CN112561904A - Method and system for reducing false detection rate of AOI (argon oxygen decarburization) defects on display screen appearance - Google Patents
Method and system for reducing false detection rate of AOI (argon oxygen decarburization) defects on display screen appearance Download PDFInfo
- Publication number
- CN112561904A CN112561904A CN202011548782.4A CN202011548782A CN112561904A CN 112561904 A CN112561904 A CN 112561904A CN 202011548782 A CN202011548782 A CN 202011548782A CN 112561904 A CN112561904 A CN 112561904A
- Authority
- CN
- China
- Prior art keywords
- image
- display screen
- defect
- appearance
- depth measurement
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 230000007547 defect Effects 0.000 title claims abstract description 161
- 238000000034 method Methods 0.000 title claims abstract description 52
- 238000001514 detection method Methods 0.000 title claims abstract description 42
- VVTSZOCINPYFDP-UHFFFAOYSA-N [O].[Ar] Chemical compound [O].[Ar] VVTSZOCINPYFDP-UHFFFAOYSA-N 0.000 title abstract description 6
- 238000005261 decarburization Methods 0.000 title abstract description 6
- 238000005259 measurement Methods 0.000 claims abstract description 57
- 239000004744 fabric Substances 0.000 claims description 17
- 238000012549 training Methods 0.000 claims description 14
- LFQSCWFLJHTTHZ-UHFFFAOYSA-N Ethanol Chemical compound CCO LFQSCWFLJHTTHZ-UHFFFAOYSA-N 0.000 claims description 12
- 239000000428 dust Substances 0.000 claims description 10
- 230000000694 effects Effects 0.000 claims description 7
- 230000011218 segmentation Effects 0.000 claims description 6
- 230000003321 amplification Effects 0.000 claims description 3
- 238000003199 nucleic acid amplification method Methods 0.000 claims description 3
- 238000011176 pooling Methods 0.000 claims description 3
- 230000002194 synthesizing effect Effects 0.000 claims description 2
- 238000010586 diagram Methods 0.000 description 5
- 238000012545 processing Methods 0.000 description 4
- 238000012216 screening Methods 0.000 description 4
- 238000004458 analytical method Methods 0.000 description 3
- 230000003287 optical effect Effects 0.000 description 3
- 238000013528 artificial neural network Methods 0.000 description 2
- 239000012535 impurity Substances 0.000 description 2
- 238000004519 manufacturing process Methods 0.000 description 2
- 238000003672 processing method Methods 0.000 description 2
- 241001112258 Moca Species 0.000 description 1
- VYPSYNLAJGMNEJ-UHFFFAOYSA-N Silicium dioxide Chemical compound O=[Si]=O VYPSYNLAJGMNEJ-UHFFFAOYSA-N 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000015556 catabolic process Effects 0.000 description 1
- 230000007797 corrosion Effects 0.000 description 1
- 238000005260 corrosion Methods 0.000 description 1
- 238000013434 data augmentation Methods 0.000 description 1
- 238000006731 degradation reaction Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- -1 fingerprints Substances 0.000 description 1
- 238000005187 foaming Methods 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 239000011521 glass Substances 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 239000000741 silica gel Substances 0.000 description 1
- 229910002027 silica gel Inorganic materials 0.000 description 1
- 238000002791 soaking Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0004—Industrial image inspection
- G06T7/001—Industrial image inspection using an image reference approach
-
- 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
- G06T2207/30108—Industrial image inspection
- G06T2207/30121—CRT, LCD or plasma display
Landscapes
- Engineering & Computer Science (AREA)
- Quality & Reliability (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Investigating Materials By The Use Of Optical Means Adapted For Particular Applications (AREA)
Abstract
The application provides a method and a system for reducing the false detection rate of AOI (argon oxygen decarburization) defects on the appearance of a display screen, wherein the method comprises the following steps: acquiring a first image of the appearance of a display screen to be detected, and extracting a suspected defect image and position information according to the first image; wiping the display screen to be detected, and acquiring a second image of the appearance of the wiped display screen to be detected; extracting an image at a position corresponding to the second image according to the position information of the suspected defect image to obtain a re-judgment image; judging the similarity of the suspected defect image and the re-judged image through a depth measurement similarity judgment model; and if the similarity result of the suspected defect image and the re-judgment image is 1, judging the suspected defect image as a defect, and outputting the position information of the suspected defect image. The method and the system can effectively distinguish the surface interference of the display screen from the surface defects of the display screen, reduce the defect false detection rate of the display screen appearance AOI system, and improve the accuracy and the efficiency of the defect detection of the display screen appearance AOI system.
Description
Technical Field
The invention belongs to the technical field of display screen appearance AOI detection, and particularly relates to a method and a system for reducing the display screen appearance AOI defect false detection rate.
Background
An Automatic Optical Inspection (AOI) system for the appearance of a display screen detects appearance defects of the display screen based on an Optical principle. In the production process of the display screen, the display screen is influenced by a processing technology and a production workshop environment, and the surface of the display screen is very easy to be polluted by interference such as dust, fingerprints, oil stains and the like. The AOI system is difficult to clearly identify the interference and the real defects, and the surface interference of dust, oil stains and the like is easily judged as the defects in the detection process, so that the false detection conditions of overdischarge, overdetection, false judgment and the like occur.
In order to reduce surface interference such as dust, oil stain and the like, a combined mode of 'direct shooting and oblique shooting' is usually adopted to acquire a detection image. As shown in figure 1, namely, one camera is opposite to the screen, and the other camera and the screen form a certain angle, and each camera collects an image, the method can output the position information of the display screen where dust, surface dirt and bonding defects are located to the image, and then the difference of the position information of the objects to be judged of the two images is compared by using a simple image processing means for distinguishing.
However, the above method can only distinguish the surface interference of the display screen from the internal defects of the display screen, but cannot distinguish the surface interference of the display screen from the surface defects of the display screen, such as scratches, surface unevenness and other defects. Moreover, the CG glass of the display screen is thinner and thinner, the height difference information of the defects and the interference is finer and finer, the distinguishing difficulty by utilizing the image position information is larger and larger, the accuracy is lower and lower, and the defect false detection rate of the AOI system of the appearance of the display screen is still higher.
Disclosure of Invention
The application provides a method and a system for reducing the false detection rate of AOI (argon oxygen decarburization) defects on the appearance of a display screen. The problem that the defect false detection rate of an existing display screen appearance AOI system is high due to the fact that the existing display screen appearance AOI detection system is difficult to distinguish display screen surface interference and display screen surface defects is solved.
In one aspect, the present application provides a method for reducing an AOI defect false-detection rate of an appearance of a display screen, including:
acquiring a first image of the appearance of a display screen to be detected, and extracting a suspected defect image and position information according to the first image;
wiping the display screen to be detected;
acquiring a second image of the appearance of the display screen to be detected after wiping, and extracting an image of a position corresponding to the second image according to the position information of the suspected defect image to obtain a re-judgment image;
judging the similarity of the suspected defect image and the re-judged image through a depth measurement similarity judgment model;
and if the similarity result between the suspected defect image and the re-judged image is 1, judging the image as a defect, and outputting the position information of the suspected defect image.
Optionally, the step of determining the similarity between the suspected defect image and the re-determined image by using a depth metric similarity determination model includes:
constructing a depth measurement learning model through a Deepsiamese network;
training the depth measurement learning model according to the positive sample set and the negative sample set to obtain the depth measurement similarity judgment model;
and inputting the suspected defect image and the re-judgment image into the depth measurement similarity judgment model, and calculating by the depth measurement similarity judgment model to obtain a similarity result.
Optionally, the depth metric learning model backbone network includes 4 convolutional layers, 2 pooling layers, and 2 full-link layers, and the data input layer of the depth metric learning model includes six-channel data pairs synthesized by two three-channel images, where three channels input the suspected defect image, and the other three channels input the re-judgment image.
Optionally, the positive sample set includes a { defect sample image, suspected defect sample image } data pair, and the negative sample set includes a { interference sample image, suspected defect sample image } data pair, where the interference image is an image of a display screen with dust and/or fingerprints and/or oil stains.
Optionally, the step of training the depth metric learning model according to the positive sample set and the negative sample set further includes: and performing data amplification on the positive sample set and the negative sample set, and respectively obtaining an original image of a sample image, a brightness change image of the sample image and a contrast change image of the sample image.
Optionally, the method for extracting the suspected defect image and the position information according to the first image adopts a dynamic threshold segmentation method.
In another aspect, the present application further provides a system for reducing the display screen appearance AOI defect false detection rate, which is used for executing the method described above, and is characterized by comprising an image acquisition device, a wiping device, and a processor connected with the image acquisition device and the wiping device;
the image acquisition device is used for acquiring a first image of the appearance of the display screen to be detected and a second image of the appearance of the display screen to be detected after wiping;
the wiping device is used for wiping the display screen to be detected;
the processor is used for extracting a suspected defect image and position information according to the first image; extracting an image at a position corresponding to the second image according to the position information of the suspected defect image to obtain a re-judgment image; judging the similarity of the suspected defect image and the re-judged image through a depth measurement similarity judgment model; and outputting the position information of the suspected defect image.
Optionally, the processor includes an offline module and an online module;
wherein the offline module is to: constructing a depth measurement learning model through a Deepsiamese network; training the depth measurement learning model according to the positive sample set and the negative sample set to obtain the depth measurement similarity judgment model;
the online module is used for: and inputting the suspected defect image and the re-judgment image into the depth measurement similarity judgment model, and calculating by the depth measurement similarity judgment model to obtain a similarity result.
Optionally, the wiping device comprises: the wet wiping elastic pressure head is provided with dustless cloth soaked in alcohol, and the dry wiping elastic pressure head is provided with dry dustless cloth.
