined number.
8. The system according to claim 6 or 7, wherein at least one second identification frame is provided around each second picture, and the frame is selected from a defect-free position adjacent to the first identification frame, the processing module is further configured to calculate a distance between the first identification frame and each second identification frame in each second picture to obtain at least one inter-frame distance, and sequentially use the corresponding second identification frame greater than a predetermined distance in the at least one inter-frame distance as a comparison frame, to determine whether images in each comparison frame and the corresponding first identification frame are similar and accumulate the number of similarity times, and the multiple screen condition includes that the number of similarity times of each second picture does not exceed a predetermined number of times.
9. The system of claim 6, wherein the processing module is further configured to perform mean filtering on each of the second pictures to remove a corresponding first identifier frame in each of the second pictures; amplifying the image of each second picture to obtain at least one corresponding third picture; sending the at least one third picture into a residual network for feature extraction, and obtaining a first prediction probability value of the preset defect type from the preset calculation model; comparing the first predicted probability value with the probability threshold value, judging whether the first predicted probability value is larger than the probability threshold value, if so, judging that the display panel has the preset defect type; or,
the processing module is also used for carrying out mean value filtering on each second picture so as to remove the corresponding first identification frame in each second picture; amplifying the image of each second picture to obtain at least one corresponding third picture; cutting the at least one first picture into at least one corresponding fourth picture by taking the corresponding first identification frame as the center, wherein each fourth picture is smaller than each second picture; respectively sending the at least one third picture and the at least one fourth picture into a residual network for feature extraction, and respectively obtaining a first prediction probability value and a second prediction probability value corresponding to the preset defect type from the preset calculation model; and comparing the first predicted probability value and the second predicted probability value with the probability threshold respectively, and judging that the preset defect type exists in the display panel if the first predicted probability value and the second predicted probability value are both larger than the probability threshold.
10. The panel detection decision system of claim 9, wherein the processing module is further configured to perform an IOU if the predicted probability value is greater than the probability threshold 3 Calculating to obtain a calculation result, judging whether the calculation result is larger than a preset result, and if so, judging that the preset defect type exists in the display panel.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202311264046.XA CN117455841A (en) | 2023-09-27 | 2023-09-27 | Panel detection judging method and system |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202311264046.XA CN117455841A (en) | 2023-09-27 | 2023-09-27 | Panel detection judging method and system |
Publications (1)
Publication Number | Publication Date |
---|---|
CN117455841A true CN117455841A (en) | 2024-01-26 |
Family
ID=89586395
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202311264046.XA Pending CN117455841A (en) | 2023-09-27 | 2023-09-27 | Panel detection judging method and system |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN117455841A (en) |
-
2023
- 2023-09-27 CN CN202311264046.XA patent/CN117455841A/en active Pending
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN118150575A (en) | Automatic optical detection method, equipment and deep learning system thereof | |
TWI669519B (en) | Board defect filtering method and device thereof and computer-readabel recording medium | |
US20230011569A1 (en) | Method and apparatus for detecting defect, device, and storage medium | |
CN116485779B (en) | Adaptive wafer defect detection method and device, electronic equipment and storage medium | |
CN115953373B (en) | Glass defect detection method, device, electronic equipment and storage medium | |
CN114841920A (en) | Flame identification method and device based on image processing and electronic equipment | |
US20200090319A1 (en) | Machine learning method implemented in aoi device | |
CN111178445A (en) | Image processing method and device | |
CN108107611B (en) | Self-adaptive defect detection method and device and electronic equipment | |
CN114943841A (en) | Method and device for assisting operation safety control based on image recognition | |
CN115083008A (en) | Moving object detection method, device, equipment and storage medium | |
CN112967224A (en) | Electronic circuit board detection system, method and medium based on artificial intelligence | |
CN109544514B (en) | Sawn timber identity identification method, device and equipment integrating apparent features | |
CN117455841A (en) | Panel detection judging method and system | |
TWI785579B (en) | Automatic model reconstruction method and automatic model reconstruction system for component recognition model | |
CN116542998A (en) | Contour detection method, device, equipment and medium for photoetching film inductance | |
CN110634124A (en) | Method and equipment for area detection | |
CN114897801A (en) | AOI defect detection method, device and equipment and computer medium | |
CN112052727B (en) | Portrait recognition analysis system and method based on big data | |
KR20230024160A (en) | System and method for semantic segmentation learning data error detection and correction | |
US9589331B2 (en) | Method and apparatus for determining a detection of a defective object in an image sequence as a misdetection | |
CN113344858B (en) | Feature detection method, device and computer storage medium | |
CN111935480A (en) | Detection method for image acquisition device and related device | |
CN111366582A (en) | Automatic judgment method based on automatic optical detection | |
CN116612088A (en) | Defect detection method and system for disassembled parts of waste service end |
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 |