CN109297973A - Defect inspecting system and defect detecting method - Google Patents
Defect inspecting system and defect detecting method Download PDFInfo
- Publication number
- CN109297973A CN109297973A CN201810800087.9A CN201810800087A CN109297973A CN 109297973 A CN109297973 A CN 109297973A CN 201810800087 A CN201810800087 A CN 201810800087A CN 109297973 A CN109297973 A CN 109297973A
- Authority
- CN
- China
- Prior art keywords
- image
- defect
- check object
- column split
- column
- 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.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/84—Systems specially adapted for particular applications
- G01N21/88—Investigating the presence of flaws or contamination
- G01N21/8851—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/01—Arrangements or apparatus for facilitating the optical investigation
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/84—Systems specially adapted for particular applications
- G01N21/88—Investigating the presence of flaws or contamination
- G01N21/8806—Specially adapted optical and illumination features
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/84—Systems specially adapted for particular applications
- G01N21/88—Investigating the presence of flaws or contamination
- G01N21/95—Investigating the presence of flaws or contamination characterised by the material or shape of the object to be examined
- G01N21/958—Inspecting transparent materials or objects, e.g. windscreens
-
- 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
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/84—Systems specially adapted for particular applications
- G01N21/88—Investigating the presence of flaws or contamination
- G01N21/8806—Specially adapted optical and illumination features
- G01N2021/8809—Adjustment for highlighting flaws
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/84—Systems specially adapted for particular applications
- G01N21/88—Investigating the presence of flaws or contamination
- G01N21/8851—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges
- G01N2021/8887—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges based on image processing techniques
Abstract
The present invention provides defect inspecting system and defect detecting method.In defect inspecting system (1), by the two dimensional image of image pickup part (3) shooting brightness change on conveying direction (X), the image data of column split image is processed by column split processing unit (9), the column split image is column split image made of sequence arrangement of the Leie for the same position that two dimensional image is divided into the column along conveying direction (X) arrangement and makes two dimensional image according to time series, the shooting image of even identical check object (T), column split image etc. also have different brightness.The data accumulated by defect classification identification part (10) based on the result of the relevant rote learning of identification to the defect classification for being included to the column split image that brightness and presentation mode are different, come identify check object (T) defect classification, the accuracy of identification of defect improves.
Description
Technical field
The present invention relates to defect inspecting system and defect detecting methods.
Background technique
As the defect inspecting system checked based on the shooting image of check object the defect of check object, example
As it has been known that there is the defect inspections of the defect of stacked film used in the diaphragm of the detection optical films such as polarizing coating and phase difference film, battery etc.
Look into system.This defect inspecting system is along conveying direction transport membrane, by the two dimensional image of discrete time shooting film, based on shooting
Two dimensional image out carries out defect inspection.For example, Japan's patent the 4726983rd system generates column split image, it should
Column split image is shot along multiple column and making of conveying direction arrangement by discrete time by being divided into two dimensional image
Two dimensional image respectively in the Leie of same position arranged according to the sequence of time series.Column split image is processed into enhancing
The defect enhancing processing image of brightness change.Pass through defect enhancing processing image, in this way it is easy to determine the presence or absence of defect of film, film
The position of defect.
In addition, even if the two dimensional image of check object is processed into defect enhancing processing image as above-mentioned technology, most
Eventually also by the identification for carrying out defect based on the judgement of people, there are rooms for improvement for the accuracy of identification of defect.
Summary of the invention
Then, the purpose of the present invention is to provide the defect inspecting systems for the accuracy of identification that can be improved defect and defect to examine
Checking method.
The present invention relates to a kind of defect inspecting systems, have: light source, to check object irradiation light;Image pickup part is pressed
Discrete time shoots two dimensional image, which is based on that check object is irradiated and penetrated from light source to check object or is being checked
On object reflect after light and formed;Delivery section, check object is opposite along conveying direction relative to light source and image pickup part
Ground conveying;And image processing part, the image data for the two dimensional image shot by image pickup part is handled, image pickup part is clapped
It takes out in two dimensional image and the changed two dimensional image of brightness on the consistent direction of conveying direction, image processing part includes
Two dimensional image, is processed into the image data of column split image by column split processing unit, and column split image is by by two dimensional image
Be divided into multiple column along conveying direction arrangement and make the two dimensional image shot by image pickup part by discrete time respectively in
The Leie of same position is arranged according to the sequence of time series;And defect classification identification part, based on to two or more
The column split image defect that is included classification the relevant rote learning of identification the data that are accumulated of result, come
Identify the classification of the defect of check object, wherein more than two column split images are handled by column split processing unit
The image arrived.
According to this structure, defect inspecting system has: light source, to check object irradiation light;Image pickup part, by it is discrete when
Between shoot two dimensional image, the two dimensional image be based on check object is irradiated and penetrated from light source to check object or in check object
Light after reflection and formed;Delivery section relatively conveys check object relative to light source and image pickup part along conveying direction;
And image processing part, the image data for the two dimensional image shot by image pickup part is handled, wherein clapped by image pickup part
It takes out in two dimensional image and the changed two dimensional image of brightness on the consistent direction of conveying direction, by the column of image processing part
Two dimensional image is processed into the image data of column split image by cutting treatmenting part, and the column split image is by dividing two dimensional image
Be segmented into multiple column along conveying direction arrangement and make the two dimensional image shot by image pickup part by discrete time respectively in phase
Leie with position is arranged according to the sequence of time series, so even being shot to identical check object
Image, each column split image also become the image with different brightness.Moreover, by the defect classification identification part of image processing part
The result of the relevant rote learning of identification based on the classification to the defect for being included to following column split image accumulated and
The data obtained, come identify check object defect classification, wherein the column split image is handled by column split processing unit
Obtained more than two column split images for being respectively provided with different brightness, therefore, even to identical check object into
The image that row shooting obtains, but based on for more than two column split images progress that brightness is different and presentation mode is different
The result of rote learning identifies the classification of defect, so as to improve the accuracy of identification of defect.
In which case it is preferable that defect classification identification part is based on to the two or more different from 10% or more brightness
The column split image defect that is included classification the relevant rote learning of identification the data that are accumulated of result, come
Identify the classification of the defect of check object.