Optionally, the image capturing device includes: high resolution industrial cameras, multi-angle light sources or three-dimensional effect light sources.
Optionally, the processor includes an offline module and an online module; wherein the offline module is to: constructing a depth measurement learning model through a Deepsiamese network; training the depth measurement learning model according to the positive sample set and the negative sample set to obtain the depth measurement similarity judgment model; the online module is used for: inputting the suspected defect image and the re-judgment image into the depth measurement similarity judgment model, and calculating by the depth measurement similarity judgment model to obtain a similarity result;
the wiping device includes: the wet wiping elastic pressure head is provided with dustless cloth soaked with alcohol, and the dry wiping elastic pressure head is provided with dry dustless cloth;
the image acquisition device includes: high resolution industrial cameras, multi-angle light sources or three-dimensional effect light sources.
According to the technical scheme, the application provides a method and a system for reducing the error detection rate of AOI (argon oxygen decarburization) defects on the appearance of a display screen, wherein the method comprises the following steps: acquiring a first image of the appearance of a display screen to be detected, and extracting a suspected defect image and position information according to the first image; wiping the display screen to be detected, and acquiring a second image of the appearance of the wiped display screen to be detected; extracting an image at a position corresponding to the second image according to the position information of the suspected defect image to obtain a re-judgment image; judging the similarity of the suspected defect image and the re-judged image through a depth measurement similarity judgment model; and if the similarity result between the suspected defect image and the re-judged image is 1, judging the image as a defect, and outputting the position information of the suspected defect image.
The method of primary screening, wiping and re-judging is adopted in the application. Firstly, extracting suspected defect images through preliminary screening, and then simulating manual work to wipe a display screen; and finally, acquiring the re-judged image after wiping, carrying out similarity judgment on the re-judged image and the suspected defect image to judge whether the image is a real defect, and constructing a model for similarity judgment on the basis of a Deepsiamese network through a depth measurement learning method. The method and the system for reducing the error detection rate of the AOI defects on the appearance of the display screen can effectively distinguish the surface interference of the display screen from the surface defects of the display screen, reduce the error detection rate of the AOI defects on the appearance of the display screen, and improve the accuracy and the efficiency of the AOI defects on the appearance of the display screen.
Drawings
In order to more clearly explain the technical solution of the present application, the drawings needed to be used in the embodiments will be briefly described below, and it is obvious to those skilled in the art that other drawings can be obtained according to the drawings without creative efforts.
FIG. 1 is a schematic structural diagram of a conventional display screen appearance AOI defect detection apparatus;
FIG. 2 is a flowchart illustrating an embodiment of a method for reducing an AOI defect false detection rate of an appearance of a display screen according to the present disclosure;
fig. 3 is a flowchart of an embodiment of a step of determining similarity between the suspected defect image and the re-determined image through a depth measure similarity determination model in the method for reducing an AOI defect false detection rate of a display screen according to the present application;
fig. 4 is a schematic structural diagram of an embodiment of a deepsiamesee network of a deep metric learning model in the method for reducing the display screen appearance AOI defect false detection rate provided by the present application;
fig. 5 is a schematic structural diagram of an embodiment of a system for reducing an AOI defect false-detection rate of an appearance of a display screen according to the present application.
Detailed Description
Reference will now be made in detail to embodiments, examples of which are illustrated in the accompanying drawings. When the following description refers to the accompanying drawings, like numbers in different drawings represent the same or similar elements unless otherwise indicated. The embodiments described in the following examples do not represent all embodiments consistent with the present application. But merely as exemplifications of systems and methods consistent with certain aspects of the application, as recited in the claims.
Referring to fig. 2, fig. 2 is a flowchart illustrating an embodiment of a method for reducing an AOI defect false-detection rate of an appearance of a display screen according to the present application. In one aspect, the present application provides a method for reducing an AOI defect false-detection rate of an appearance of a display screen, including:
s1: the method comprises the steps of obtaining a first image of the appearance of a display screen to be detected, and extracting a suspected defect image and position information according to the first image. Firstly, defects of a display screen to be detected are preliminarily screened, a suspected defect image is extracted from a first image according to an image processing algorithm, the suspected defect image comprises real defects or interferences such as dust, fingerprints and oil stains, and further analysis and judgment are needed.
The position information is convenient for extracting the re-judgment image, and when the suspected defect is determined to be a real defect, the position information can be quickly positioned, and the position of the real defect can be found for replacement or maintenance. The positioning method of the position information includes but is not limited to: the coordinate value of the upper left corner of the suspected defect image is obtained as position information, the coordinate value of the center point of the suspected defect image is obtained as position information, and the like.
S2: and wiping the display screen to be detected. In this embodiment, a mode of manually wiping the display screen is simulated, and after the suspected defect image is preliminarily screened and extracted from the display screen to be detected, the display screen to be detected is wiped. The wiping method comprises the following steps: firstly, the display screen is wiped and cleaned by soaking high-purity alcohol in dust-free cloth, so that dirt and dust are wiped off or dissolved and separated from the display screen. Then wiping with clean dust-free cloth, and wiping off the alcohol dissolved with impurities with the dry dust-free cloth to ensure that no residual alcohol is left on the screen. The interference of the display screen can be effectively removed through wiping, and the screen is cleaned.
S3: and acquiring a second image of the appearance of the display screen to be detected after wiping, and extracting an image of a corresponding position of the second image according to the position information of the suspected defect image to obtain a re-judgment image. And the re-judgment image is an image of the corresponding position of the suspected defect image after being wiped, and is compared with the suspected defect image to judge whether the suspected defect is a real defect.
S4: and judging the similarity of the suspected defect image and the re-judged image through a depth measurement similarity judgment model.
Referring to fig. 3, fig. 3 is a flowchart illustrating a step of determining similarity between the suspected defect image and the re-determined image according to a depth measure similarity determination model in the method for reducing an AOI defect false detection rate of a display screen according to an embodiment of the present disclosure. Optionally, in some embodiments, the method specifically includes:
s41: and constructing a depth metric learning model through a Deepsiamese network. The DeepSiamese network is a convolutional network structure composed of a set of twin structures sharing weights. A DeepSiamese network may be used to measure the degree of similarity of two inputs. The twin neural network has two inputs that are entered into two neural networks that respectively map the inputs to new spaces, forming representations of the inputs in the new spaces. Through calculation, the similarity of the two inputs is evaluated.
Referring to fig. 4, fig. 4 is a schematic structural diagram of an embodiment of a DeepSiamese network of a deep metric learning model in the method for reducing the display screen appearance AOI defect false detection rate according to the present application.
Optionally, the depth metric learning model is constructed by a DeepSiamese network, the DeepSiamese network constructing a backbone network including 4 convolutional layers Convol1_1, Convol1_2, Convol2_1, Convol2_ 2; 2 layers of pooling layers Pool1, Pool 2; the depth measurement learning model comprises 2 layers of full connection layers FC6 and FC7, wherein the data input layer of the Deepsiamase network of the depth measurement learning model comprises six-channel data pairs formed by synthesizing two three-channel images, the suspected defect image is input into three channels, and the repeated judgment image is input into the other three channels. The loss function uses the contrast loss.
S42: and training the depth measurement learning model according to the positive sample set and the negative sample set to obtain the depth measurement similarity judgment model.
Optionally, the positive sample set includes a { defect sample image, suspected defect sample image } data pair, and the negative sample set includes a { interference sample image, suspected defect sample image } data pair, where the interference image is an image of a display screen with dust and/or fingerprints and/or oil stains. In this embodiment, to improve the accuracy of the training of the depth metric learning model, the positive sample set and the negative sample set each include more than ten thousand sets of data pairs.
Optionally, the step of training the depth metric learning model according to the positive sample set and the negative sample set further includes: and performing data amplification on the positive sample set and the negative sample set, and respectively obtaining an original image of a sample image, a brightness change image of the sample image and a contrast change image of the sample image.
In the embodiment, due to the fact that optical system degradation and instability factors exist in an industrial scene, the situation that the image brightness difference is large and the contrast difference is large before and after wiping easily occurs, the brightness change and the contrast change factors are directly displayed and applied to a trained positive sample set and a trained negative sample set as data augmentation modes, more prior knowledge is given to the deep metric learning model Deepsiamase network, and rapid convergence is facilitated.
Specifically, the images of the positive sample set input into the six channels of the depth metric learning model are respectively: the image processing method comprises the steps of obtaining an original image of a defect sample image, a brightness change image of the defect sample image, a contrast change image of the defect sample image, an original image of a suspected defect sample image, a brightness change image of the suspected defect sample image and a contrast change image of the suspected defect sample image.
The images of the negative sample set input into the six channels of the depth metric learning model are respectively as follows: the image processing method comprises the steps of obtaining an original image of an interference sample image, a brightness change image of the interference sample image, a contrast change image of the interference sample image, an original image of a suspected defect sample image, a brightness change image of the suspected defect sample image and a contrast change image of the suspected defect sample image.
S43: and inputting the suspected defect image and the re-judgment image into the depth measurement similarity judgment model, and calculating by the depth measurement similarity judgment model to obtain a similarity result.
S5: and if the similarity result between the suspected defect image and the re-judged image is 1, judging the image as a defect, and outputting the position information of the suspected defect image. If the similarity result is 0, the interference is determined.
Optionally, the method for extracting the suspected defect image and the position information according to the first image adopts a dynamic threshold segmentation method. The dynamic threshold segmentation method is used for segmenting an object from a background according to the difference between the current pixel point and the local neighborhood point of the image, so that the dynamic threshold segmentation method can detect the defect that the size of 1 pixel is minimum and the contrast is 1 gray value. In this embodiment, a suspected defect image including suspected defects with a size of 64 × 64 pixels is extracted by a dynamic threshold segmentation method.
Referring to fig. 5, fig. 5 is a schematic structural diagram of an embodiment of a system for reducing an AOI defect false-detection rate of an appearance of a display screen according to the present application.
In another aspect, the present application further provides a system for reducing the false detection rate of AOI defects on the appearance of a display screen, which is used for executing the method described above, and is characterized by comprising an image acquisition device 1, a wiping device 2, and a processor 3 connected with the image acquisition device 1 and the wiping device 2;
the image acquisition device 1 is used for acquiring a first image of the appearance of a display screen to be detected and a second image of the appearance of the display screen to be detected after wiping;
the wiping device 2 is used for wiping the display screen to be detected;
the processor 3 is configured to extract a suspected defect image and position information according to the first image; extracting an image at a position corresponding to the second image according to the position information of the suspected defect image to obtain a re-judgment image; judging the similarity of the suspected defect image and the re-judged image through a depth measurement similarity judgment model; and outputting the position information of the suspected defect image.
Optionally, the processor 3 includes an offline module and an online module.
Wherein the offline module is to: constructing a depth measurement learning model through a Deepsiamese network; training the depth measurement learning model according to the positive sample set and the negative sample set to obtain the depth measurement similarity judgment model;
the online module is used for: and inputting the suspected defect image and the re-judgment image into the depth measurement similarity judgment model, and calculating by the depth measurement similarity judgment model to obtain a similarity result.
In this embodiment, the offline module is mainly used for training the depth metric learning model, and provides a model for real-time processing and analysis of the online module and updates the model, so as to improve the generalization performance of the model and improve the recognition rate. The on-line module analyzes and processes the collected images in real time by using the trained model and outputs a similarity analysis result.
Optionally, the wiping device 2 comprises: the wet wiping elastic pressure head is provided with dustless cloth soaked in alcohol, and the dry wiping elastic pressure head is provided with dry dustless cloth. The display screen is firstly wiped and cleaned by using the wet wiping elastic pressure head, so that dirt and dust are wiped off or dissolved and separated from the display screen. Then, the display screen is wiped by using a dry wiping elastic pressure head, and the dry dust-free cloth wipes off the alcohol dissolved with impurities, so that no residual alcohol is left on the screen.
In this embodiment, wet wipe the elasticity pressure head with dry wipe the elasticity pressure head and adopt the softness and have certain resistant alcohol corrosion's foaming silica gel material, ensure that dustless cloth can wipe the vast majority on display screen 2.5D arc limit under elasticity pressure head pressure, improve the area that the display screen effectively cleaned.
Optionally, the image capturing apparatus 1 includes: high resolution industrial cameras, multi-angle light sources or three-dimensional effect light sources. In this embodiment, the high-resolution industrial camera may adopt a 16K black-and-white camera to obtain a high-resolution image, so as to improve the accuracy of the similarity determination between the suspected defect image and the re-judgment image. The multi-angle light source and the three-dimensional effect light source can provide multi-angle and multi-azimuth light for the display screen, so that the imaging of the display screen is clearer and more real, and the accuracy of similarity judgment is further improved.
Optionally, the MoCA device may further include the following components in the above embodiment: the processor comprises an offline module and an online module; wherein the offline module is to: constructing a depth measurement learning model through a Deepsiamese network; training the depth measurement learning model according to the positive sample set and the negative sample set to obtain the depth measurement similarity judgment model; the online module is used for: inputting the suspected defect image and the re-judgment image into the depth measurement similarity judgment model, and calculating by the depth measurement similarity judgment model to obtain a similarity result;
the wiping device includes: the wet wiping elastic pressure head is provided with dustless cloth soaked with alcohol, and the dry wiping elastic pressure head is provided with dry dustless cloth;
the image acquisition device includes: high resolution industrial cameras, multi-angle light sources or three-dimensional effect light sources.
According to the technical scheme, the application provides a method and a system for reducing the error detection rate of AOI (argon oxygen decarburization) defects on the appearance of a display screen, wherein the method comprises the following steps: acquiring a first image of the appearance of a display screen to be detected, and extracting a suspected defect image and position information according to the first image; wiping the display screen to be detected, and acquiring a second image of the appearance of the wiped display screen to be detected; extracting an image at a position corresponding to the second image according to the position information of the suspected defect image to obtain a re-judgment image; judging the similarity of the suspected defect image and the re-judged image through a depth measurement similarity judgment model; and if the similarity result between the suspected defect image and the re-judged image is 1, judging the image as a defect, and outputting the position information of the suspected defect image.
The method of primary screening, wiping and re-judging is adopted in the application. Firstly, extracting suspected defect images through preliminary screening, and then simulating manual work to wipe a display screen; and finally, acquiring the re-judged image after wiping, carrying out similarity judgment on the re-judged image and the suspected defect image to judge whether the image is a real defect, and constructing a model for similarity judgment on the basis of a Deepsiamese network through a depth measurement learning method. The method and the system for reducing the error detection rate of the AOI defects on the appearance of the display screen can effectively distinguish the surface interference of the display screen from the surface defects of the display screen, reduce the error detection rate of the AOI defects on the appearance of the display screen, and improve the accuracy and the efficiency of the AOI defects on the appearance of the display screen.
The embodiments provided in the present application are only a few examples of the general concept of the present application, and do not limit the scope of the present application. Any other embodiments extended according to the scheme of the present application without inventive efforts will be within the scope of protection of the present application for a person skilled in the art.
Claims (11)
1. A method for reducing the false detection rate of AOI defects on the appearance of a display screen is characterized by comprising the following steps:
acquiring a first image of the appearance of a display screen to be detected, and extracting a suspected defect image and position information according to the first image;
wiping the display screen to be detected;
acquiring a second image of the appearance of the display screen to be detected after wiping, and extracting an image of a position corresponding to the second image according to the position information of the suspected defect image to obtain a re-judgment image;
judging the similarity of the suspected defect image and the re-judged image through a depth measurement similarity judgment model;
and if the similarity result between the suspected defect image and the re-judged image is 1, judging the image as a defect, and outputting the position information of the suspected defect image.
2. The method according to claim 1, wherein the step of determining the similarity between the suspected defect image and the re-determined image by a depth measure similarity determination model comprises:
constructing a depth measurement learning model through a Deepsiamese network;
training the depth measurement learning model according to the positive sample set and the negative sample set to obtain the depth measurement similarity judgment model;
and inputting the suspected defect image and the re-judgment image into the depth measurement similarity judgment model, and calculating by the depth measurement similarity judgment model to obtain a similarity result.
3. The method for reducing the display screen appearance AOI defect false detection rate according to claim 2, wherein the depth metric learning model backbone network comprises 4 convolutional layers, 2 pooling layers and 2 full-link layers, a data input layer of the depth metric learning model comprises a six-channel data pair formed by synthesizing two three-channel images, wherein the suspected defect image is input through three channels, and the repeated judgment image is input through the other three channels.
4. The method according to claim 2 or 3, wherein the positive sample set includes { defect sample image, suspected defect sample image } data pairs, the negative sample set includes { interference sample image, suspected defect sample image } data pairs, and the interference image is an image with dust and/or fingerprints and/or greasy dirt on the display screen.
5. The method for reducing the display screen appearance AOI defect false detection rate according to claim 4, wherein the step of training the depth metric learning model according to the positive sample set and the negative sample set further comprises: and performing data amplification on the positive sample set and the negative sample set, and respectively obtaining an original image of a sample image, a brightness change image of the sample image and a contrast change image of the sample image.
6. The method for reducing the display screen appearance AOI defect false detection rate according to claim 5, wherein the method for extracting the suspected defect image and the position information according to the first image adopts a dynamic threshold segmentation method.
7. A system for reducing the false detection rate of AOI defects on the appearance of a display screen is used for executing the method of any one of claims 1 to 6, and is characterized by comprising an image acquisition device, a wiping device and a processor connected with the image acquisition device and the wiping device;
the image acquisition device is used for acquiring a first image of the appearance of the display screen to be detected and a second image of the appearance of the display screen to be detected after wiping;
the wiping device is used for wiping the display screen to be detected;
the processor is used for extracting a suspected defect image and position information according to the first image; extracting an image at a position corresponding to the second image according to the position information of the suspected defect image to obtain a re-judgment image; judging the similarity of the suspected defect image and the re-judged image through a depth measurement similarity judgment model; and outputting the position information of the suspected defect image.
8. The system for reducing the display screen appearance AOI defect false detection rate according to claim 7, wherein the processor comprises an offline module and an online module;
wherein the offline module is to: constructing a depth measurement learning model through a Deepsiamese network; training the depth measurement learning model according to the positive sample set and the negative sample set to obtain the depth measurement similarity judgment model;
the online module is used for: and inputting the suspected defect image and the re-judgment image into the depth measurement similarity judgment model, and calculating by the depth measurement similarity judgment model to obtain a similarity result.
9. The system for reducing the display screen appearance AOI defect false detection rate according to claim 7, wherein the wiping device comprises: the wet wiping elastic pressure head is provided with dustless cloth soaked in alcohol, and the dry wiping elastic pressure head is provided with dry dustless cloth.
10. The system for reducing the display screen appearance AOI defect false detection rate according to claim 7, wherein the image acquisition device comprises: high resolution industrial cameras, multi-angle light sources or three-dimensional effect light sources.
11. The system for reducing the display screen appearance AOI defect false detection rate according to claim 7, wherein the processor comprises an offline module and an online module; wherein the offline module is to: constructing a depth measurement learning model through a Deepsiamese network; training the depth measurement learning model according to the positive sample set and the negative sample set to obtain the depth measurement similarity judgment model; the online module is used for: inputting the suspected defect image and the re-judgment image into the depth measurement similarity judgment model, and calculating by the depth measurement similarity judgment model to obtain a similarity result;
the wiping device includes: the wet wiping elastic pressure head is provided with dustless cloth soaked with alcohol, and the dry wiping elastic pressure head is provided with dry dustless cloth;
the image acquisition device includes: high resolution industrial cameras, multi-angle light sources or three-dimensional effect light sources.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011548782.4A CN112561904A (en) | 2020-12-24 | 2020-12-24 | Method and system for reducing false detection rate of AOI (argon oxygen decarburization) defects on display screen appearance |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011548782.4A CN112561904A (en) | 2020-12-24 | 2020-12-24 | Method and system for reducing false detection rate of AOI (argon oxygen decarburization) defects on display screen appearance |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112561904A true CN112561904A (en) | 2021-03-26 |
Family
ID=75033205
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011548782.4A Pending CN112561904A (en) | 2020-12-24 | 2020-12-24 | Method and system for reducing false detection rate of AOI (argon oxygen decarburization) defects on display screen appearance |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112561904A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113379746A (en) * | 2021-08-16 | 2021-09-10 | 深圳荣耀智能机器有限公司 | Image detection method, device, system, computing equipment and readable storage medium |
| CN114324330A (en) * | 2021-12-24 | 2022-04-12 | 深圳一信泰质量技术有限公司 | Ultrahigh-definition intelligent interactive display terminal performance detection device and method |
| CN114820521A (en) * | 2022-04-27 | 2022-07-29 | 电子科技大学 | Defect detection method for complex picture of display screen |
| CN116609345A (en) * | 2023-07-19 | 2023-08-18 | 北京阿丘机器人科技有限公司 | Battery cover plate defect detection method, device, equipment and storage medium |
Citations (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101324713A (en) * | 2007-05-30 | 2008-12-17 | 日本麦可罗尼克斯股份有限公司 | Liquid crystal panel inspection method and apparatus |
| CN105259181A (en) * | 2015-10-26 | 2016-01-20 | 华为技术有限公司 | Display screen display defect detecting method, device and equipment |
| CN106683081A (en) * | 2016-12-17 | 2017-05-17 | 复旦大学 | Brain glioma molecular marker nondestructive prediction method and prediction system based on radiomics |
| WO2018000731A1 (en) * | 2016-06-28 | 2018-01-04 | 华南理工大学 | Method for automatically detecting curved surface defect and device thereof |
| CN107967255A (en) * | 2017-11-08 | 2018-04-27 | 北京广利核***工程有限公司 | A kind of method and system for judging text similarity |
| CN109902560A (en) * | 2019-01-15 | 2019-06-18 | 浙江师范大学 | A kind of fatigue driving method for early warning based on deep learning |
| CN110108713A (en) * | 2019-04-26 | 2019-08-09 | 武汉精立电子技术有限公司 | A kind of Superficial Foreign Body defect fast filtering method and system |
| US20200134800A1 (en) * | 2018-10-29 | 2020-04-30 | International Business Machines Corporation | Precision defect detection based on image difference with respect to templates |
| CN111208148A (en) * | 2020-02-21 | 2020-05-29 | 凌云光技术集团有限责任公司 | Dig hole screen light leak defect detecting system |
| CN111242185A (en) * | 2020-01-03 | 2020-06-05 | 凌云光技术集团有限责任公司 | Defect rapid preliminary screening method and system based on deep learning |
| CN111275736A (en) * | 2020-01-07 | 2020-06-12 | 中国科学院大学 | Unmanned aerial vehicle video multi-target tracking method based on target scene consistency |
| CN111612781A (en) * | 2020-05-27 | 2020-09-01 | 歌尔股份有限公司 | Screen defect detection method and device and head-mounted display equipment |
| CN111612763A (en) * | 2020-05-20 | 2020-09-01 | 重庆邮电大学 | Mobile phone screen defect detection method, device and system, computer equipment and medium |
-
2020
- 2020-12-24 CN CN202011548782.4A patent/CN112561904A/en active Pending
Patent Citations (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101324713A (en) * | 2007-05-30 | 2008-12-17 | 日本麦可罗尼克斯股份有限公司 | Liquid crystal panel inspection method and apparatus |
| CN105259181A (en) * | 2015-10-26 | 2016-01-20 | 华为技术有限公司 | Display screen display defect detecting method, device and equipment |
| WO2018000731A1 (en) * | 2016-06-28 | 2018-01-04 | 华南理工大学 | Method for automatically detecting curved surface defect and device thereof |
| CN106683081A (en) * | 2016-12-17 | 2017-05-17 | 复旦大学 | Brain glioma molecular marker nondestructive prediction method and prediction system based on radiomics |
| CN107967255A (en) * | 2017-11-08 | 2018-04-27 | 北京广利核***工程有限公司 | A kind of method and system for judging text similarity |
| US20200134800A1 (en) * | 2018-10-29 | 2020-04-30 | International Business Machines Corporation | Precision defect detection based on image difference with respect to templates |
| CN109902560A (en) * | 2019-01-15 | 2019-06-18 | 浙江师范大学 | A kind of fatigue driving method for early warning based on deep learning |
| CN110108713A (en) * | 2019-04-26 | 2019-08-09 | 武汉精立电子技术有限公司 | A kind of Superficial Foreign Body defect fast filtering method and system |
| CN111242185A (en) * | 2020-01-03 | 2020-06-05 | 凌云光技术集团有限责任公司 | Defect rapid preliminary screening method and system based on deep learning |
| CN111275736A (en) * | 2020-01-07 | 2020-06-12 | 中国科学院大学 | Unmanned aerial vehicle video multi-target tracking method based on target scene consistency |
| CN111208148A (en) * | 2020-02-21 | 2020-05-29 | 凌云光技术集团有限责任公司 | Dig hole screen light leak defect detecting system |
| CN111612763A (en) * | 2020-05-20 | 2020-09-01 | 重庆邮电大学 | Mobile phone screen defect detection method, device and system, computer equipment and medium |
| CN111612781A (en) * | 2020-05-27 | 2020-09-01 | 歌尔股份有限公司 | Screen defect detection method and device and head-mounted display equipment |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113379746A (en) * | 2021-08-16 | 2021-09-10 | 深圳荣耀智能机器有限公司 | Image detection method, device, system, computing equipment and readable storage medium |
| CN113379746B (en) * | 2021-08-16 | 2021-11-02 | 深圳荣耀智能机器有限公司 | Image detection method, device, system, computing equipment and readable storage medium |
| CN114324330A (en) * | 2021-12-24 | 2022-04-12 | 深圳一信泰质量技术有限公司 | Ultrahigh-definition intelligent interactive display terminal performance detection device and method |
| CN114324330B (en) * | 2021-12-24 | 2023-09-12 | 深圳一信泰质量技术有限公司 | Device and method for detecting performance of ultra-high definition intelligent interactive display terminal |
| CN114820521A (en) * | 2022-04-27 | 2022-07-29 | 电子科技大学 | Defect detection method for complex picture of display screen |
| CN114820521B (en) * | 2022-04-27 | 2023-04-07 | 电子科技大学 | Defect detection method for complex picture of display screen |
| CN116609345A (en) * | 2023-07-19 | 2023-08-18 | 北京阿丘机器人科技有限公司 | Battery cover plate defect detection method, device, equipment and storage medium |
| CN116609345B (en) * | 2023-07-19 | 2023-10-17 | 北京阿丘机器人科技有限公司 | Battery cover plate defect detection method, device, equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112561904A (en) | Method and system for reducing false detection rate of AOI (argon oxygen decarburization) defects on display screen appearance | |
| CN113450307B (en) | Product edge defect detection method | |
| CN111223088B (en) | Casting surface defect identification method based on deep convolutional neural network | |
| WO2021143343A1 (en) | Method and device for testing product quality | |
| CN108416766B (en) | Double-side light-entering type light guide plate defect visual detection method | |
| CN111383209A (en) | Unsupervised flaw detection method based on full convolution self-encoder network | |
| CN111223093A (en) | AOI defect detection method | |
| CN110555842A (en) | Silicon wafer image defect detection method based on anchor point set optimization | |
| CN112233090B (en) | Film flaw detection method based on improved attention mechanism | |
| CN108682012B (en) | 3D curved surface glass surface flatness defect detection method based on line scanning laser | |
| CN108665458A (en) | Transparent body surface defect is extracted and recognition methods | |
| CN110021012B (en) | Mobile phone lens window glass defect detection method based on machine vision technology | |
| CN112132196B (en) | Cigarette case defect identification method combining deep learning and image processing | |
| CN112991271B (en) | Aluminum profile surface defect visual detection method based on improved yolov3 | |
| TWI710762B (en) | An image classification system | |
| CN114677601A (en) | Dam crack detection method based on unmanned aerial vehicle inspection and combined with deep learning | |
| CN112070746A (en) | Steel strip defect detection method and device | |
| CN116228651A (en) | Cloth defect detection method, system, equipment and medium | |
| CN115170475A (en) | Non-woven fabric defect detection method based on deep learning method | |
| CN116612106A (en) | Method for detecting surface defects of optical element based on YOLOX algorithm | |
| CN116563262A (en) | Building crack detection algorithm based on multiple modes | |
| CN115187544A (en) | DR-RSBU-YOLOv 5-based fabric flaw detection method | |
| CN114331961A (en) | Method for defect detection of an object | |
| CN117115171B (en) | Slight bright point defect detection method applied to subway LCD display screen | |
| CN116523916B (en) | Product surface defect detection method and device, electronic equipment and storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |