CN107305190A - Defect detecting method and defect inspecting system - Google Patents
Defect detecting method and defect inspecting system Download PDFInfo
- Publication number
- CN107305190A CN107305190A CN201710240030.3A CN201710240030A CN107305190A CN 107305190 A CN107305190 A CN 107305190A CN 201710240030 A CN201710240030 A CN 201710240030A CN 107305190 A CN107305190 A CN 107305190A
- Authority
- CN
- China
- Prior art keywords
- defect
- image
- pixel
- region
- coloured light
- 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
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/84—Systems specially adapted for particular applications
- G01N21/88—Investigating the presence of flaws or contamination
-
- 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
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/90—Determination of colour characteristics
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/84—Systems specially adapted for particular applications
- G01N21/88—Investigating the presence of flaws or contamination
- G01N21/8851—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges
- G01N2021/8887—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges based on image processing techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10024—Color image
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30168—Image quality inspection
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Theoretical Computer Science (AREA)
- Analytical Chemistry (AREA)
- Life Sciences & Earth Sciences (AREA)
- Chemical & Material Sciences (AREA)
- Health & Medical Sciences (AREA)
- Biochemistry (AREA)
- General Health & Medical Sciences (AREA)
- Immunology (AREA)
- Pathology (AREA)
- Quality & Reliability (AREA)
- Signal Processing (AREA)
- Investigating Materials By The Use Of Optical Means Adapted For Particular Applications (AREA)
- Image Processing (AREA)
- Image Analysis (AREA)
Abstract
A kind of defect detecting method and defect inspecting system, the defect of checked property is easily and securely detected by simply constituting.Defect detecting method includes：The process for checked property (W) shoot obtaining shooting image；Two kinds of color components in red, green, blue are regard as the process on the two-dimensional distribution of the longitudinal axis and transverse axis with the color component of the pixel of shooting image is configured at.On two-dimensional distribution, the pixel in defect distribution region and the pixel in noise profile region are divided into, the pixel in be partitioned into defect distribution region is selected.The pixel selected is applied to the original location of pixel to generate restriction shooting image, uses and limits shooting image progress defect inspection.
Description
Technical field
Present embodiment is directed to use with the defect inspection side that image processing techniques is checked come the defect to shooting image
Method and defect inspecting system, the defect more particularly to easily distinguished in image are detected with noise to real defect
Defect detecting method and defect inspecting system.
Background technology
All the time, as being shot using filming apparatus such as CCD cameras to the surface of checked property and be based on the bat
The defect detecting method for taking the photograph image to check defect, it is known to various methods.
However, there are the following problems for any defect detecting method：The image procossing of complexity is needed, and can not be so clear
Defect is detected clearly.
The content of the invention
Problems to be solved by the invention
Present embodiment considers this point and completed, and its object is to can pass through simple structure there is provided one kind
Into the defect detecting method and defect inspecting system of the easily and securely defect of detection checked property.
Technical teaching for solving the problem was
Present embodiment is a kind of defect detecting method, it is characterised in that including：Light shine the surface of checked property
Illumination process；The surface of checked property shoot so as to obtain the shooting process of shooting image；Carried from shooting image
Taking out detect the defect distribution region of defect and need not detect the noise profile region of defect, and by each region
The color component of pixel is configured on using two kinds of color components in red, green, blue as the two-dimensional distribution of the longitudinal axis and transverse axis
Arrangement step；The pixel in defect distribution region will be constituted by segmentation straight line on two-dimensional distribution and noise profile region is constituted
The segmentation process that pixel is separated；Select to be divided on two-dimensional distribution in two regions that line segmentation is opened and belong to scarce
Fall into the choice process of the pixel of distributed areas side；The pixel selected in process is selected is applied to the pixel of this in shooting image
The location of originally generate the restriction process for limiting shooting image；And perform defect inspection using shooting image is limited
Inspection perform process.
Present embodiment is a kind of defect detecting method, it is characterised in that just white light.
Present embodiment is a kind of defect detecting method, it is characterised in that the just light of different two or more colors.
Present embodiment is a kind of defect detecting method, it is characterised in that just the 1st coloured light, the 2nd color of different colours
Light, the 3rd coloured light.
Present embodiment is a kind of defect detecting method, it is characterised in that the 1st coloured light, the 2nd coloured light, the 3rd coloured light are respectively
Any one in red, green, blue.
Present embodiment is a kind of defect detecting method, it is characterised in that in illumination process, to checked property irradiation the
1 coloured light, the 2nd coloured light, the height difference of the position of the 3rd coloured light.
Present embodiment is a kind of defect detecting method, it is characterised in that be set to y, transverse axis in the longitudinal axis of two-dimensional distribution
It is set to x and when a and b are set to real number, segmentation straight line is represented by expression of first degree y=ax+b.
Present embodiment is a kind of defect inspecting system, it is characterised in that possessed：Light shine the surface of checked property
Lighting device；The surface of checked property shoot so as to obtain the filming apparatus of shooting image；And to from shooting
The shooting image real-time image processing of device carries out the flaw detection apparatus of defect inspection, and flaw detection apparatus possesses：Configuration
Portion, it extracts the defect distribution region that detect defect and the noise profile area that need not detect defect from shooting image
Domain, and the color component of the pixel in each region is configured at using two kinds of color components in red, green, blue is used as the longitudinal axis and horizontal stroke
On the two-dimensional distribution of axle；Cutting part, it will constitute the pixel in defect distribution region on two-dimensional distribution by splitting straight line
Separated with the pixel for constituting noise profile region；Choice portion, it selects the divided line segmentation on two-dimensional distribution to open
Two regions in the pixel for belonging to defect distribution area side；Limited section, it should by the pixel selected in process is selected
Use the original location of the pixel of this in shooting image and generate and limit shooting image；And enforcement division is checked, it uses limit
Determine shooting image to perform defect inspection.
Present embodiment is a kind of defect inspecting system, it is characterised in that just white light.
Present embodiment is a kind of defect inspecting system, it is characterised in that the just light of different two or more colors.
Present embodiment is a kind of defect inspecting system, it is characterised in that just the 1st coloured light, the 2nd color of different colours
Light, the 3rd coloured light.
Present embodiment is a kind of defect inspecting system, it is characterised in that the 1st coloured light, the 2nd coloured light, the 3rd coloured light are respectively
Any one in red, green, blue.
Present embodiment is a kind of defect inspecting system, it is characterised in that in illumination process, to checked property irradiation the
1 coloured light, the 2nd coloured light, the height difference of the position of the 3rd coloured light.
Present embodiment is a kind of defect inspecting system, it is characterised in that the longitudinal axis of two-dimensional distribution is being set into y, horizontal stroke
Axle is set to x and when a and b are set to real number, and segmentation straight line is represented by expression of first degree y=ax+b.
The effect of invention
Present embodiment as according to more than, can easily and securely detect checked property by simply constituting
Defect.
Brief description of the drawings
Fig. 1 is the flow chart of the defect detecting method of present embodiment.
Fig. 2 (a), (b), (c) are the explanation figures of the principle of present embodiment.
Fig. 3 is the explanation figure of the principle of present embodiment.
Fig. 4 is the explanation figure of the principle of present embodiment.
Fig. 5 is the explanation figure of the principle of present embodiment.
Fig. 6 is the stereogram of chip shape electronic unit.
Fig. 7 (a), (b) are the lighting device of illumination light to be irradiated to chip shape electronic unit and to chip shape electronic unit
The explanation figure for the filming apparatus that surface is shot.
Fig. 8 is the flow chart of the defect detecting method of comparative example 1.
Fig. 9 (a), (b), (c), (d) are from the perspective of for obtaining red image, green image, blue image from shooting image
Bright figure.
Figure 10 (a), (b), (c), (d), (e), (f), (g), (h), (i) are the images of the defect detecting method of comparative example 1
The explanation figure of processing.
Figure 11 (a), (b) are to irradiate the lighting device of illumination light to chip shape electronic unit and come using the lighting device
To the explanation figure used in the shooting image of chip shape electronic unit.
Figure 12 is the explanation figure of the image procossing of the defect detecting method of comparative example 2.
Figure 13 is the explanation figure of the image procossing of the defect detecting method of comparative example 3.
Description of reference numerals
1 framework
1b big openings portion
The small opening portions of 1n
1s inner concave surfaces
2a light in part
10a, 10b lighting device
20 filming apparatus
Dd defect distributions region
Dn noise profiles region
The high position illuminations of LH
LI, LIw incident light
A LIH high positions put incident light
LIL lower position incident lights
LIM centre positions incident light
LM centre positions are illuminated
LL lower positions are illuminated
LR reflected lights
Lw white light-emitting diodes
W workpiece
Embodiment
Embodiment
Hereinafter, the embodiment referring to the drawings to defect detecting method and defect inspecting system is illustrated.First, according to
Fig. 6 is to being that the chip shape electronic unit (hereinafter also referred to " workpiece ") of checked property is illustrated as check object.Such as Fig. 6 institutes
Show, workpiece W is hexahedral shape, is had：The main body Wd formed by insulator；With the two ends for the longer direction for being formed at main body Wd
The electrode Wa, the Wb that are formed by electric conductor in portion.
The elements such as resistance, electric capacity have been internally formed in main body Wd, and element and external circuit are connected through electrode Wa, Wb
External circuit is connected to carry out.
Then, the defect detecting method on the surface to workpiece W is illustrated.First, the white of illumination light will be used as
Coloured light is irradiated on workpiece W, and its surface is shot.
The filming apparatus that Fig. 7 shows the lighting device of irradiation illumination light and shot to the surface of workpiece.Here, figure
7 (a) is the stereogram observed by lighting device from lower surface, in addition, Fig. 7 (b) is to represent to use lighting device and bat
Take the photograph the side view for the situation that device is shot come the surface to workpiece.
In Fig. 7 (a), lighting device 10a has the framework 1 for being formed as semi-spherical shape, the central side of the hemisphere of the framework 1
With the lower surface 1d for being formed as plane.Moreover, from lower surface 1d, the hemisphere in the inner side as framework 1 is internally formed
The inner concave surface 1s of the semi-spherical shape same with the profile of framework 1.In addition, foring the central side to hemisphere in framework 1
That is the big opening portion 1b of lower surface 1d openings.In addition, in the inner concave surface 1s of framework 1 from big opening portion 1b sides, in circle
The ground of central part insertion framework 1 of the framework 1 of shape forms the diameter circular small opening portion 1n shorter than big opening portion 1b.Here,
Small opening portion 1n's in Fig. 7 (a) is shaped as circle, and still, small opening portion 1n shape is not limited to circle, as long as energy
Enough shapes for making the reflected light from workpiece pass through framework 1 as aftermentioned.Internally concave surface 1s surface is in concentric circles
Multiple white light-emitting diodes (LED) Lw of white light can be irradiated shape to big opening portion 1b by being arranged at equal intervals.This is concentric
Round center is consistent with small opening portion 1n center, and the quantity of concentric circles is total up to seven.
In Fig. 7 (b), lighting device 10a is configured to：So that lower surface 1d and big opening portion 1b are downward so that
Small opening portion 1n is upward.Moreover, big opening portion 1b center substantially directly below from the lower surface 1d of framework 1 every
The position for the workpiece W for the distance of very little and being placed in mounting table (not shown) is：Workpiece W one side Wu and big opening portion
1b is relative.
On the other hand, in lighting device 10a surface, filming apparatus 20 makes light in part 2a and small opening portion 1n relatively
It is configured at small opening portion 1n top.Filming apparatus 20 is taken into from the light of workpiece W outgoing through light in part 2a, thus, it is possible to
Access shooting image.
Then, flaw detection apparatus 20A is sent to by shooting image obtained from filming apparatus 20, in the defect inspection
Defect inspection is performed in device 20A.In addition, being connected to monitor (monitor) 20B in flaw detection apparatus 20A.
In this case, defect is constituted by lighting device 10a, filming apparatus 20, flaw detection apparatus 20A, monitor 20B
Inspection system.
Then shooting image, for one side Wu that workpiece W is obtained such as Fig. 7 (b) Suo Shi and based on shooting image come pair
The step of defect detecting method that defect is detected, illustrated using Fig. 1 flow chart.First, such as the S101 in Fig. 1
(illumination process) like that, using the white light-emitting diode Lw for the lighting device 10a being configured at shown in Fig. 7 (b), white light is shone
It is mapped on workpiece W.Now, workpiece W one side Wu is irradiated to from the incident light LI of white light-emitting diode Lw outgoing, and it is anti-herein
Penetrate and turn into reflected light LR.Moreover, reflected light LR is placed through the small opening portion 1n of the framework 1 of workpiece W top, from light in part
2a is taken into filming apparatus 20.Thus, as shown in Fig. 1 S102 (shooting process), workpiece W one side Wu (upper surface) is clapped
Take the photograph to obtain shooting image.
Then, flaw detection apparatus 20A is sent to by shooting image obtained from filming apparatus 20, in the defect inspection
Defect inspection is performed in device 20A.
That is, for by shooting image obtained from filming apparatus 20, as S33 (abstraction process), from the shooting figure
The defect distribution region that detect defect and the noise profile region that defect need not be detected are extracted as in.Fig. 2 (a) is shown
The situation.Image shown in Fig. 2 (a) is the image shown by monitor 20B, is the workpiece W obtained in Fig. 1 S102
Shooting image.Image shown by operating personnel one side visual monitoring device 20B, while extracting detect the defect point of defect
Cloth region is simultaneously marked.Here, extracting the region Ae surrounded by chain-dotted line as defect distribution region and being marked.And
And, the Zone Full outside the Ae of region, which turns into, need not detect the noise profile region of defect.
So divide behind two kinds of regions, then implement Fig. 1 S34 (arrangement step).Here, by shooting image
The color component of whole pixels in defect distribution region and noise profile region is configured at the longitudinal axis and takes blue component, transverse axis to take red
On the two-dimensional distribution of component.
Fig. 2 (b) shows the result for the S34 that Fig. 1 is implemented to Fig. 2 (a) shooting image.In Fig. 2 (b), by dot-dash
The region that line Dd is surrounded is：The pixel that defect distribution region in pie graph 2 (a) is Ae is configured on the two-dimensional distribution and
Obtained region.In addition, in Fig. 2 (b), being by the chain-dotted line Dn regions surrounded：By the noise profile area in pie graph 2 (a)
The pixel in domain is configured at region obtained from the two-dimensional distribution.Here, in Fig. 2 (b), constituting defect distribution region Dd's
Pixel belongs to region Ae in Fig. 2 (a), therefore the symbol immediately above using set of the two-dimensional distribution in Fig. 2 (b) is designated as Dd
={ Ae }.
After the S34 for so terminating Fig. 1, then on the two-dimensional distribution being made up of the S34, such as S35 (segmentation works
Sequence) shown in, pixel of the pixel in defect distribution region with constituting noise profile region will be constituted by straight line (segmentation straight line)
Separate.Fig. 2 (b) is while be also shown for the situation of the segmentation.In Fig. 2 (b), defect distribution region Dd and noise profile region
Dn shape is the ellipse for having major axis in substantially upper right.In the case that there is such shape in two regions,
Consider that described two regions are in the 1st quadrant being made up of the positive direction of the longitudinal axis (B axle) and the positive direction of transverse axis (R axles), draw
The straight line to upper right side that defect distribution region Dd as shown in Fig. 2 (b) and noise profile region Dn are separated.If this is straight
The slope of line is set to a, the straight line is set into b with the coordinate that the longitudinal axis (B axle) intersects, then by B and R expression formula
B=aR+b (1)
Represent straight line.Wherein b is the threshold value for splitting defect distribution region Dd and noise profile region Dn.Now, table
The formula for showing i.e. defect distribution region Dd more upper than straight line is：
B ＞ aR+b (2)
I.e.
B-aR-b ＞ 0 (3).
It was found from Fig. 2 (b), the whole pixels for constituting noise profile region Dn are eliminated in the region of formula (3).
Then, as Fig. 1 S36 (choice process), two separated in Fig. 2 (b) by straight line are selected
It is in region, belong to not include belong to noise profile region pixel area side whole pixels, that is, select defect distribution
Whole pixels in region Dd (formula (3)).
Then, then as shown in Fig. 1 S37 (restriction process), by the composition defect distribution region Dd's so selected
The original location of the pixel that pixel is applied in shooting image, generation is only defined in the defect distribution region Dd selected
The restriction shooting image of interior pixel.
Fig. 2 (c) shows the restriction shooting image generated in the S37 based on Fig. 2 (b).The restriction shooting image by
Monitor 20B is shown.In the restriction shooting image that Fig. 2 (c) is obtained, the pixel outside the Ae of region not including Fig. 2 (a).Cause
This, after Fig. 1 S37 process terminates, using the restriction shooting image by Fig. 2 (c) obtained from S37, such as S38 (is checked
Perform process) like that by operating personnel one side visual monitoring device 20B, while carrying out defect inspection.By in such manner, it is possible to remove structure
Into Fig. 2 (b) noise profile region Dn pixel implement defective inspection.
In addition, in S35 (segmentation process), in order to determine straight line, coming with the visual two-dimensional distribution of operating personnel
The method of decision and the method for being used as image processing software using existing known discriminant analysis.Use the feelings of the discriminant analysis
Straight line under condition is for example determined as：It is equal straight according to the other mahalanobis distance of any sort in defect classification and noise class
Line.
Here, being illustrated using Fig. 3 to the principle of the invention shown in S33 to S37 in Fig. 1.Fig. 3 is to represent the present invention
Principle Fig. 2 (b) schematic diagram.It is recited as：The defect distribution region Dd surrounded by solid line and the noise point surrounded by dotted line
Cloth region Dn is in the positive direction and transverse axis (R by the longitudinal axis (B axle) as the ellipse in substantially upper right with major axis
Axle) positive direction constitute 1st quadrant.
In figure 3, it is considered to which the straight line shown in the origin dotted line of 45 ° of inclined upward to the right is passed through as B=R straight line.If
For：The straight line (B=R+BT) of solid line obtained from only moving threshold value BT (positive number) to the positive direction of the longitudinal axis by making the straight line, from
And defect distribution region Dd can be separated with noise profile region Dn.Noise profile region Dn is not at the upside of the straight line i.e.
B ＞ R+BT scope.Therefore, the defect distribution region Dd pixel of the scope in B ＞ R+BT is constituted it may be said that entirely table
Show the pixel of real defect.That is, the scope of B ＞ R+BT is the region of the object as defect inspection.Select in the region
Whole pixels, and the pixel that is applied to as the S37 in Fig. 1 in shooting image is the location of original, thus, it is possible to
Generation limits shooting image.Moreover, the restriction shooting image so generated does not include the noise profile region Dn's in Fig. 3 completely
Pixel.Therefore, operating personnel can one side visual monitoring device 20B, while using limit shooting image implement defective
Check.
So, according to present embodiment, defect inspection can be easily and securely performed without complicated image procossing
Look into.
In addition, in order that the explanation of the above simply, will split defect distribution region Dd and noise profile region Dn straight line
Gradient be set to 45 °, still, according to actual defect distribution region Dd and noise profile region Dn shape, the gradient
Generation various change.For example, the slope of the straight line shown in Fig. 2 (b) is a, if a takes the value beyond 1, the gradient of straight line is no longer
It is -45 °.
In the above description, to being configured at Fig. 7 (a), the LED of lighting device 10a shown in (b) said for white
It is bright.But, in Fig. 7 (a), (b), the light for being irradiated to workpiece W is not limited to white.In the presence of the three primary colors used as light
Red, green, blue each LED as other examples lighting device.Figure 11 shows the lighting device and the surface to workpiece
The filming apparatus shot.Figure 11 (a) is the stereogram observed by lighting device from lower surface.In addition, Figure 11 (b)
It is the side view of situation for representing to shoot the surface of checked property using the lighting device and filming apparatus.Figure 11 (a),
(b) LED color is only that with Fig. 7 (a), the difference of (b), therefore, is omitted on not having the detailed description of discrepant part.
In Figure 11 (a), lighting device 10b inner concave surface 1s surface concentric circles be arranged light emitting diode
(LED).The quantity of concentric circles is total up to seven.Arranging red at substantially equal intervals on big opening portion 1b four concentric circles
The LED of color (R), constitutes lower position illumination LL.Similarly, since concentric with the 6th on the 5th concentric circles big opening portion 1b sides
The LED of green (G) is arranged on circle at substantially equal intervals, centre position illumination LM is constituted.Moreover, since big opening portion 1b sides
The 7th concentric circles on arranging blue (B) at substantially equal intervals on the concentric circles closest to small opening portion 1n position
LED, constitute high position illumination LH.
In Figure 11 (b), lighting device 10b is configured to：Lower surface 1d and big opening portion 1b are below, makes small open
Oral area 1n is in top.Moreover, big opening portion 1b center lower surface 1d substantially directly below from framework 1 across very
Small distance and the chip-type electronic component (work for being placed in the hexahedral shape as checked property in mounting table (not shown)
Part) W position is：Make workpiece W one side Wu relative with big opening portion 1b.
Use the flow chart in the case of Figure 11 lighting device 10b and filming apparatus 20 identical with Fig. 1.Here, right
The effect that lighting device 10b configures three kinds of LED of red, green, blue and produced is illustrated.
In Figure 11 (b), workpiece W one side Wu is irradiated to from the incident light LI of each illumination (LED) outgoing.Now, because low
In position illumination LL, centre position illumination LM, high position illumination LH position, incident light LI LL outgoing is illuminated from lower position
Lower position incident light LIL is irradiated to workpiece W one side Wu from low angle.Similarly, being illuminated from centre position in incident light LI
One side Wus of the centre position incident light LIM of LM outgoing from the angular illumination higher than lower position incident light LIL to workpiece W.Enter one
Step, the high position for illuminating LH outgoing from high position in incident light LI put incident light LIH from subvertical high angular illumination to
Workpiece W one side Wu.So, incident light LI is made up of three kinds of different incident lights of incidence angle.Moreover, incident light LI is in workpiece
W one side Wu reflects and turns into reflected light LR, is shooting image as obtained from filming apparatus 20 shoots reflected light LR.Cause
This, the three kinds of monochrome images obtained from shooting image are that the light of the color as each monochrome image is incident from different height (angle)
To workpiece W Wu and this while Wu reflection obtained from monochrome image.That is, each monochrome image performance is with one side Wu's
The corresponding intrinsic reflection characteristic of shape.The characteristic is referred to as angular response, can select to be considered most show workpiece W one side
The monochrome image of the angular response of Wu defect or as image obtained from the operation result between each monochrome image, is used as inspection
Image (Fig. 8 S105 or S106) is looked into, to carry out defect inspection.All it can utilize such angular response in use all the time
Illumination configuration.
In addition, when being illustrated using Fig. 3 to the principle of the present invention, defect distribution region Dd and noise profile region Dn
Without common ground, therefore, it is possible to by the complete Ground Splits of straight line B=R+BT.But, using the shooting different from Fig. 2 (a)
In the case of two-dimensional distribution as Fig. 2 (b) is made in image, defect distribution region Dd also has sometimes with noise profile region Dn
There is common ground.The example of the dividing method in two regions in this case is illustrated using Fig. 4 and Fig. 5.
Fig. 4 is the two-dimensional distribution that some shooting image is made.Defect distribution region Dd and noise profile region Dn has
There is common ground Dc.Therefore, it is impossible to using straight line by defect distribution region Dd and the complete Ground Splits of noise profile region Dn
Open.Herein, it is contemplated that following situation：The region for the object for turning into defect inspection as constituting is the pixel for limiting shooting image, i.e.,
Make that the whole pixels for belonging to defect distribution region Dd can not be selected, it is considered to remove in the region preferably from the object as defect inspection
Go to belong to whole pixels in noise profile region.In this case, by the straight line B=R+BTa shown in Fig. 4 by common ground Dc
Removed together with the Dn of noise profile region from the object of choice.Then, as long as only selecting in the Dd of defect distribution region
Straight line upside region be B ＞ R+BTa region pixel come generate limit shooting image.Then, in contrast,
Contemplate following situation：Even if the whole pictures for belonging to noise profile region can not be removed from the region of the object as defect inspection
Element, is the pixel for limiting shooting image as the region of object of the composition as defect inspection, it is considered to preferably select and belong to scarce
Fall into distributed areas Dd whole pixels.In this case, by the straight line B=R+BTb shown in Fig. 5 select common ground Dc and
Defect distribution region Dd.Then, as long as only selecting the region i.e. B ＞ R+ of the upside of the straight line in the Dn of noise profile region
The pixel in BTa region limits shooting image to be made.
In addition, the establishing method for the judgment standard split as following segmentation as to carrying out Fig. 4 or progress Fig. 5
One illustrate.For example, it is contemplated that：The characteristics such as the shape of defect having to the shooting image due to certain workpiece and
By probability P 1 that rejected region false judgment is normal portions with being rejected region by normal portions false judgment during defect inspection
Probability P 2 be compared and consider to set benchmark.In the case of the defect of the characteristic with P1 ＞ P2, it is considered to if as scarce
Fall into the region of the object checked includes many defect distribution regions as far as possible, then can increase not by the number of the defect of false judgment
Amount.In this case, split as Fig. 5.In addition, in the case of the defect of the characteristic with P1 ＜ P2, examining
Many noise profile regions are removed as far as possible from the region of the object as defect inspection if considering, and can be increased and not sentenced by mistake
The quantity of disconnected normal portions.In this case, split as Fig. 4.
In the above description, the two-dimensional distribution in the present invention is made on blueness and red, but it is also possible to
The distribution map or red and green two-dimensional distribution of green and blueness is made.In this case, from the distribution map being made
Selection can be most readily by the distribution map that straight line opens defect distribution region and noise profile region segmentation.Then, only
Use the distribution map to select the pixel in defect distribution region and limit shooting image to be made.By two-dimensional distribution
When the longitudinal axis is set to y, transverse axis is set to x and a and b are set into real number, straight line now is generally by expression of first degree y=ax+b tables
Show.In the above-described embodiment, the longitudinal axis is blue (B), and transverse axis is red (R), therefore the expression of first degree is carried out as follows expression.
B=aR+b (1)
In addition, in the above description, it is contemplated to which a region only in shooting image is as defect distribution region, still
It is also contemplated that being used as defect distribution region by more than two multiple regions formed in shooting image.
In addition, in the above description, illustrating following situations：Defect in two-dimensional distribution is split by straight line
When distributed areas are with noise profile region, either one in two regions opened by line segmentation includes whole defect distribution regions
Or at least one party in whole noise profile regions.But, the segmentation based on straight line is not limited to this.For example can also be：
In Fig. 4 or Fig. 5, draw straight line the common ground in defect distribution region and noise profile region is divided further into two
Point.
In addition, in the above description, illustrating following situations：The lower position illumination being irradiated to workpiece is set to red
Color, centre position illumination is set to green, and the illumination of high position is set to blueness, still, and the position of three kinds of illuminations and the correspondence of color are simultaneously
It is not limited to this.
In addition, in the above description, to the light that is irradiated on workpiece for the situation of i.e. a kind of color of white light and by it is red,
The situation that green, blue three-color is constituted is illustrated, and still, the species of the color of the light of irradiation is not limited to this, as long as
It is two or more arbitrary colors.
Comparative example 1
Then, the comparative example 1 of present embodiment is illustrated according to Fig. 8 flow chart.
In the flow chart of figure 8, the lighting device 10a shown in Fig. 7 (a), (b) is configured at chip shape electronic unit first
Upside, chip shape electronic unit is irradiated (S101).Then, using filming apparatus 20 to the upper of chip shape electronic unit
Surface is shot.
Then, defect detecting method is performed by flaw detection apparatus 20A.Defect device 20A is connected with monitor 20B.
First, as shown in monochrome image generation process (S103), extracted red (R), green (G), blue (B) from shooting image
Each color component, generation red image, green image, blue image are used as three kinds of monochrome images.The extraction of each color component is led to
Cross and carry out assorted filtering process to shooting image to perform using software.Fig. 9 represents the explanation figure of monochrome image generation process.
Fig. 9 (b), (c), (d) respectively illustrate red image, the green obtained by each color component for the shooting image for extracting Fig. 9 (a)
Image, blue image.
Then, carry out selecting most differentiate from the three kinds of monochrome images generated the check image selection of the image of defect
Process.S104 to S106 of the check image selection process equivalent to Fig. 8.First, in S104, monitor 20B institutes by visual observation
Whether the image of display to having in red image, green image, blue image can most differentiate that the image of defect is compared and grind
Study carefully.Now, preferably：Prepare multiple times for being considered not having the workpiece of defective certified products and be considered existing defects respectively
The workpiece of product, after S101 to S103 is performed, the mutual comparative studies of multiple monochrome images to being generated as its result.
In the case where S104 judged result is "Yes", into S105, corresponding monochrome image is selected to scheme as inspection
Picture.Enter S106 in the case of being "No" in S104 judged result, some computings are carried out between monochrome image, and to them
Result be compared to select check image.
The concrete example on the computing between monochrome image is illustrated below.Now, green image is designated as G, will
Blue image is designated as B, and G-B is designated as using image is removed as its poor normal reflection component.Now, blue image is substantially only by positive and negative
Penetrate to be formed, therefore, the brightness by carrying out green image is subtracted obtained from brightness (G-B) of blue image this image procossing
It is the image for eliminating normal reflection substantially that normal reflection component, which removes image (G-B images),.If from the G-B image zooming-out defects, by
In eliminating normal reflection substantially, therefore compared with extracting the situation of defect in itself from shooting image, the precision of extraction is improved.But
It is, it is impossible to which it is whole defects on the surface of the object to ensure the defect gone out from G-B image zooming-outs.
Therefore, brightness by visual observation to such as R-B images in addition to G-B images or each monochrome image of progress
Between addition obtained from be compared such as G+R images, selection be judged as that the image that can most extract defect well is
Final inspection image.
After it have selected check image by S105 or S106, into the S107 as extracted region process.Here, to choosing
The check image selected carries out the image procossing based on software, extracts the check object area as the object for carrying out defect inspection
Domain.
Then, into the S108 that process is made as brightness histogram, brightness is made in the pixel on check object region
Histogram.Then, in the S109 of process is determined as binary-state threshold, by the histogram of the visual brightness of monitor 20B,
Determine that appropriate brightness is used as binary-state threshold.
Figure 10 represents the S107 to S109 of above explanation figure.Figure 10 (a) is the inspection figure of the workpiece W as certified products
Picture.Here, workpiece W check object region is set to electrode Wa.In this case, extracted from Figure 10 (a) check image
Electrode Wa is used as check object region.Figure 10 (b) shows the situation as schematic diagram.Carried by the electrode Wa shown in solid line
Take, main body Wd and electrode Wb shown by dashed lines is not extracted.
So, the step of generation check image shown in Figure 10 (a) to Figure 10 (b) for Fig. 8 S107.In addition, Figure 10 (c)
Represent in Fig. 8 S108, whole pixels included by Figure 10 (b) Wa are made brightness histogram (histogram, point
Butut) obtained by figure.Distribution from low-light level to high brightness can be described as substantially normal distribution.On the other hand, on scarce
The workpiece W, Figure 10 (d) of sunken substandard products show check image, and Figure 10 (e) represents the signal in the check object region being extracted
Figure, Figure 10 (f) represents the histogram of the brightness of whole pixels on check object region.
As shown in Figure 10 (d) and Figure 10 (e), in the workpiece W of substandard products electrode Wa existing defects WD1.Electrode Wa brightness
It is higher, therefore defect WD1 brightness relatively step-down.Therefore, in Figure 10 (f), in terms of as the Ld of low-light level number of times
There is Δ F1 (=Fd1-Fn1) peak.Figure 10 (f) is compared with Figure 10 (c), in expression and the workpiece W of certified products electricity
In histogrammic Figure 10 (c) corresponding pole Wa, in the absence of the peak of brightness Ld number of times.
So, the picture in check object region the workpiece W of certified products and the workpiece W of substandard products is made in the S108 by Fig. 8
After the histogram of element, the S109 that process is determined as binary-state threshold is subsequently entered.In S109, on monitor 20B visually
Histogram shown in Figure 10 (c) and Figure 10 (f).Then, determine to be used to extract to include producing on number of times because of certified products and substandard products
Brightness Ld scope is used as binary-state threshold as the optimal brightness of defect candidate region as raw obvious difference.Tp1 is fitted
Cooperate as binary-state threshold in this case.
In addition, in order to compare, Figure 10 (g) to Figure 10 (i) shows the example for the image for being not suitable as check image.It is right
The defect WD2 and defect WD1 in Figure 10 (d) in Figure 10 (g) is compared, and defect WD2 profile and scope is than defect WD1
Light, therefore, it is difficult to distinguish Figure 10 (a) certified products and Figure 10 (g) substandard products by visual observation.That is, Fig. 8 S104 judgement is passed through
Benchmark, the workpiece W of substandard products appear as Figure 10 (g) as image be not suitable as check image.Figure 10 (h) shows reality
The check object region extracted on border by Fig. 8 S107 from Figure 10 (g) image, passes through in addition, Figure 10 (i) is shown
Brightness histograms of Fig. 8 S108 to Figure 10 (h) check object Area generation.
In Figure 10 (i) and Figure 10 (f), understood if being compared to brightness Ld number of times, in Figure 10 (i), brightness
Ld number of times is only small as Δ F2 (=Fd2-Fn2) all to claim not to peak relative to the amount that the brightness before and after it changes
On degree value.That is, the histogram from corresponding brightness will also realize that image is unsuitable for check image as Figure 10 (g).
So, after Fig. 8 S109 is gone to, the S110 as defect candidate region abstraction process is subsequently entered.
In S110, binaryzation is carried out to the pixel in check object region using the threshold value Tp1 shown in Figure 10 (c), Figure 10 (f).It is tied
Really, the defect candidate region being made up of the low pixels of brightness specific luminance Tp1 is extracted.
Then, as shown in Fig. 8 S111, the pixel of the defect candidate region to extracting carries out Morphological scale-space.Herein
Morphological scale-space is illustrated.Morphological scale-space is the general name of the image procossing carried out to binary image (black white image).
Combination carries out expansion and contraction described later several times.Its object is to：The smoothing of binary image (reduces concavo-convex and is allowed to flat
It is sliding), remove isolated point (filling up), the removing of jut, the separation of bound fraction, combination of cut-off parts etc..Expansion is to make
Figure in black white image expands the processing of the amount of a pixel vertically and horizontally.Shrink with described expansion process on the contrary, being
The figure in black white image is set to shrink the processing of the amount of a pixel vertically and horizontally.
As the processing for being combined with expansion and shrinking, with opening operation (opening) and closed operation (closing).Open fortune
N times contraction is first carried out at last, carry out the computing of n times expansion, removing and/or bound fraction in the jut of figure again afterwards
Separation in terms of have effect.With opening operation on the contrary, closed operation is first to carry out n times expansion, carry out the fortune of n times contraction again afterwards
Calculate, figure fill up and/or the combination of cut-off parts in terms of have effect.By carrying out Morphological scale-space, so that binaryzation
The periphery of image is smoothened, interconnects the part that original one has but been cut off, and can remove image-independent
Noise.
That is, if, can be as S112 by certain area using binary image obtained from Morphological scale-space is passed through
Regional determination above is defect.Defect inspection process is constituted by S111 and S112.
In addition, the problem of that has following in above-mentioned comparative example 1 point.Expect to make by Fig. 7 (a), (b) or Figure 11 (a),
(b) the flaw detection apparatus 20A automations that lighting device 10a, 10b and filming apparatus 20 shown in are constituted, it is desirable in workpiece W amounts
Shorten the review time during production.It is corresponding to the expectation, as long as realizing Fig. 8 flow chart by software and the software being equipped on into defect
Check device 20A.When designing the software, designer prepares multiple to be considered do not have defective conjunction as described above respectively
The workpiece of lattice product and the workpiece for being considered defective substandard products.Then, after Fig. 8 S101 to S103 is performed to these workpiece,
To the multiple monochrome images generated as its result or as image obtained from the operation result between each monochrome image
Mutually compare, check image is selected from these images.
Here, the species of the defect detected as the shape and wanting of the workpiece of check object in inspection is related to a lot
Aspect.Therefore, which image is most in each monochrome image and as image obtained from the operation result between each monochrome image
Properly as check image, accordingly occurs various change with the shape of these workpiece and the species of defect.Above-mentioned is green
Color image (G)-example of computing as blue image (B) is the poor subtraction for taking two kinds of monochrome images.But it is possible to generate
The computing of optimal check image such as red image (R)-green image (G)+blue image (B) is probably like that to three kinds
Monochrome image combination carries out the computing of computing not of the same race.That is, designer needs to rely on based on the intuition of past experience to determine
Determine the content of computing.So, the substantial amounts of computing based on intuition is performed, the substantial amounts of image generated by various computings is entered
Row relatively selects check image.And as described above, most suitable as the image of check image, with the shape of workpiece and lacking
Various change accordingly occurs for sunken species.Therefore, whenever the shape and the species of defect of the workpiece as check object
, it is necessary to re-execute following operations when changing：Computing, choosing are determined by the intuition based on past experience as described above
Select optimal check image.Therefore, the automation for being checked software design aspect, it is necessary to many labour
And the time.
On the other hand, as described above, can be easily and accurate without complicated image procossing according to present embodiment
Really perform defect inspection.
Comparative example 2
Then, as the comparative example 2 relative to present embodiment, on being used in the defect detecting method of conventional art
The extraction (Fig. 8 S107 to S110) of the defect candidate region of the binary image of check image is illustrated.Figure 12 represents logical
One of the check image crossed Fig. 8 S104 to S106 and selected.
Figure 12 is the shooting from the workpiece W shown in Fig. 7 (b) the i.e. upper surface of chip shape electronic unit (Wu in Fig. 7 (b))
The blue image that image is obtained.If defect candidate regions can be extracted to the check image binaryzation by appropriate threshold value
Domain.In the left and right of image it can be seen that the region of white lengthwise be electrode Wa and electrode Wb respectively.Be formed at electrode Wa with
Between electrode Wb, by edge E1 and the edge E2 (being surrounded by chain-dotted line) of white clip above and below the region grown crosswise be main body
Wd。
Here, in fig. 12, white part is in edge E1 and edge E2 as main body Wd in thin banding of growing crosswise
It is neighbouring on the main body Wd that chain-dotted line is surrounded.In addition, in addition, there are three from left to right thin on main body Wd
Banding region.In fig. 12, above-below direction in these three regions, in main body Wd is shown by chain-dotted line with surrounding
On substantial middle part a thin banding region Bd1.It is that defect is either made an uproar on these regions E1, E2, Bd1
Sound, distinguishes it is extremely difficult by visual observation.The difficult expression of the discrimination of defect and noise in such check image：For
The setting for extracting the threshold value of the binaryzation of defect candidate region is highly difficult.Make binary image optimal by adjusting the threshold value
Change, it is possible to can obtain easily carrying out the binary image of defect inspection.But, reach the tune of the threshold value untill above-mentioned degree
It is whole to be highly desirable to the time, and also there is the defect that easily progress defect inspection can not be also extracted even if by the adjustment sometimes
The situation of candidate region.On the other hand, as described above, can be without complicated image procossing according to present embodiment
Easily and securely perform defect inspection.
In the above description, check image is blue image, but can fully be thought：For will be from same shooting image
Obtained red image or green image is respectively as the setting of the binary-state threshold in the case of check image, defect candidate regions
The extraction in domain is similarly difficult.Moreover, using the image for carrying out computing between monochrome image and generating to scheme as inspection
As in the case of, the setting of threshold value when making check image binaryzation is also difficult as described above.That is, it can obtain following knot
Really：Which kind of check image no matter is selected, implements defective and checks all difficult.
Comparative example 3
Then, as the comparative example 3 compared with present embodiment, following situations are illustrated using Figure 13：Than
Compared with example 1, being made distribution map, and split by straight line defect distribution region and noise profile region in distribution map.Than
Compared with example 1, as described above, from shooting image obtain red image, green image, blue image as monochrome image (Fig. 8's
S103).In the image that computing is carried out from the monochrome image so obtained or between monochrome image and is generated, check image is selected
(Fig. 8 S104 to S106).Then, check object region is extracted from check image, binaryzation (figure is carried out to its whole pixel
8 S107 to S110).Here, Figure 13 is that the defect candidate regions in being extracted by binaryzation are shown in the same manner as Fig. 3
The two-dimensional distribution of the respective Luminance Distribution of defect and noise in domain.The longitudinal axis takes the brightness B of blue component, and transverse axis takes red component
Brightness R.The region Dd of solid line is the defect distribution region of the set of the Luminance Distribution as defect, and the region Dn of dotted line is to make
For the noise profile region of the set of the Luminance Distribution of noise.It is uncommon in defect distribution region Dd and noise profile region Dn
Part, therefore on the distribution map, both can separate in visual scope.
Here, the longitudinal axis of Figure 13 distribution map corresponds to the Luminance Distribution of Fig. 9 (d) blue image, transverse axis corresponds to Fig. 9
(b) Luminance Distribution of red image.That is, selection Fig. 9 (b) red image is as check image, for will be from check image
In binary image obtained from whole pixel binaryzations in check object region for extracting, if some threshold value can be applied
RTT makes the defect distribution region Dd in Figure 14 distribution map be separated with noise profile region Dn, then can by such separation
Determine defect candidate region and accurately implement defect inspection.
In addition, in Figure 13 distribution map, when taking threshold value RTT on transverse axis (red brightness) and drafting and the longitudinal axis are (blue
The brightness of color) parallel straight line R=RTT when, noise profile region Dn all belongs to R ＞ RTT region, defect distribution region Dd
Belong to R ＞ RTT region and this two side of R ＜ RTT region.This is represented：Also can not be just even if the brightness R given thresholds to red
Really separate defect distribution Dd and noise profile region Dn.
Similarly, in Figure 13 distribution map, when taking threshold value BTT and drafting and transverse axis on the longitudinal axis (blue brightness)
During (red brightness) parallel straight line B=BTT, defect distribution region Dd all belongs to B ＞ BTT region, noise profile area
Domain Dn belongs to B ＞ BTT region and this two side of B ＜ BTT region.This is represented：Even if to blueness brightness B given thresholds also without
Defect distribution Dd and noise profile region Dn is correctly separated in method.
That is, select check image in the case of two-dimensional distribution is made, no matter to make from monochrome image in comparative example 1
With Fig. 9 (b) red image as check image, or use Fig. 9 (d) blue image as check image, all it is extremely difficult certainly
Determine defect candidate region, it is impossible to implement defective inspection.
Below, although eliminate detailed description, still, even from the defect distribution region Dd in Figure 13 and noise point
Cloth region Dn mutual position relationship carries out binaryzation to Fig. 9 (c) green image, and there is also will can not lack as described above
Fall into the situation that distributed areas Dd and noise profile region Dn is correctly separated.The situation with S104 in the flow chart of figure 8
It is difficult to select the image this case that can most differentiate defect corresponding from monochrome image.As a result, as Fig. 8 S106 that
Select most differentiate the image of defect in sample, the image generated from the computing carried out between monochrome image.Moreover, in this case
Also occur the phenomenon same with the phenomenon being illustrated according to Figure 13, therefore, be still difficult to generate from as operation result
Image in select most differentiate the image of defect.
On the other hand, according to present embodiment, the present invention by the pixel in shooting image be divided into defect distribution region with
Noise profile region, and the pixel for constituting each region is made the two-dimensional distribution of the brightness based on two kinds of colors.Moreover, at this
By straight line by two region segmentations on distribution map, so as to select the whole in the region for belonging to the object as defect inspection
Pixel, shooting image is limited by the pixel generation selected.Due to being such simple process, therefore, it is possible to will easily make an uproar
Except sound distributed areas (exclusion), implement defective inspection.In addition, i.e. become check object workpiece shape with
And the species of defect changes, process is also identical, therefore, there is no need to re-execute the optimal inspection figure of selection as comparative example 1
The operation of picture.That is, the labour wanted for the design of the software of the automation checked diminishes, design want when
Between also shorten.
Claims (14)
1. a kind of defect detecting method, it is characterised in that including：
Process is illuminated, the surface of checked property is light shone；
Process is shot, the surface of checked property shoot so as to obtain shooting image；
Arrangement step, being extracted from shooting image detect the defect distribution region of defect and need not detect making an uproar for defect
Sound distributed areas, and the color component of the pixel in each region is configured at two kinds of color components work in red, green and blue
On two-dimensional distribution for the longitudinal axis and transverse axis；
Segmentation process, will constitute the pixel in defect distribution region by segmentation straight line on two-dimensional distribution and constitutes noise profile area
The pixel in domain is separated；
Process is selected, selects in two regions that divided line segmentation is opened on two-dimensional distribution and belongs to defect distribution area
The pixel of domain side；
Limit process, by process is selected the pixel selected be applied to the pixel of this in shooting image it is the location of original and
Generation limits shooting image；And
Check and perform process, defect inspection is performed using shooting image is limited.
2. defect detecting method according to claim 1, it is characterised in that
Just white light.
3. defect detecting method according to claim 1, it is characterised in that
The just light of different two or more colors.
4. defect detecting method according to claim 1, it is characterised in that
Just the 1st coloured light, the 2nd coloured light and the 3rd coloured light of different colours.
5. defect detecting method according to claim 4, it is characterised in that
1st coloured light, the 2nd coloured light and the 3rd coloured light are any one in red, green and blue respectively.
6. the defect detecting method according to claim 4 or 5, it is characterised in that
In illumination process, the height that the position of the 1st coloured light, the 2nd coloured light and the 3rd coloured light is irradiated to checked property is different.
7. defect detecting method according to claim 1, it is characterised in that
X is set to the longitudinal axis of two-dimensional distribution is set into y, transverse axis and when a and b are set to real number, and segmentation straight line is by expression of first degree y=ax
+ b is represented.
8. a kind of defect inspecting system, it is characterised in that possess：
Lighting device, light shines the surface of checked property；
Filming apparatus, to the surface of checked property shoot so as to obtain shooting image；And
Flaw detection apparatus, defect inspection is carried out to the shooting image real-time image processing from filming apparatus,
Flaw detection apparatus possesses：
Configuration section, being extracted from shooting image detect the defect distribution region of defect and need not detect the noise of defect
Distributed areas, and by the color component of the pixel in each region be configured at using two kinds of color components in red, green and blue as
On the two-dimensional distribution of the longitudinal axis and transverse axis；
Cutting part, pixel and the composition noise profile area in defect distribution region will be constituted on two-dimensional distribution by splitting straight line
The pixel in domain is separated；
Choice portion, selects in two regions that divided line segmentation is opened on two-dimensional distribution and belongs to defect distribution region
The pixel of side；
Limited section, is applied to the original location of the pixel of this in shooting image by the pixel selected in process is selected and gives birth to
Into restriction shooting image；And
Enforcement division is checked, defect inspection is performed using shooting image is limited.
9. defect inspecting system according to claim 8, it is characterised in that
Just white light.
10. defect inspecting system according to claim 8, it is characterised in that
The just light of different two or more colors.
11. defect inspecting system according to claim 8, it is characterised in that
Just the 1st coloured light, the 2nd coloured light and the 3rd coloured light of different colours.
12. defect inspecting system according to claim 11, it is characterised in that
1st coloured light, the 2nd coloured light and the 3rd coloured light are any one in red, green and blue respectively.
13. the defect inspecting system according to claim 11 or 12, it is characterised in that
In illumination process, the height that the position of the 1st coloured light, the 2nd coloured light and the 3rd coloured light is irradiated to checked property is different.
14. defect inspecting system according to claim 8, it is characterised in that
X is set to the longitudinal axis of two-dimensional distribution is set into y, transverse axis and when a and b are set to real number, and segmentation straight line is by expression of first degree y=ax
+ b is represented.
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| JP2016-086315 | 2016-04-22 | ||
| JP2016086315A JP6625475B2 (en) | 2016-04-22 | 2016-04-22 | Defect inspection method and defect inspection system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107305190A true CN107305190A (en) | 2017-10-31 |
| CN107305190B CN107305190B (en) | 2020-08-04 |
Family
ID=60150758
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710240030.3A Active CN107305190B (en) | 2016-04-22 | 2017-04-13 | Defect inspection method and defect inspection system |
Country Status (4)
| Country | Link |
|---|---|
| JP (1) | JP6625475B2 (en) |
| KR (1) | KR101898947B1 (en) |
| CN (1) | CN107305190B (en) |
| TW (1) | TWI618925B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108508053A (en) * | 2018-04-26 | 2018-09-07 | 武汉新芯集成电路制造有限公司 | A kind of detection method of the atomic physical imperfection of systematicness |
| CN111175302A (en) * | 2018-11-13 | 2020-05-19 | 晶彩科技股份有限公司 | Optical image automatic acquisition method with composite detection conditions |
Families Citing this family (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP7214432B2 (en) * | 2018-10-22 | 2023-01-30 | キヤノン株式会社 | Image processing method, image processing program, recording medium, image processing apparatus, production system, article manufacturing method |
| CN109584239B (en) * | 2018-12-13 | 2024-02-06 | 华南理工大学 | High-light object surface defect detection system and method based on reflected light |
| US10853932B2 (en) * | 2019-01-16 | 2020-12-01 | Applied Material Israel, Ltd. | Method of defect detection on a specimen and system thereof |
| CN112150412B (en) * | 2020-08-31 | 2022-11-22 | 国网浙江省电力有限公司常山县供电公司 | Insulator self-explosion defect detection method based on projection curve analysis |
| JP2022146141A (en) * | 2021-03-22 | 2022-10-05 | 株式会社安永 | Visual inspection method and visual inspection device |
| CN116625239B (en) * | 2023-07-19 | 2023-09-15 | 深圳灿维科技有限公司 | Step hole detection method, device and system based on image processing |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101409825A (en) * | 2007-10-10 | 2009-04-15 | 中国科学院自动化研究所 | Nighttime vision monitoring method based on information fusion |
| CN103559507A (en) * | 2013-11-18 | 2014-02-05 | 沈阳工业大学 | Method for detecting traffic signs based on combination of color feature and shape feature |
| CN104198497A (en) * | 2014-09-12 | 2014-12-10 | 苏州大学 | Surface defect detection method based on visual saliency map and support vector machine |
| CN105372267A (en) * | 2014-08-11 | 2016-03-02 | 东京威尔斯股份有限公司 | Device and method for inspecting appearance of transparent substrate |
Family Cites Families (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPH02213711A (en) * | 1989-02-14 | 1990-08-24 | Omron Tateisi Electron Co | Parts inspecting device |
| JPH0993443A (en) * | 1995-05-16 | 1997-04-04 | Sanyo Electric Co Ltd | Color-monochromatic image conversion method and edge position detection method for object to be inspected |
| JP2002243655A (en) * | 2001-02-13 | 2002-08-28 | Mitsubishi Materials Corp | Method and equipment for visual inspection of electronic component |
| US7113313B2 (en) * | 2001-06-04 | 2006-09-26 | Agilent Technologies, Inc. | Dome-shaped apparatus for inspecting a component or a printed circuit board device |
| JP2006004715A (en) * | 2004-06-16 | 2006-01-05 | Matsushita Electric Works Ltd | Wiring device |
| JP4539964B2 (en) * | 2004-07-21 | 2010-09-08 | 大日本スクリーン製造株式会社 | Image segmentation |
| JP4750444B2 (en) * | 2005-03-24 | 2011-08-17 | 株式会社日立ハイテクノロジーズ | Appearance inspection method and apparatus |
| JP2008064715A (en) * | 2006-09-11 | 2008-03-21 | Omron Corp | Defect inspection device and defect inspection method |
| JP4399494B2 (en) * | 2006-12-28 | 2010-01-13 | シャープ株式会社 | Defect detection apparatus, defect detection method, image sensor device, and image sensor module |
| JP5056263B2 (en) * | 2007-08-14 | 2012-10-24 | オムロン株式会社 | Image inspection apparatus, image inspection method, program for causing computer to function as image inspection apparatus, and computer-readable recording medium |
| JP5007979B2 (en) * | 2008-05-22 | 2012-08-22 | 独立行政法人産業技術総合研究所 | Defect inspection method and defect inspection apparatus |
| JP2014062837A (en) * | 2012-09-21 | 2014-04-10 | Hitachi High-Technologies Corp | Defect inspection device and defect reviewing device |
| US9619876B2 (en) * | 2013-03-12 | 2017-04-11 | Kla-Tencor Corp. | Detecting defects on wafers based on 2D scatter plots of values determined for output generated using different optics modes |
| US10127652B2 (en) * | 2014-02-06 | 2018-11-13 | Kla-Tencor Corp. | Defect detection and classification based on attributes determined from a standard reference image |
-
2016
- 2016-04-22 JP JP2016086315A patent/JP6625475B2/en active Active
-
2017
- 2017-03-07 TW TW106107407A patent/TWI618925B/en active
- 2017-03-15 KR KR1020170032315A patent/KR101898947B1/en active IP Right Grant
- 2017-04-13 CN CN201710240030.3A patent/CN107305190B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101409825A (en) * | 2007-10-10 | 2009-04-15 | 中国科学院自动化研究所 | Nighttime vision monitoring method based on information fusion |
| CN103559507A (en) * | 2013-11-18 | 2014-02-05 | 沈阳工业大学 | Method for detecting traffic signs based on combination of color feature and shape feature |
| CN105372267A (en) * | 2014-08-11 | 2016-03-02 | 东京威尔斯股份有限公司 | Device and method for inspecting appearance of transparent substrate |
| CN104198497A (en) * | 2014-09-12 | 2014-12-10 | 苏州大学 | Surface defect detection method based on visual saliency map and support vector machine |
Non-Patent Citations (1)
| Title |
|---|
| 余锋: ""基于视觉的工业机器人协作装配方法研究"", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108508053A (en) * | 2018-04-26 | 2018-09-07 | 武汉新芯集成电路制造有限公司 | A kind of detection method of the atomic physical imperfection of systematicness |
| CN108508053B (en) * | 2018-04-26 | 2021-01-12 | 武汉新芯集成电路制造有限公司 | Method for detecting systematic infinitesimal physical defects |
| CN111175302A (en) * | 2018-11-13 | 2020-05-19 | 晶彩科技股份有限公司 | Optical image automatic acquisition method with composite detection conditions |
Also Published As
| Publication number | Publication date |
|---|---|
| TWI618925B (en) | 2018-03-21 |
| JP2017194425A (en) | 2017-10-26 |
| KR20170121048A (en) | 2017-11-01 |
| TW201741649A (en) | 2017-12-01 |
| CN107305190B (en) | 2020-08-04 |
| KR101898947B1 (en) | 2018-09-17 |
| JP6625475B2 (en) | 2019-12-25 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107305190A (en) | Defect detecting method and defect inspecting system | |
| CN114723701B (en) | Gear defect detection method and system based on computer vision | |
| DE69817580T2 (en) | DEVICE FOR THE OPTICAL INSPECTION OF PACKAGING FILMS | |
| JP5706860B2 (en) | Method for measuring the appearance quality of white rice and brown rice | |
| CN108491892A (en) | fruit sorting system based on machine vision | |
| CN105466950A (en) | Bamboo dice online detection method and system based on vision | |
| CN108765402A (en) | Non-woven fabrics defects detection and sorting technique | |
| Kumar et al. | Automatic detection of acute myeloid leukemia from microscopic blood smear image | |
| CN105223208B (en) | A kind of circuit board detecting template and preparation method thereof, circuit board detecting method | |
| DE102012001868B4 (en) | Method for setting up a system for the optical identification of objects, laboratory image recording system for carrying out such a method and arrangement comprising the laboratory image recording system and the installation | |
| CN106971141A (en) | Cell compartment determines method, cell camera system and cell image processing apparatus | |
| KR20100121250A (en) | Vision system for inspection of qfn semiconductor package and classifying method | |
| CN100492413C (en) | Method for calibrating camera and colour reference thereby | |
| CN106960424A (en) | Tubercle bacillus image segmentation and identification method and device based on optimized watershed algorithm | |
| KR20100122558A (en) | Method of measuring an area of a measurement target on a printed circuit board | |
| Cusano et al. | Intensity and color descriptors for texture classification | |
| CN108537815A (en) | A kind of video image foreground segmentation method and device | |
| KR20150114614A (en) | System and method for classificating surface mount device | |
| Jin et al. | High-quality tea flushes detection under natural conditions using computer vision | |
| JP4195980B2 (en) | Appearance inspection method and appearance inspection apparatus using color image | |
| JP2014186710A (en) | Shiitake mushroom grading apparatus using image processing technology | |
| CN104484641B (en) | A kind of system for recognizing color three dimension code | |
| CN115908242B (en) | Chip gold thread whole line detection method and system for multichannel image fusion | |
| KR102120351B1 (en) | Light Setting Method of Vision system | |
| CN109948543A (en) | A kind of leading line extraction method of field close shot image |
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 |