According to this structure, defect classification identification part is based on wrapping the two column split images different from 10% or more brightness
The data that the result of the relevant rote learning of identification of the classification of the defect contained is accumulated, to identify lacking for check object
Sunken classification, therefore, the image that even identical check object is shot, but based on substantially different for brightness
Result up to 10% or more and the presentation mode rote learning that substantially different two column split images carry out identifies defect
Classification, therefore the accuracy of identification of defect can be further increased.
It is further preferred, that defect inspecting system is also equipped with occulter, the occulter be located at light source and check object it
Between, and a part of the light irradiated from light source to check object is blocked, thus being shot by image pickup part by discrete time
Two dimensional image on form bright portion and dark portion, delivery section by check object relative to light source, occulter and image pickup part along with it is bright
Portion and the line of demarcation of dark portion intersection conveying direction relatively convey, defect classification identification part be based on to following column split image
The data that the result of the relevant rote learning of identification of the classification for the defect for being included is accumulated, to identify check object
Defect classification, column split image refers to: arranging the Leie of the position in the bright portion in two dimensional image according to the sequence of time series
Column split image made of column;And arrange the Leie of the position of dark portion in two dimensional image according to the sequence of time series
Column split image.
According to this structure, from the occulter between light source and check object to the light irradiated from light source to check object
A part blocked, so that bright portion and dark portion are formed on the two dimensional image shot by image pickup part by discrete time, by defeated
Send portion by check object relative to the conveying side intersected with light source, occulter and image pickup part along the line of demarcation with bright portion and dark portion
Each position of check object into a series of two dimensional image for relatively conveying, therefore shooting by discrete time enters bright
Portion and dark portion this two side.In addition, defect classification identification part is based on the classification to the defect for being included with following column split image
Identify the data that the result of relevant rote learning is accumulated, come identify check object defect classification, wherein institute
It states column split image to refer to: made of arranging the Leie of the position in the bright portion in two dimensional image according to the sequence of time series
Column split image;And pass through column made of arranging the Leie of the position of the dark portion in two dimensional image according to the sequence of time series
Segmented image, thus based on for be belonging respectively to bright portion and dark portion and presentation mode substantially different two column split images carry out
The result of rote learning identify the classification of defect, the accuracy of identification of defect can be further increased.
On the other hand, the present invention relates to a kind of defect detecting methods comprising: from the light source of defect inspecting system to inspection
The irradiation process of object irradiation light;The camera shooting process of two dimensional image is shot by discrete time by the image pickup part of defect inspecting system,
Wherein, two dimensional image is based on irradiating and penetrate check object from light source to check object in irradiation process or in check object
Light after reflection and formed;By the delivery section of defect inspecting system by check object relative to light source and image pickup part along conveying side
To the conveying operation relatively conveyed;And by defect inspecting system image processing part to camera shooting process in shoot two
The image procossing process that is handled of image data of dimension image is shot in two dimensional image and conveying in the camera shooting process
The changed two dimensional image of brightness on the consistent direction in direction, includes: to be processed into two dimensional image in image procossing process
The column split treatment process of the image data of column split image, wherein column split image by by two dimensional image be divided into along
Multiple column of conveying direction arrangement and make in the camera shooting process two dimensional image shot by discrete time respectively in identical bits
The Leie set is arranged according to the sequence of time series;And based on to the defect for being included with more than two column split images
Classification the relevant rote learning of identification the data that are accumulated of result, come identify check object defect classification
Defect classification identify process, wherein more than two column split images are the figures handled in column split treatment process
Picture.
In which case it is preferable that in defect classification identification process, based on to different from 10% or more brightness two
The data that the result of the relevant rote learning of identification of the classification for the defect that a column split image is included is accumulated, come
Identify the classification of the defect of check object.
It is further preferred, that in irradiation process, by occulter in two shot by imaging process by discrete time
Bright portion and dark portion are formed on dimension image, occulter irradiates between light source and check object, and to from light source to check object
A part of light blocked, in conveying operation, by check object relative to light source, occulter and image pickup part along with it is bright
Portion and the line of demarcation of dark portion intersection conveying direction relatively convey, defect classification identification process in, based on to it is such as following
The data that the result of the relevant rote learning of identification of the classification for the defect that segmented image is included is accumulated, to identify
The classification of the defect of check object, column split image refer to: by making the Leie of the position in the bright portion in two dimensional image shine the time
Column split image made of the sequence arrangement of sequence;And by making the Leie of the position of the dark portion in two dimensional image shine time sequence
Column split image made of the sequence arrangement of column.
Detailed description of the invention
Fig. 1 is the perspective view for indicating the defect inspecting system of embodiment.
Fig. 2 is the figure of the configuration of the light source for indicating the defect inspecting system of Fig. 1, image pickup part, occulter and check object.
Fig. 3 is the block diagram for indicating the details of the image processing part of defect inspecting system of Fig. 1.
Fig. 4 is the flow chart for indicating the process of defect detecting method of embodiment.
Fig. 5 is the flow chart for indicating the details of image procossing process of Fig. 4.
(A), (B), (C), (D), (E), (F), (G) of Fig. 6 is indicated by the image processing part of the defect inspecting system of Fig. 1
Column split processing unit processing image figure.
(A) of Fig. 7 is the figure for indicating the two dimensional image of time series, and (B) is to indicate to make the Leie of each position according to time sequence
The figure of each column split image made of the sequence arrangement of column, (C) are shown so that each column split image of (B) indicates check object
The mode of same position will be staggered constantly the resulting figure to bit image.
Fig. 8 is the figure for indicating convolutional neural networks.
Specific embodiment
Hereinafter, explaining the preferred reality of defect inspecting system and defect detecting method of the invention in detail referring to attached drawing
Apply mode.As shown in Figures 1 and 2, the defect inspecting system 1 of embodiments of the present invention has light source 2, image pickup part 3, delivery section
4, image processing part 5, occulter 6, parallel light lens 7 and display device 8.The defect inspecting system of present embodiment is by polarizing coating
And the films such as stacked film used in the diaphragm of the optical films such as phase difference film, battery detect check object T's as check object T
Defect.Check object T extends along the conveying direction X of delivery section 4, has on the width direction Y orthogonal with conveying direction X pre-
The width first set.Refer to the state different from desired state in the defect that check object T is generated, such as can enumerate different
Object, scratch, bubble (bubble for generating etc. in forming), foreign bubble (bubble generated by being mixed into for foreign matter etc.), scar,
Crackle (crackle etc. generated by broken line trace etc.) and striped (striped etc. generated by the difference of thickness).Defect inspection
The classification of these defects of the identification of system 1.Defect inspecting system 1 is other than the identification of the classification of defect, additionally it is possible to determine defect
It is to be generated in which face of check object T.
As shown in Figures 1 and 2, light source 2 is to check object T irradiation light.It is parallel with width direction Y that light source 2 is configured to irradiation
Linear light.As light source 2, as long as the irradiations such as metal halide lamp, halogen transmission lamp, fluorescent lamp are not given as inspection pair
As the lamp for the light that the composition and property of the film of T affect, it is not particularly limited.
Image pickup part 3 shoots two dimensional image by discrete time, which is based on irradiating simultaneously from light source 2 to check object T
Light after reflecting through check object T or on check object T and formed.There are image pickup part 3 multiple optical components and photoelectricity to turn
Change element.Optical component includes optical lens, optical gate etc., makes to penetrate as the light after the film of check object T in photoelectric conversion element
The surface of part is imaged.Photo-electric conversion element is the CCD (Charge Coupled Device) or CMOS by shooting two dimensional image
The face sensor that photographing elements such as (Complementary Metal-Oxide Semiconductor) are constituted.Image pickup part 3 can also
To be the component either shot in two dimensional image and colorful two dimensional image without color.
Delivery section 4 relatively conveys check object T-phase light source 2 and image pickup part 3 along conveying direction X.Delivery section 4
Such as has and, along the conveying direction X outlet roller conveyed and receiving roll, rotary encoder etc. will be passed through as the film of check object T
To measure conveying distance.In the present embodiment, delivery section 4 is set to edge to the check object T conveying speed conveyed
Conveying direction X be 2~100m/ minutes this degree.The conveying speed of delivery section 4 is by the equal setting of image processing part 5 and control.
The image data for the two dimensional image that the processing of image processing part 5 is shot by image pickup part 3.As long as image processing part 5
The component for carrying out the image procossing of two-dimensional image data, is just not particularly limited, such as can be applicable in and be equipped with image processing software
PC (personal computer), carry the FPGA (Field Programmable Gate Array) of image processing circuit on the books
Image pick-up card etc..
Occulter 6 passes through one to the light irradiated from light source 2 to check object T between light source 2 and check object T
Part is blocked, to form bright portion and dark portion on the two dimensional image shot by image pickup part 3 by discrete time.By occulter
6, image pickup part 3 is shot in two dimensional image and the changed two dimensional image of brightness on the consistent direction conveying direction X.More
For body, delivery section 4 by check object T-phase for light source 2, parallel light lens 7, occulter 6 and image pickup part 3 along with bright portion with
The conveying direction X of the line of demarcation intersection of dark portion is relatively conveyed.In the present embodiment, line of demarcation is parallel to and conveying direction X
Vertical width direction Y.It should be noted that as long as image pickup part 3 can be shot in the consistent with conveying direction X of two dimensional image
Direction on the changed two dimensional image of brightness, may not possess occulter 6.Parallel light lens 7 make from light source 2 to
The direction of travel for the light that check object T and occulter 6 irradiate is parallel.Parallel light lens 7 for example can be by telecentric optical system structure
At.
The display device 8 connecting with image processing part 5 is constituted such as by PC (personal computer), will be by image processing part
The classification of 5 defects identified is shown in LC (Liquid Crystal) display panel, plasma display panel, EL
(Electro Luminescence) display panel etc..It is handled it should be noted that image processing part 5 also can have display
The display device of the image arrived.
Hereinafter, illustrating the details of image processing part 5.As shown in figure 3, image processing part 5 has column split processing unit 9
With defect classification identification part 10.Two dimensional image is processed into the image data of column split image, the column point by column split processing unit 9
Image is cut by the way that two dimensional image to be divided into multiple column along conveying direction X arrangement and makes to be clapped by image pickup part 3 by discrete time
The two dimensional image taken out respectively in the Leie of same position arranged according to the sequence of time series.Defect classification identification part 10
Based on to the identification phase with the classification by treated the defect that more than two column split images are included of column split processing unit 9
The data that the result of the rote learning of pass is accumulated, come identify check object T defect classification.To rote learning
As a result the data accumulated are stored in the storage devices such as the hard disk of PC comprising defect classification identification part 10, with mechanical
The result of study and be updated.
It should be noted that in the present embodiment, to the identification phase of the classification for the defect for being included with column split image
The data that the result of the rote learning of pass is accumulated are in addition to including to the image pickup part with the inside by defect inspecting system 1
The knowledge of the classification for the defect that column split image after the 3 a series of two dimensional images shot by discrete time are processed is included
Other than the data that the result of not relevant rote learning is accumulated, further include to in the outside of defect inspecting system 1
The result of the relevant rote learning of identification of the classification for the defect that the column split image separately generated is included is accumulated to obtain
Data.That is, in the present embodiment, in addition to including in the state that the inside of defect inspecting system 1 has carried out rote learning
It further include based on not yet carrying out rote learning in the inside of defect inspecting system 1 other than the scheme for identifying the classification of defect
The data that the result of the rote learning separately generated under state in the outside of defect inspecting system 1 is accumulated, to identify
The scheme of the classification of defect.
Identification phase of the defect classification identification part 10 based on the classification to the defect for being included with following two column split images
The data that the result of the rote learning of pass is accumulated, come identify check object T defect classification, it is described two column point
Cutting image is two different column split images of 10% or more brightness.In addition, defect classification identification part 10 be based on to it is such as following
The data that the result of the relevant rote learning of identification of the classification for the defect that segmented image is included is accumulated, to identify
The classification of the defect of check object T, the column split image refer to: by making in the two dimensional image obtained by occulter 6
Column split image made of sequence arrangement of the Leie of the position in bright portion according to time series;And by making to obtain by occulter 6
To two dimensional image in dark portion position Leie according to time series sequence arrangement made of column split image.
Hereinafter, illustrating the defect detecting method of present embodiment.As shown in figure 4, carrying out the light source from defect inspecting system 1
2 to check object T irradiation light irradiation process (S1).As shown in (A) of Fig. 6, in irradiation process, using be located at light source 2 with
The defect inspecting system 1 blocked between check object T and to a part of the light irradiated from light source 2 to check object T
Occulter 6, camera shooting process in by discrete time shooting two dimensional image F (t1) on formed with line of demarcation b be demarcate bright portion 1
With dark portion d.As shown in (A) of Fig. 6, for the two dimensional image F (t1) under moment t=t1, the light from light source 2 is by occulter
6 block, thus with reach conveying direction X downstream side and the brightness in two dimensional image F (t1) is got higher.In addition, in two dimension
Image F (t1) shows the defect D on the film for having check object T.Moment t=t2, t3 ..., two dimensional image F (t2), F under tm
(t3) ..., F (tm) is also same (m is arbitrary natural number).
As shown in figure 4, carrying out camera shooting process (S2) by the image pickup part 3 of defect inspecting system 1, in the camera shooting process, press
Discrete time shoots two dimensional image F (t1), and two dimensional image F (t1) is based on shining in irradiation process from light source 2 to check object T
Penetrate and penetrate check object T or on check object T reflect after light and formed.As shown in (A) of Fig. 6, in camera shooting process,
A part of the light irradiated from light source 2 to check object T is blocked from occulter 6, thus shoot two dimensional image F (t1) with
The changed two dimensional image F (t1) of brightness on the consistent direction conveying direction X.Two dimensional image under moment t=t2, t3 ... tm
F (t2), F (t3) ..., F (tm) be also same.
In addition, as shown in figure 4, being carried out check object T-phase for light source 2 and being taken the photograph by the delivery section 4 of defect inspecting system 1
The conveying operation (S3) relatively conveyed as portion 3 along conveying direction X.As shown in (A) of Fig. 6, in conveying operation, it will check
Object T-phase intersects light source 2, parallel light lens 7, occulter 6 and image pickup part 3 along with the line of demarcation b in bright portion 1 and dark portion d
Conveying direction X relatively convey.In the present embodiment, line of demarcation b is parallel to the width direction Y orthogonal with conveying direction X,
But line of demarcation b and conveying direction X angulation are also possible to the angle other than 90 °.In addition, line of demarcation b is not necessarily stringent
Line of demarcation, the maximum position of brightness and the X-Y scheme comprising dark portion d that line of demarcation b refers to the two dimensional image F (t1) comprising bright portion 1
As the line of the centre at the smallest position of the brightness of F.
As shown in figure 4, being carried out by the image processing part 5 of defect inspecting system 1 to the two dimension shot in camera shooting process
The image procossing process (S4) that image F (t1)~F (tm) image data is handled.Hereinafter, illustrating image treatment process
Details.As shown in figure 5, being handled in image procossing process by the column split of the image processing part 5 of defect inspecting system 1
Portion 9 carries out column split treatment process (S41).As shown in (B) of Fig. 6, in column split treatment process, column split processing unit 9 will
Two dimensional image F (t1), which is divided into, arranges L1 (t1)~jth column Lj (t1)~kth column along multiple column the i.e. the 1st of conveying direction X arrangement
Lk (t1) (j and k are arbitrary natural number, j≤k).Arrange the width of the conveying direction X of L1 (t1)~column Lk (t1) and at the moment
T1, moment t2 ..., moment tj ..., each in moment tm when the frame period inscribed in by check object T along conveying direction X
The distance of conveying is identical.To under moment t=t2, t3 ... tm two dimensional image F (t2), F (t3) ..., F (tm) also carries out similarly
Processing.
Two dimensional image F (t1)~F (tm) is processed into the image data of column split image, column split by column split processing unit 9
Image pass through make in camera shooting process two dimensional image F (t1)~F (tm) for being shot by discrete time respectively in same position
Column L1 (t1), L1 (t2) etc. arranged according to the sequence of time series.The 1st column split image is enumerated to be illustrated.Such as
Shown in (C) of Fig. 6, column split processing unit 9 make the two dimensional image F (t1) shot by discrete time, F (t2), F (t3) ... it is each
The most downstream side of conveying direction X in the 1st column L1 (t1), L1 (t2), L1 (t3) ... it is (defeated according to the sequence of time series
Send direction X) arrangement.As shown in (D) of Fig. 6, column split processing unit 9 make two dimensional image F (t1)~F (tm) respectively in the 1st column
L1 (t1)~L1 (tm) is arranged according to the sequence of time series and is generated the 1st column split image DL1 (t1).
As shown in (G) of (E) of Fig. 6, (F) of Fig. 6 and Fig. 6, column split processing unit 9 is also to two dimensional image F (t1)~F
(tm) respectively in the 1st column L1 (t1)~L1 (tm) ..., jth column Lj (t1)~Lj (tm) ..., kth column Lk (t1)~Lk (tm)
Similarly handled, generate the 1st column split image DL1 (t1) ..., jth column split image DLJ (t1) ..., kth column split
Image DLk (t1).As shown in (E) of Fig. 6, column split image DL1 (t1) is by making in two dimensional image F (t1)~F (tm)
Column split image made of sequence arrangement of the column L1 (t1) of the position in the bright portion 1~L1 (tm) according to time series.In addition, as schemed
Shown in 6 (F), column split image DLj (t1) is by making the position near the line of demarcation b in two dimensional image F (t1)~F (tm)
Column split image made of sequence arrangement of the column Lj (t1) the set~L1 (tm) according to time series.In addition, such as (G) institute of Fig. 6
Show, column split image DLk (t1) is column Lk (t1)~Lk by making the position of the dark portion d in two dimensional image F (t1)~F (tm)
(tm) column split image made of the sequence arrangement according to time series.
As shown in (G) of (E)~Fig. 6 of Fig. 6, column split image DL1 (t1)~DLk (t1) is to make to shoot by discrete time
Column L1 (t1)~Lk (t1) of same position during two dimensional image F (t1) out~F (tm) is respective is respectively in accordance with time series
Sequence arrange made of column split image, therefore it is identical at the time of range column split image DL1 (t1)~DLk (t1) indicate
The position of the different positions of check object T, the defects of column split image DL1 (t1)~DLk (t1) D also deviates respectively.In
Be, in the present embodiment, by make shoot range respectively at different times two dimensional image respectively in phase
With position Leie according to time series sequence arrangement made of column split image, thus so that each column split image indicate check
The mode of the same position of object T is aligned.
As shown in (A) of Fig. 7, in camera shooting process, two dimensional image F (t1)~F (tm) is shot by discrete time.Inspection pair
As T is conveyed along conveying direction X, therefore the position of the defects of two dimensional image F (t1)~F (tm) D deviates respectively.Such as Fig. 7
(B) shown in, as described above generate column split image DL1 (t1)~DLj (t1)~DLk (t1).Range at the time of identical
The different positions column split image DL1 (t1)~DLk (t1) expression check object T, therefore column split image DL1 (t1)~
The position of the defects of DLk (t1) D also deviates respectively.
Relative to from the downstream side of conveying direction X the 1st column L1 (t1)~L1 (tm), such as it is identical at the time of range
Indicated from the jth column Lj (t1) the downstream side of conveying direction X~Lj (tm) to be checked in the frame period of the amount of (j-1) pair
The position deviated as the transported distance of T to the upstream side of the conveying direction X of check object T.Therefore, as shown in (C) of Fig. 7,
Relative to the column split image DL1 (tm) of the 1st column L1 (tm)~L1 (t (m+ (m-1))), such as the column split image with regard to jth column
For, t (m- (j- at the time of obtained by the time of the frame period of the amount of (j-1) has been recalled relative to moment t1~moment tm range
1) the column split image DLj (t (m- (j-1))) of the range of)~moment t (m+ (m-j)) indicates the same position of check object T.
Similarly, the column split image DL1 (tm) relative to the 1st column L1 (tm)~L1 (t (m+ (m-1))), such as with regard to kth
For the column split image of column, the time institute of the frame period of the amount of (k-1) has been recalled relative to moment t1~moment tm range
The column split image DLk (t (m- (k-1))) of the range of t (m- (k-1))~moment t (m+ (m-k)) indicates inspection pair at the time of obtaining
As the same position of T.
Alternatively, the column split image DL1 (t1) relative to the 1st column L1 (t1)~L1 (t (1+ (m-1))), such as jth is arranged
Column split image for, the column split image DLj (t (1- (j- of the range of moment t (1- (j-1))~moment t (1+ (m-j))
1) same position of check object T)) is indicated.In addition, the column split figure relative to the 1st column L1 (t1)~L1 (t (1+ (m-1)))
As DL1 (t1), such as the column split image of kth column, the range of moment t (1- (k-1))~moment t (1+ (m-k))
The same position of column split image DLk (t (1- (k-1))) expression check object T.By making the range at moment be staggered like this,
So as to be aligned in a manner of making each column split image indicate the same position of check object T.
It should be noted that the size of situation or column split image known to amount in positional shift is enough relative to defect
In the case where big, defect is centainly fallen into column split image, so even also can be by the column comprising defect without contraposition
Segmented image is used for rote learning.It therefore, in this case, can also be without contraposition.
As shown in figure 5, carrying out defect classification by the defect classification identification part 10 of the image processing part 5 of defect inspecting system 1
It identifies process (S42).Defect classification identification process in, defect classification identification part 10 be based on to more than two column split figures
As the data that the result of the relevant rote learning of identification of the classification for the defect for being included is accumulated, to identify inspection pair
As the classification of the defect D of T, described two above column split images be handled in column split treatment process two with
On column split image DL1 (t1) ..., DLj (t (1- (j-1))) ..., DLk (t (1- (k-1))).
In defect classification identification process, defect classification identification part 10 is based on to two column different from 10% or more brightness
The result of the relevant rote learning of identification of the classification for the defect that segmented image DL1 (t1), DLk (t1) are included accumulate
The data arrived, come identify check object T defect D classification.More specifically, in defect classification identification process, defect class
The result of the relevant rote learning of identification of the other identification part 10 based on the classification to the defect for being included to following column split image
The data accumulated, come identify check object T defect D classification, the column split image refers to: by making two dimension
Made of sequence arrangement of the column L1 (t1) of the position in the bright portion 1 in image F (t1)~F (the tm)~L1 (tk) according to time series
Column split image DL1 (t1);And the position by making the dark portion d in two dimensional image F (t (1- (k-1)))~F (t (1+ (m-k)))
Column split image DLk made of sequence arrangement of column Lk (t (1- (k-1)))~Lk (t (1+ (m-k))) set according to time series
(t(1-(k-1))).Rote learning is for example carried out by convolutional neural networks.It should be noted that as long as rote learning can be passed through
Identify the classification of defect, it can also be using the neural network or other methods other than convolutional neural networks.
As shown in figure 8, convolutional neural networks 100 have input layer 110, hidden layer 120 and output layer 130.It is examined by defect
The image processing part 5 of system 1 is looked by the column split image DL1 (t1) handled in column split treatment process~DLk (t (1-
(k-1)) more than two column split images in) are inputted to input layer 110.Hidden layer 120 have based on weight filter into
The convolutional layer 121 of row image procossing, 123, reduce in length and breadth and remain from the two-dimensional array that convolutional layer 121,123 exports
The full articulamentum 124 of the weight coefficient n of the pond layer 122 and each layer of update of the processing of the value of effect.It is defeated in output layer 130
Recognition result of the rote learning to the classification of defect D out.In convolutional neural networks 100, by the recognition result and normal solution of output
The error of value learns the weight of each layer to the inverse propagation of reverse R.
For example, in advance that multiple column split images are defeated to image processing part 5 together with the normal solution of the identification of the classification of defect D
Enter and learn image processing part 5, thus successively identifies the classification that column split image DL1 (t1) newly inputted etc. is included
Whether it is the classification of specific defect D, and is sequentially output recognition result.The error of the recognition result and normal solution that are sequentially output is to inverse
It is propagated to R is inverse, successively updates the weight coefficient n of each layer and accumulated as data.Successively having updated the weight of each phase
Under state, further successively identify whether the classification that column split image DL1 (t1) newly inputted etc. is included is specific defect
Classification, and be sequentially output recognition result, the error based on the recognition result and normal solution that are sequentially output successively updates each layer
Weight coefficient n is simultaneously accumulated as data, and repeatedly, thus the error of recognition result and normal solution becomes smaller, the classification of defect D
Identification precision improve.
According to the present embodiment, it is related to a kind of defect inspecting system 1, which has: light source 2, to inspection
It checks as T irradiation light;Image pickup part 3, by discrete time shooting two dimensional image F (t1)~F (tm) etc., two dimensional image F (t1)
~F (tm) etc. based on check object T is irradiated and penetrated from light source 2 to check object T or on check object T reflect after light and
It is formed;Delivery section 4 relatively conveys check object T-phase light source 2 and image pickup part 3 along conveying direction X;And figure
As processing unit 5, the image data of two dimensional image F (t1)~F (tm) shot by image pickup part 3 etc. is handled, wherein by imaging
It shoots in two dimensional image F (t1)~F (tm) etc. and the changed two dimension of brightness on the consistent direction conveying direction X in portion 3
Two dimensional image F (t1)~F (tm) etc. is processed by image F (t1)~F (tm) etc. by the column split processing unit 9 of image processing part 5
The image data of column split image DL1 (t1)~DLk (t (1- (k-1))), column split image DL1 (t1)~DLk (t (1- (k-
1))) by the way that two dimensional image F (t1)~F (tm) etc. is divided into multiple column L1 (the t1)~Lk (tm) arranged along conveying direction X
Deng and make two dimensional image F (the t1)~F (tm) shot by image pickup part 3 by discrete time etc. respectively in same position column
L1 (t1)~L1 (tm) etc. is arranged according to the sequence of time series, so even being shot to identical check object
Obtained image, each column split image DL1 (t1)~DLk (t (1- (k-1))) also become the image with different brightness.And
And by the defect classification identification part 10 of image processing part 5 based on the classification to the defect for being included with following column split image
Identify the data that the result of relevant rote learning is accumulated, come identify check object T defect D classification, wherein
The column split image is to handle obtained more than two column splits for being respectively provided with different brightness by column split processing unit 9
Image DL1 (t1)~DLk (t (1- (k-1))), therefore, the image that even identical check object T is shot,
But based on for more than two column split image DL1 (t1)~DLk (t (1- (k- that brightness is different and presentation mode is different
)) etc. 1) result of the rote learning of progress identifies the classification of defect D, so as to improve the accuracy of identification of defect D.
In addition, according to the present embodiment, defect classification identification part 10 be based on included to following two column split images
Defect classification the relevant rote learning of identification the data that are accumulated of result, to identify lacking for check object T
Fall into D classification, wherein described two column split images be two different column split image DL1 (t1) of 10% or more brightness,
DLk (t (1- (k-1))), therefore, the image that even identical check object T is shot, but based on for brightness
Substantially it is different up to 10% or more and presentation mode substantially different two column split image DL1 (t1), DLk (t (1- (k-1))) into
The result of capable rote learning identifies the classification of defect D, so as to further increase the accuracy of identification of defect D.
In addition, according to the present embodiment, from the occulter 6 between light source 2 and check object T to from light source 2 to inspection
It checks as T a part of light irradiated is blocked, thus in the two dimensional image F (t1) for pressing discrete time shooting by image pickup part 3
Bright portion 1 and dark portion d are formed on~F (tm) etc., by delivery section 4 by check object T-phase for light source 2, occulter 6 and image pickup part 3
One for relatively conveying along the conveying direction X intersected with the line of demarcation b in bright portion 1 and dark portion d, therefore being shot by discrete time
Each position of check object T in two dimensional image F (t1)~F (tm) etc. of series enters bright portion 1 and dark portion d this two side.In addition,
Defect classification identification part 10 is based on to the relevant rote learning of the identification of classification of defect for being included to following column split image
The data that are accumulated of result, come identify check object T defect D classification, wherein the column split image is
Refer to: by making column L1 (t1)~L1 (tm) of the position in the bright portion 1 in two dimensional image F (t1)~F (tm) etc. according to time series
Sequence arrangement made of column split image DL1 (t1);And by making two dimensional image F (t (1- (k-1)))~F (t (1+ (m-
K) column Lk (t (1- (k-1)))~Lk (t (1+ (m-k))) of the position of the dark portion d in)) according to time series sequence arrangement and
At column split image DLk (t (1- (k-1))), therefore, based on for being belonging respectively to bright portion 1 and dark portion d and presentation mode substantially
The rote learning that different two column split image DL1 (t1), DLk (t (1- (k-1))) are carried out as a result, to identify defect D's
Classification, thus, it is possible to further increase the accuracy of identification of defect D.
It this concludes the description of embodiments of the present invention, but the present invention is not limited to above embodiment, it can be with various sides
Formula is implemented.For example, in the above-described embodiment, it is illustrated centered on the case where by check object T being film, but of the invention
Defect inspecting system and defect detecting method for example can be suitable for being filled in the loading inspection of the liquid of container in production line
It looks into.By the defect inspecting system 1 and defect detecting method of present embodiment, it is able to detect liquid and does not reach institute's phase in container
The position of prestige or liquid are less than the defects of desired position in container.
In addition, the defect inspecting system 1 and defect detecting method of present embodiment can be suitable for glass in production line
The visual examinations such as fracture, the scar of product etc..In the case where glass product has the defects of fracture, scar, can utilize bright
This case higher than other positions is spent to extract defect.
Claims (6)
1. a kind of defect inspecting system, which is characterized in that have:
Light source, to check object irradiation light;
Image pickup part shoots two dimensional image by discrete time, which is based on shining from the light source to the check object
Penetrate and penetrate the check object or in the check object reflect after the light and formed;
Delivery section relatively conveys the check object relative to the light source and the image pickup part along conveying direction;
And
Image processing part handles the image data for the two dimensional image shot by the image pickup part,
The image pickup part is shot in the changed with brightness on the consistent direction of the conveying direction of the two dimensional image
The two dimensional image,
Described image processing unit includes
The two dimensional image is processed into the described image data of column split image, the column split figure by column split processing unit
As by the way that the two dimensional image to be divided into multiple column along conveying direction arrangement and is made by the image pickup part by described
The two dimensional image that discrete time is shot respectively in same position the Leie according to time series sequence arrangement and
At;And
Defect classification identification part, the identification based on the classification to the defect for being included with more than two column split images
The data that the result of relevant rote learning is accumulated, come identify the check object defect classification, wherein two
A above column split image is image obtained from being handled as the column split processing unit.
2. defect inspecting system according to claim 1, wherein
Defect classification identification part is lacked based on included to two column split images different from 10% or more brightness
The data that the result of the relevant rote learning of identification of sunken classification is accumulated, to identify the defect of the check object
Classification.
3. defect inspecting system according to claim 1 or 2, wherein
The defect inspecting system is also equipped with occulter, and the occulter is and right between the light source and the check object
A part of the light irradiated from the light source to the check object is blocked, thus discrete being pressed by the image pickup part
Bright portion and dark portion are formed on the two dimensional image of time shooting,
The delivery section by the check object relative to the light source, the occulter and the image pickup part along with stated clearly
The conveying direction of portion and the intersection of the line of demarcation of the dark portion relatively conveys,
Defect classification identification part is based on to the relevant machine of the identification of classification of defect for being included to following column split image
The data that are accumulated of result of tool study, come identify the check object defect classification, the column split image
Refer to: making column point made of sequence arrangement of the Leie of the position in stated clearly the portion in the two dimensional image according to time series
Cut image;And arrange the Leie of the position of the dark portion in the two dimensional image according to the sequence of time series
Column split image.
4. a kind of defect detecting method characterized by comprising
From the light source of defect inspecting system to the irradiation process of check object irradiation light;
The camera shooting process of two dimensional image is shot by discrete time by the image pickup part of the defect inspecting system, wherein the two dimension
Image is based on irradiating and penetrate the check object from the light source to the check object in the irradiation process or in institute
It states the light after reflecting in check object and is formed;
By the delivery section of the defect inspecting system by the check object relative to the light source and the image pickup part along defeated
Send the conveying operation that direction relatively conveys;And
Figure by the image processing part of the defect inspecting system to the two dimensional image shot in the camera shooting process
As the image procossing process that data are handled,
In the camera shooting process, shoots and occur in the two dimensional image with brightness on the consistent direction of the conveying direction
The two dimensional image of variation,
Include: in described image treatment process
The two dimensional image is processed into the column split treatment process of the described image data of column split image, wherein the column
The two dimensional image by being divided into multiple column along conveying direction arrangement and being made in the camera shooting work by segmented image
The two dimensional image shot in sequence by the discrete time respectively in same position the Leie according to time series
Sequence arranges;And
Based on to the relevant rote learning of the identification of classification of defect for being included to more than two column split images
As a result the data accumulated, come identify the check object defect classification defect classification identify process, wherein
More than two column split images are the images handled in the column split treatment process.
5. defect detecting method according to claim 4, wherein
In defect classification identification process, based on being wrapped to the two column split images different from 10% or more brightness
The data that the result of the relevant rote learning of identification of the classification of the defect contained is accumulated, to identify the check object
Defect classification.
6. defect detecting method according to claim 4 or 5, wherein
In the irradiation process, by occulter in the two dimensional image shot by the camera shooting process by discrete time
Upper to form bright portion and dark portion, the occulter is positioned at light source and check object between, and to from the light source to the inspection pair
As a part of the light of irradiation is blocked,
In the conveying operation, by the check object relative to the light source, the occulter and the image pickup part along
The conveying direction intersected with the line of demarcation in stated clearly portion and the dark portion relatively conveys,
In defect classification identification process, the identification phase based on the classification to the defect for being included with following column split image
The data that the result of the rote learning of pass is accumulated, come identify the check object defect classification, it is described column point
It cuts image to refer to: by arranging the Leie of the position in stated clearly the portion in the two dimensional image according to the sequence of time series
Made of column split image;And by making the Leie of the position of the dark portion in the two dimensional image shine time series
Sequence arrangement made of column split image.
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
JP2017142923A JP6970549B2 (en) | 2017-07-24 | 2017-07-24 | Defect inspection system and defect inspection method |
JP2017-142923 | 2017-07-24 |
Publications (2)
Publication Number | Publication Date |
---|---|
CN109297973A true CN109297973A (en) | 2019-02-01 |
CN109297973B CN109297973B (en) | 2022-12-23 |
Family
ID=65172600
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810800087.9A Active CN109297973B (en) | 2017-07-24 | 2018-07-19 | Defect inspection system and defect inspection method |
Country Status (4)
Country | Link |
---|---|
JP (1) | JP6970549B2 (en) |
KR (1) | KR20190011199A (en) |
CN (1) | CN109297973B (en) |
TW (1) | TWI788387B (en) |
Families Citing this family (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
DE112020006158T5 (en) | 2019-12-16 | 2022-09-29 | Mitsubishi Electric Corporation | Image good/bad learning device, image good/bad determining device and image reading device |
KR102372714B1 (en) * | 2020-01-31 | 2022-03-10 | 한국생산기술연구원 | Automatic defect inspection system based on deep learning |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JPH08220021A (en) * | 1995-02-10 | 1996-08-30 | Central Glass Co Ltd | Defect detecting method for transparent plate-shaped body |
JP2004163416A (en) * | 2002-10-23 | 2004-06-10 | Nippon Steel Corp | Defect inspection device |
CN1520512A (en) * | 2001-07-05 | 2004-08-11 | 日本板硝子株式会社 | Method and device for inspecting defect of sheet-shaped transparent body |
CN102630299A (en) * | 2009-10-30 | 2012-08-08 | 住友化学株式会社 | Image processing device for defect inspection and image processing method for defect inspection |
CN104956210A (en) * | 2013-01-30 | 2015-09-30 | 住友化学株式会社 | Image generating device, defect inspecting device, and defect inspecting method |
Family Cites Families (13)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP3476913B2 (en) * | 1994-07-08 | 2003-12-10 | オリンパス株式会社 | Defect type determination device and process management system |
JPH08247954A (en) * | 1995-03-15 | 1996-09-27 | Asahi Glass Co Ltd | Measuring method of optical distortion |
JP4220595B2 (en) * | 1998-08-10 | 2009-02-04 | 株式会社日立製作所 | Defect classification method and teaching data creation method |
JP2001134763A (en) * | 1999-11-09 | 2001-05-18 | Hitachi Ltd | Method for sorting defect on basis of picked-up image and method for displaying the result |
US7105848B2 (en) * | 2002-04-15 | 2006-09-12 | Wintriss Engineering Corporation | Dual level out-of-focus light source for amplification of defects on a surface |
JP2004191283A (en) * | 2002-12-13 | 2004-07-08 | Ricoh Co Ltd | Defect detecting method and defect detection device |
WO2004104566A1 (en) * | 2003-05-19 | 2004-12-02 | Micro-Epsilon Messtechnik Gmbh & Co. Kg | Method and device for optically controlling the quality of objects having a preferably circular edge |
JP2004354250A (en) * | 2003-05-29 | 2004-12-16 | Nidek Co Ltd | Defect inspection device |
JP4694914B2 (en) * | 2005-08-03 | 2011-06-08 | 昭和電工株式会社 | Surface inspection method and apparatus |
JP4739044B2 (en) * | 2006-02-09 | 2011-08-03 | 三菱原子燃料株式会社 | Appearance inspection device |
EP2177898A1 (en) * | 2008-10-14 | 2010-04-21 | Fraunhofer-Gesellschaft zur Förderung der angewandten Forschung e.V. | Method for selecting an optimized evaluation feature subset for an inspection of free-form surfaces and method for inspecting a free-form surface |
KR20170107952A (en) * | 2015-01-29 | 2017-09-26 | 가부시키가이샤 데크시스 | Optical appearance inspection device and optical appearance inspection system using same |
US10359371B2 (en) * | 2015-08-24 | 2019-07-23 | Kla-Tencor Corp. | Determining one or more characteristics of a pattern of interest on a specimen |
-
2017
- 2017-07-24 JP JP2017142923A patent/JP6970549B2/en active Active
-
2018
- 2018-07-17 KR KR1020180082656A patent/KR20190011199A/en not_active Application Discontinuation
- 2018-07-19 CN CN201810800087.9A patent/CN109297973B/en active Active
- 2018-07-20 TW TW107125176A patent/TWI788387B/en active
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JPH08220021A (en) * | 1995-02-10 | 1996-08-30 | Central Glass Co Ltd | Defect detecting method for transparent plate-shaped body |
CN1520512A (en) * | 2001-07-05 | 2004-08-11 | 日本板硝子株式会社 | Method and device for inspecting defect of sheet-shaped transparent body |
JP2004163416A (en) * | 2002-10-23 | 2004-06-10 | Nippon Steel Corp | Defect inspection device |
CN102630299A (en) * | 2009-10-30 | 2012-08-08 | 住友化学株式会社 | Image processing device for defect inspection and image processing method for defect inspection |
CN104956210A (en) * | 2013-01-30 | 2015-09-30 | 住友化学株式会社 | Image generating device, defect inspecting device, and defect inspecting method |
Also Published As
Publication number | Publication date |
---|---|
CN109297973B (en) | 2022-12-23 |
JP2019023588A (en) | 2019-02-14 |
TWI788387B (en) | 2023-01-01 |
KR20190011199A (en) | 2019-02-01 |
TW201908718A (en) | 2019-03-01 |
JP6970549B2 (en) | 2021-11-24 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109297972A (en) | Defect inspecting system and defect detecting method | |
US20130128026A1 (en) | Image processing device for defect inspection and image processing method for defect inspection | |
CN107462580A (en) | Defect inspecting system, film manufacturing device and defect detecting method | |
JP5274622B2 (en) | Defect inspection apparatus and method | |
TWI457556B (en) | Surface inspection method and apparatus for steel plate with resin coating film | |
CN107735674A (en) | The manufacture method of surface defect detection apparatus, detection method of surface flaw and steel | |
CN102253053B (en) | Appearance inspection device | |
KR102246301B1 (en) | Appearance inspection apparatus | |
CN109297971A (en) | Defect inspecting system and defect detecting method | |
CN109297973A (en) | Defect inspecting system and defect detecting method | |
JPWO2012153662A1 (en) | Inspection method for minute defects of translucent plate and inspection apparatus for minute defects of translucent plate | |
TW201140043A (en) | End face inspection method for light-pervious rectangular sheets and end face inspection apparatus | |
JP2018146251A (en) | Foreign matter detection system, foreign matter detection method and program thereof | |
CN110412035A (en) | A kind of high reflecting material surface inspecting method and system | |
JP7151469B2 (en) | Sheet defect inspection device | |
KR20140089200A (en) | Method of detecting defect and reflective optical inspection device using the same | |
CN108204982A (en) | Multi-view image capturing device and multi-view image detecting equipment | |
KR20170129077A (en) | Transmissive optical inspection device and method of detecting defect using the same | |
CN102235982A (en) | End face inspection method and end face inspection device of light-transmitting rectangular plate-shaped object | |
JP2005070012A (en) | Foreign substance inspection method of liquid filled in container and its device | |
US20210100256A1 (en) | Foreign Material Inspection System | |
JP2016205929A (en) | Foreign substance discrimination device and foreign substance discrimination method | |
Ramanathan et al. | Development of an Intelligent Robotized Machine Vision Automated System for Bacterial Growth Monitoring | |
DE102014225775B4 (en) | Method and device for testing a correct adhesive application | |
JP2012163533A (en) | Defect inspection apparatus and method for lenticular sheet |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |