CN106651857A - Printed circuit board patch defect detection method - Google Patents
Printed circuit board patch defect detection method Download PDFInfo
- Publication number
- CN106651857A CN106651857A CN201710005017.XA CN201710005017A CN106651857A CN 106651857 A CN106651857 A CN 106651857A CN 201710005017 A CN201710005017 A CN 201710005017A CN 106651857 A CN106651857 A CN 106651857A
- Authority
- CN
- China
- Prior art keywords
- formula
- image
- numeral
- region
- area
- 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
- 230000007547 defect Effects 0.000 title claims abstract description 31
- 238000001514 detection method Methods 0.000 title claims abstract description 23
- 238000000034 method Methods 0.000 claims abstract description 103
- 238000004458 analytical method Methods 0.000 claims abstract description 18
- 238000004364 calculation method Methods 0.000 claims abstract description 10
- 238000005516 engineering process Methods 0.000 claims abstract description 7
- 101100129500 Caenorhabditis elegans max-2 gene Proteins 0.000 claims description 16
- 101100083446 Danio rerio plekhh1 gene Proteins 0.000 claims description 15
- 238000012512 characterization method Methods 0.000 claims description 14
- 238000002372 labelling Methods 0.000 claims description 14
- 229910000679 solder Inorganic materials 0.000 claims description 14
- 238000007689 inspection Methods 0.000 claims description 12
- 238000006243 chemical reaction Methods 0.000 claims description 6
- 239000003550 marker Substances 0.000 claims description 5
- 238000012360 testing method Methods 0.000 claims description 5
- 238000009432 framing Methods 0.000 claims description 4
- 230000002950 deficient Effects 0.000 claims description 3
- 239000003086 colorant Substances 0.000 abstract description 3
- 238000003466 welding Methods 0.000 abstract description 2
- 230000011218 segmentation Effects 0.000 description 6
- ATJFFYVFTNAWJD-UHFFFAOYSA-N Tin Chemical compound [Sn] ATJFFYVFTNAWJD-UHFFFAOYSA-N 0.000 description 5
- 238000010586 diagram Methods 0.000 description 4
- 230000004069 differentiation Effects 0.000 description 4
- 230000002902 bimodal effect Effects 0.000 description 2
- 230000001419 dependent effect Effects 0.000 description 2
- 238000006073 displacement reaction Methods 0.000 description 2
- 239000004744 fabric Substances 0.000 description 2
- 238000003709 image segmentation Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 238000000280 densification Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 230000000877 morphologic effect Effects 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0004—Industrial image inspection
-
- G06T5/70—
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30108—Industrial image inspection
- G06T2207/30141—Printed circuit board [PCB]
Abstract
The invention relates to the fields of machine vision and printed circuit board detection, in particular to a machine vision technology-based printed circuit board patch defect detection method. According to the method, identification is performed based on the positions, colors and quantity of pixel points meeting a threshold condition, and an identification algorithm is realized through a relatively small calculation amount. The method comprises the following steps of performing patch locating and size calculation; performing patch region color analysis; and performing patch type identification. The method is simple and convenient; component whole bodies, welding points and patch numbers are identified according to a process; detection items are comprehensive; different detection conditions can be met; and detection programs can be designed for different patches.
Description
Technical field
The present invention relates to field of machine vision and printed circuit board (PCB) detecting field, more particularly to it is a kind of based on machine vision skill
The printed circuit board (PCB) paster defect inspection method of art.
Background technology
Circuit board is the important component part of Modern Industry Products, and in the processing technology of printed circuit board (PCB), Product checking is
One of important operation.Paster is that Modern circuit boards realize one of elementary cell of function.The development trend of current paster is essence
Densification and miniaturization, need the detection method of accurate, efficient printed circuit board (PCB) (PCB).
The picture that general camera shoots is two dimensional image, but the surface features of paster two ends scolding tin are three-dimensional detection mesh
Mark.The widely used RGB three-color light sources of modern PCB detection industries irradiate paster, and to the scolding tin of different gradients different colors are reflected
Coloured silk, so as to react scolding tin surface nature.
Printed circuit board (PCB) paster has a fine distinction in the features such as patch location, angle, solder joint shape, but phase
For the size of paster, this nuance impact can not ignore.And traditional image-recognizing method, such as image comparison
Method, it is excessively sensitive to nuance, easily cause erroneous judgement.But other morphologic image recognition algorithms, such as wavelet transformation, Kazakhstan
Husband converts, then exist computationally intensive, and the shortcomings of affected big by ambient interferences, testing result can not fully meet requirement.
Based on the problems referred to above, the present invention proposes a kind of method of detection paster defect, is welded using the irradiation of RGB three-color light sources
The image that stannum is obtained carries out image recognition, can detect paster common deficiency, such as Short Item, wrong part, Component Displacement, set up a monument, solder joint
Defect etc..User uses template paster, and according to algorithm routine threshold value is calculated, and is capable of achieving the detection of paster defect.
The content of the invention
The present invention is a kind of printed circuit board (PCB) paster defect inspection method, is mainly solved in standard industry detection platform environment
Under, the identification problem of printed circuit board (PCB) paster defect, can detect Short Item, wrong part, Component Displacement, set up a monument, welding point defect etc. lacks
Fall into.
The purpose of the present invention is achieved through the following technical solutions:
A kind of printed circuit board (PCB) paster defect inspection method, comprises the steps:
1) paster positioning and Size calculation:
1a) according to PCB patch location information, by framing technology, the overall patch location of positioning simultaneously intercepts paster figure
Picture;Separate picture red channel, the red channel image to isolating carries out binarization operation, index for selection threshold value, and to figure
As carrying out a Denoising disposal.
1b) according to above-mentioned patch location and the red component of image, using the metrics-thresholds chosen, using Threshold segmentation skill
Art, is partitioned into the red area of the end pad of paster two, as flat site.
1c) all connected regions in the region that is partitioned into are marked, left side white portion is designated as 1st area, and right side is white
Zone domain is designated as 2nd area;In the horizontal and vertical coordinate in 1st area minima and maximum be designated as respectively minx1, miny1, maxx1,
maxy1;Minima and maximum are designated as respectively minx2, miny2, maxx2, maxy2 in the horizontal and vertical coordinate in 2 area;According to
Formula 1 distinguishes parameter t1、t2、t3、t4、t5、t6, and compared with corresponding index upper limit threshold and index lower threshold
Compared with.
If index is in the range of index upper limit threshold and index lower threshold, judge that patch size is correct, otherwise
It is recorded as not meeting the project of metrics-thresholds bound.
1d) according to the image upper left corner and the angular coordinate in the lower right corner, image interception is carried out, intercept method is:It is left with image
Used as coordinate origin, abscissa is y directions at upper angle, and the paster y directions length after cutting is y1;Vertical coordinate be x directions, cutting
Paster x directions length afterwards is x1;Wherein, intercepting angular coordinate choose (miny1-2, minx1-2) and (maxy2+2,
Maxx2+2), it is (maxy2-miny1+4) × (maxx2-minx1+4) to intercept region area.
2) paster region color analysis:
2a) from step 1d) intercept after paster image on intercept 6 block feature regions, region A, region B, area are designated as respectively
Domain C, region D, region E, region F, record the area starting point coordinate and area for intercepting.
2b) to step 2a) in 6 block feature regions carry out color analysis respectively, using pixel red component in region
Meansigma methodss (mcol) are used as index, and formula is as follows:
Wherein, col is each pixel red color component value in region;N is pixel number in region.
Every piece of Regional Red component meansigma methodss are calculated according to formula 2.
When pixel red component meansigma methodss (mcol) meet red component upper limit threshold and red point in every piece of region
During amount lower threshold, then solder joint zero defect is judged;If there is not meeting red component upper limit threshold and red component lower limit threshold
The situation of value, then judge that solder joint is defective, and records defect area.
3) paster type identifier:
3A. numeral identifications:After solder joint detection in paster region color analysis are completed, digital identification is carried out.
3a) in step 1d) in intercept after paster image on the basis of, according to the size of template image in testing image
Intercept numeric area image;Binary conversion treatment, agriculture products threshold value are carried out to digital block area image, and image is once gone
Noise processed.
3b) using the method for zone marker, three numeric areas in image are respectively labeled as into 1st ' area, 2nd ' area, 3rd ' area.Meter
Calculate maximum and minimum value in y, the x coordinate value of each connected domain, be designated as (max1 ' y, max1 ' x), (max2 ' y, max2 ' x),
(max3 ' y, max3 ' are x), (min1 ' y, min1 ' are x), (min2 ' y, min2 ' are x), (min3 ' y, min3 ' is x);According to corresponding y, x
The maximum and minima of coordinate intercepts corresponding numeric area as the upper left corner and bottom right angular coordinate;The numeric area of intercepting is big
In calculated region;For connected component labeling for 1 ' region, angular coordinate be (max1 ' y+2, max1 ' x+2) and
(min1 ' y-2, min1 ' x-2);For connected component labeling for 2 ' region, angular coordinate be (max2 ' y+2, max2 ' x+2) and
(min2 ' y-2, min2 ' x-2);For connected component labeling for 3 ' region, angular coordinate be (max3 ' y+2, max3 ' x+2) and
(min3 ' y-2, min3 ' x-2).
Upper left, lower-left, upper right, the white portion of bottom right of image are taken, A ', B ', C ', D ' is respectively labeled as, each region
White pixel point quantity is denoted as respectively NA ', NB ', NC ', ND '.
Then, individual digit image is identified;Connected domain quantity is as index in image with numeric area as background
The numeral that image may be characterized is classified, sorting technique is as follows:
If 3b1) connected domain quantity is 3, the numeral of characterization image is 8;
If 3b2) connected domain quantity is 2, the numeral of characterization image is 9,6, one of 0, now carries out further numeral and knows
Not, NA ', NB ', NC ', ND ' are compared with the template threshold value of formula 3-5, it is concrete to judge so as to judge the numeral of characterization image
Flow process is:When NA ', NB ', NC ', ND ' coincidence formulas 3, then judge numeral as 9;Work as NA ', NB ', NC ', ND ' and do not meet formula
3 and during coincidence formula 4, then judge numeral as 6;When NA ', NB ', NC ', ND ' do not meet formula 3 and 4 and coincidence formula 5, then
Judge numeral as 0.
Wherein, formula 3,4,5 is as follows:
(NA’+NC’+ND’)/3-NB’>φ1Formula 3
(NA’+NB’+ND’)/3-NC’>φ2Formula 4
|NA’-(NA’+NB’+NC’+ND’)|+|NB’-(NA’+NB’+NC’+ND’)|+|NC’-(NA’+NB’+NC’+
ND’)|+|ND’-(NA’+NB’+NC’+ND’)|>φ3
Formula 5
Wherein, φ1、φ2、φ3The respectively template threshold value of formula 3,4,5.
According to above-mentioned flow process and formula, pass sequentially through template matching and judge whether numeral is 9,6,0;If above-mentioned formula is equal
It is unsatisfactory for, then process decision chart is as None- identified.
If 3b3) connected domain quantity is 1, the numeral of characterization image is 1,2,3,4,5, one of 7, is now carried out further
Numeral identification, picture traverse and NA ', NB ', NC ', ND ' is compared with the template threshold value of formula 6-12, so as to judge image
The numeral of sign, specifically judges that flow process is:
When picture traverse coincidence formula 6, then judge numeral as 1;When picture traverse does not meet formula 6, according to NA ',
NB ', NC ', ND ' proceed to judge;When NC ' coincidence formulas 7, continuation judges NA ', NC ', ND ' whether coincidence formula 8;When
When NA ', NC ', ND ' coincidence formulas 8, then judge numeral as 4;When NA ', NC ', ND ' do not meet formula 8 and coincidence formula 9, then
Judge numeral as 7;When NC ' coincidence formulas 7 and NA ', NC ', ND ' do not meet formula 8 and formula 9, then image fails identification.
When NC ' does not meet formula 7, continuation judges NA ', NC ', ND ' whether coincidence formula 10;During coincidence formula 10, then
Judge numeral as 5;When NA ', NC ', ND ' do not meet formula 10 and NA ', NB ', NC ', ND ' coincidence formulas 11, then numeral is judged
For 2;When NA ', NB ', NC ', ND ' do not meet formula 10 and formula 11 and coincidence formula 12, then judge numeral as 3;Work as NA ',
When NB ', NC ', ND ' do not meet formula 7, formula 10, formula 11 and formula 12, then image fails identification.
Wherein formula 6-12 is as follows:
width<η1Formula 6
NC’<η2Formula 7
|NA’-(NA’+NC’+ND’)/3|<η3Formula 8
|NC’-(NA’+ND’)/2|>η4Formula 9
|NA’-(NA’+NC’+ND’)/3|<η5Formula 10
NB’+NC’-NA’-ND’>η6Formula 11
NB’+ND’-NA’-NC’>η7Formula 12
Wherein, width is picture traverse;η1、η2、η3、η4、η5、η6、η7Respectively formula 6 arrives the template threshold value of formula 12.
According to above-mentioned flow process and formula, the steps such as picture traverse calculating, template matching are passed sequentially through, judge that whether numeral is
1、2、3、4、5、7;If above-mentioned formula is unsatisfactory for, process decision chart is as None- identified.
3B. compares with template numeral, judges whether numeral is correct, so as to whether decision element model is correct.
Numeral identification number of times in a kind of paster model identification step of described printed circuit board (PCB) paster defect inspection method
Recognize three times for each repetition of figures.
A kind of binarization operation of described printed circuit board (PCB) paster defect inspection method adopts Two-peak method, metrics-thresholds with
Template Threshold adopts normal distribution method.
The beneficial effects of the present invention is:
This method adaptability is stronger, can detect paster defect kind more comprehensive.
This method is made using being identified based on the position of the pixel number for meeting threshold condition, color and quantity
Recognizer is realized with less amount of calculation.
This method is easy, and according to flow process, respectively, solder joint overall to element and paster numeral are identified, and detection project is complete
Face, is adapted to different detection operating modes, can be directed to different patch design detection programs.
Description of the drawings
Fig. 1 is the hardware device composition schematic diagram of the present invention;
Fig. 2 is a kind of flow chart of printed circuit board (PCB) paster defect inspection method of the invention;
Fig. 3 is pending paster image of the invention;
Fig. 4 is the red area image after pending image segmentation of the invention;
Fig. 5 is pending image cropping bak stay image of the invention;
Fig. 6 is pending paster feature regional images of the invention;
Fig. 7 is the numeric area image after pending image cropping of the invention;
Fig. 8 is numeral identification process figure of the invention;
Fig. 9 is the area schematic after pending image digitization area image of the invention is intercepted;
Figure 10 is printing digital angle point typical case legend of the present invention;
Figure 11 is overhaul flow chart when connected domain is 2 in numeral identification process of the invention;
Figure 12 is overhaul flow chart when connected domain is 1 in numeral identification process of the invention;
Figure 13 is binarization operation Two-peak method schematic diagram of the present invention;
Figure 14 is the pending paster feature regional images of the embodiment of the present invention;
Figure 15 is the red area image after the pending image segmentation of the embodiment of the present invention;
Figure 16 is the pending paster feature regional images of the embodiment of the present invention;
Figure 17 is the area schematic after the pending image digitization area image of the embodiment of the present invention is intercepted.
Reference
The industrial camera of 11 computer 12
The light source of 13 telecentric lens 14
The area of 1 connected region, 1st area, 2 connected region 2
A characteristic area A B characteristic area B
C characteristic area C D characteristic area D
E characteristic area E F characteristic area F
The numeric area 2 ' of 1 ' numeric area 1 ' 2 '
The white portion A ' that the A ' of 3 ' numeric area 3 ' is intercepted
The white portion C ' that the white portion B ' C ' that B ' is intercepted are intercepted
The workbench of white portion D ' 15 that D ' is intercepted
Specific embodiment
With reference to the accompanying drawings and examples, the specific embodiment of the present invention is described in further detail.
The emphasis of the present invention is defect identification method, based on Blob analytic process and color analysis method, by related objective
The feature such as size, pixel quantity, color component value analysis, calculated based on several template images using normal distribution model
Dependent thresholds, then compare target to be detected and dependent thresholds, by the method for characteristic matching, recognize paster defect.
The hardware device composition of the present invention by kilomega network using computer 11 as shown in figure 1, control industrial camera 12 pairs
Printed circuit board (PCB) carries out image acquisition.Industrial camera 12 connects telecentric lens 13.Light source 14 is carried out using arbitrary source control machine
Control, is that AOI detects special RGB annular light sources, be irradiated to scolding tin surface can cause the plane reflection of different inclination angle it is red,
Green, blue three-color light.
The hardware device course of work of the present invention:
(1) device debugging:Adjust the relevant parameter of industrial camera 12 (white balance, time of exposure) and the combination of light source lighting color
(central region field color is adjusted to by white by the combination of tri- kinds of colors of RGB in the present invention);
(2) PCB pasters to be measured are placed on workbench 15, it is corresponding with the light source 14 above it.Using computer controls work
Industry camera 12 gathers PCB pasters image and is stored in computer 11;
(3) image procossing is carried out, paster defect is recognized.
As shown in Fig. 2 a kind of printed circuit board (PCB) paster defect inspection method of the invention, comprises the steps:
1) paster positioning and Size calculation:
Paster to be measured is as shown in Figure 3.
1a) according to PCB patch location information, by framing technology, the overall patch location of positioning simultaneously intercepts paster figure
Picture.Separate picture red channel, the red channel image to isolating carries out binarization operation, index for selection threshold value, and to figure
As carrying out a Denoising disposal.
Wherein, binarization operation adopts Two-peak method, metrics-thresholds to determine that method adopts normal distribution method.
Binarization operation adopts Two-peak method:
Figure 13 is the coordinate diagram of image Color Channel or gray scale, wherein, x-axis characterizes color component value, y-axis phenogram picture
The quantity of the pixel of middle correspondence x-axis color component, segmentation threshold select background and prospect bimodal between minimum point, i.e. pixel
The point of minimum number.
Preferably, the binary-state threshold of above-mentioned binarization operation is 0.233 × 255.
Metrics-thresholds determine that method adopts normal distribution method:
(1) multiple template image is selected, using algorithm the sample value (a at threshold value differentiation is calculated1,a2,a3……an) (can
Being the quantity of size, or qualified point);
(2) normal distribution model is used, the mean μ and variance ε of normal distribution is calculated;
(3) altimetric image to be checked is analyzed, calculates numerical value a at metrics-thresholds differentiation, and compared with normal distribution model, such as a
Fall in the range of the ε of u ± 3, then judge to meet metrics-thresholds condition, otherwise, it is determined that herein parameter does not meet metrics-thresholds.
1b) according to above-mentioned patch location and the red component of image, using the metrics-thresholds chosen, using Threshold segmentation skill
Art, is partitioned into the red area of the end pad of paster two, as flat site.Red area after segmentation is as shown in Figure 4.
1c) all connected regions in the region that is partitioned into are marked, left side white portion is designated as 1st area, and right side is white
Zone domain is designated as 2nd area.In the horizontal and vertical coordinate in 1st area minima and maximum be designated as respectively minx1, miny1, maxx1,
maxy1;Minima and maximum are designated as respectively minx2, miny2, maxx2, maxy2 in the horizontal and vertical coordinate in 2 area.Then
Parameter t is distinguished according to formula 11、t2、t3、t4、t5、t6, and carry out with corresponding index upper limit threshold and index lower threshold
Relatively, if index is in the range of index upper limit threshold and index lower threshold, judge that patch size is correct, otherwise record
Not meet the project of metrics-thresholds bound.
1d) according to the image upper left corner and the angular coordinate in the lower right corner, image interception is carried out, intercept method is:It is left with image
Used as coordinate origin, abscissa is y directions at upper angle, and the paster y directions length after cutting is y1, vertical coordinate be x directions, cutting
Paster x directions length afterwards is x1.Intercept bak stay image as shown in Figure 5.Considered based on algorithm robustness, wherein, intercepting
Angular coordinate chooses (miny1-2, minx1-2) and (maxy2+2, maxx2+2), and intercepting region area is (maxy2-miny1+
4)×(maxx2-minx1+4)。
2) paster region color analysis
2a) as shown in fig. 6, from step 1d) intercept after paster image on intercept 6 block feature regions, region is designated as respectively
A, region B, region C, region D, region E, region F, record the area starting point coordinate and area for intercepting.The picture starting of intercepting
Point and area are as shown in table 1.
Starting point and area intercept method that the image representative region of table 1 is intercepted
2b) to step 2a) in 6 block feature regions carry out color analysis respectively, using pixel red component in region
Meansigma methodss (mcol) as index, formula is as follows:
Wherein, col is each pixel red color component value in region;N is pixel number in region.
Every piece of Regional Red component meansigma methodss are calculated according to formula 2.Red component Threshold is using normal state point
Cloth method is identical with the determination step of metrics-thresholds.
When pixel red component meansigma methodss (mcol) meet red component upper limit threshold and red point in every piece of region
During amount lower threshold, then solder joint zero defect is judged;If there is not meeting red component upper limit threshold and red component lower limit threshold
The situation of value, then judge that solder joint is defective, and records defect area.
3) paster type identifier:
3A. numeral identifications:After solder joint detection in paster region color analysis are completed, digital identification is carried out, numeral is known
Other flow chart is as shown in Figure 8.
3a) in step 1d) in intercept after paster image on the basis of, according to the size of template image in testing image
Numeric area image is intercepted, the numeric area image after intercepting is as shown in Figure 7.Binary conversion treatment is carried out to digital block area image,
Agriculture products threshold value, and a Denoising disposal is carried out to image.
Wherein, the determination method and step 1a of binary conversion treatment and metrics-thresholds) in it is identical.
3b) using the method for zone marker, by numeric area difference labelling in image.As shown in fig. 7, three numeric areas
It is labeled as 1st ' area, 2nd ' area, 3rd ' area.The maximum and minimum value in y, the x coordinate value of each connected domain is calculated, be designated as (max1 ' y,
Max1 ' x), (max2 ' y, max2 ' x), (max3 ' y, max3 ' x), (min1 ' y, min1 ' x), (min2 ' y, min2 ' x),
(min3 ' y, min3 ' is x).It is right to be intercepted as the upper left corner and bottom right angular coordinate according to corresponding y, the maximum of x coordinate and minima
The numeric area answered.In view of algorithm robustness, the numeric area of intercepting is more than calculated region.As shown in figure 9, for
Connected component labeling is 1 ' region, and angular coordinate is (max1 ' y+2, max1 ' x+2) and (min1 ' y-2, min1 ' x-2);For
Connected component labeling is 2 ' region, and angular coordinate is (max2 ' y+2, max2 ' x+2) and (min2 ' y-2, min2 ' x-2);For
Connected component labeling is 3 ' region, and angular coordinate is (max3 ' y+2, max3 ' x+2) and (min3 ' y-2, min3 ' x-2).
Upper left, lower-left, upper right, the white portion of bottom right of image in Fig. 9 are intercepted, A ', B ', C ', D ' is respectively labeled as, is cut
Take method as follows:As shown in figure 9, using the image upper left corner as coordinate origin, abscissa is y directions, the paster y side after cutting
It is y to length2, vertical coordinate is x directions, and the paster x directions length after cutting is x2, the starting point coordinate of cut out areas and face
Product is as shown in table 2:
The starting point coordinate and area of the cut out areas of table 2
Each region white pixel point quantity is denoted as respectively NA ', NB ', NC ', ND '.
From the shape facility of printing digital, what four angles of digital picture were likely to occur is shaped as shown in Figure 10
Three types:A types, b types and c types, sort by the quantity of white pixel point, and at most, b types take second place c types, and a types are minimum.Inhomogeneity
White pixel point quantity difference is larger between the angle point of type, and the white pixel quantitative difference between the angle point of same type is less.
Then, individual digit image is identified.Connected domain quantity n is as index in image with numeric area as background
(for example, in Fig. 9, numeral is 0, and, then 2), to classify to the numeral that image may be characterized, sorting technique is such as connected domain quantity
Under:
If 3b1) connected domain quantity n is 3, the numeral of characterization image is 8;
If 3b2) connected domain quantity n is 2, the numeral of characterization image is 9,6, one of 0, now carries out further numeral and knows
Not, flow process as shown in figure 11, NA ', NB ', NC ', ND ' is compared with the template threshold value of formula 3-5, so as to judge characterization image
Numeral, specifically judge that flow process is:When NA ', NB ', NC ', ND ' coincidence formulas 3, then judge numeral as 9;Work as NA ', NB ',
NC ', ND ' do not meet formula 3 and during coincidence formula 4, then judge numeral as 6;Work as NA ', NB ', NC ', ND ' and do not meet formula 3 and 4
And during coincidence formula 5, then judge numeral as 0.
Wherein, formula 3,4,5 is as follows:
(NA’+NC’+ND’)/3-NB’>φ1(formula 3)
(NA’+NB’+ND’)/3-NC’>φ2(formula 4)
|NA’-(NA’+NB’+NC’+ND’)|+|NB’-(NA’+NB’+NC’+ND’)|+|NC’-(NA’+NB’+NC’+
ND’)|+|ND’-(NA’+NB’+NC’+ND’)|>φ3
(formula 5)
Wherein, φ1、φ2、φ3, the respectively template threshold value of formula 3,4,5.Template Threshold is using normal state point
Cloth method is identical with the determination step of metrics-thresholds.
According to above-mentioned flow process and formula, pass sequentially through template matching and judge whether numeral is 9,6,0.If above-mentioned formula is equal
It is unsatisfactory for, then process decision chart is as None- identified.
If 3b3) connected domain quantity n is 1, the numeral of characterization image is 1,2,3,4,5, one of 7, is now carried out further
Numeral identification, flow process is as shown in figure 12, by picture traverse and NA ', NB ', NC ', ND ' compared with the template threshold value of formula 6-12
Compared with so as to judge the numeral of characterization image, specifically judging that flow process is:
When picture traverse coincidence formula 6, then judge numeral as 1;When picture traverse does not meet formula 6, according to NA ',
NB ', NC ', ND ' proceed to judge;When NC ' coincidence formulas 7, continuation judges NA ', NC ', ND ' whether coincidence formula 8;When
When NA ', NC ', ND ' coincidence formulas 8, then judge numeral as 4;When NA ', NC ', ND ' do not meet formula 8, coincidence formula 9, then
Judge numeral as 7;When NC ' coincidence formulas 7 and NA ', NC ', ND ' do not meet formula 8 and formula 9, then image fails identification.
When NC ' does not meet formula 7, continuation judges NA ', NC ', ND ' whether coincidence formula 10;During coincidence formula 10, then
Judge numeral as 5;When NA ', NC ', ND ' do not meet formula 10 and NA ', NB ', NC ', ND ' coincidence formulas 11, then numeral is judged
For 2;When NA ', NB ', NC ', ND ' do not meet formula 10 and formula 11 and coincidence formula 12, then judge numeral as 3;Work as NA ',
When NB ', NC ', ND ' do not meet formula 7, formula 10, formula 11 and formula 12, then image fails identification.
Wherein formula 6-12 is as follows:
width<η1(formula 6)
NC’<η2(formula 7)
|NA’-(NA’+NC’+ND’)/3|<η3(formula 8)
|NC’-(NA’+ND’)/2|>η4(formula 9)
|NA’-(NA’+NC’+ND’)/3|<η5(formula 10)
NB’+NC’-NA’-ND’>η6(formula 11)
NB’+ND’-NA’-NC’>η7(formula 12)
Wherein, width is picture traverse;η1、η2、η3、η4、η5、η6、η7Respectively formula 6 arrives the template threshold value of formula 12.
Template Threshold adopts normal distribution method, identical with the determination step of metrics-thresholds.
According to above-mentioned flow process and formula, the steps such as picture traverse calculating, template matching are passed sequentially through, judge that whether numeral is
1、2、3、4、5、7.If above-mentioned formula is unsatisfactory for, process decision chart is as None- identified.
According to above-mentioned flow process, each repetition of figures is recognized three times 3B., respectively to three numeral identifications of Fig. 7, then with mould
Plate numeral is compared, and judges whether numeral is correct, so as to whether decision element model is correct.
Finally, it is complete through paster positioning and Size calculation, paster region color analysis, three steps of paster type identifier
The detection of paster in pairs.
Embodiment
The implementation process of the present invention is illustrated below by way of specific embodiment:
According to a kind of printed circuit board (PCB) paster defect inspection method as shown in Figure 2, through paster positioning and Size calculation,
Paster region color analysis, three steps of paster type identifier, realize paster defects detection.
The paster that the present embodiment detection object is size 0402 (metric system), paster type figure is " 002 ", such as Figure 14 institutes
Show.Detecting step is as follows:
1) paster positioning and Size calculation:
1a) according to PCB patch location information, by framing technology, the overall patch location of positioning simultaneously intercepts paster figure
Picture.Separate picture red channel, the red channel image to isolating carries out binarization operation, index for selection threshold value, and to figure
As carrying out a denoising.
Wherein, binarization operation adopts Two-peak method, metrics-thresholds to determine that method adopts normal distribution method.
Binarization operation adopts Two-peak method:
Figure 13 is the coordinate diagram of image Color Channel or gray scale, wherein, x-axis characterizes color component value, y-axis phenogram picture
The quantity of the pixel of middle correspondence x-axis color component, segmentation threshold select background and prospect bimodal between minimum point, i.e. pixel
The point of minimum number.
Preferably, the binary-state threshold selects to be 0.233 × 255.
Metrics-thresholds determine that method adopts normal distribution method:
(1) multiple template image is selected, using algorithm the sample value (a at threshold value differentiation is calculated1,a2,a3……an) (can
Being the quantity of size, or qualified point);
(2) normal distribution model is used, the mean μ and variance ε of normal distribution is calculated;
(3) altimetric image to be checked is analyzed, calculates numerical value a at metrics-thresholds differentiation, and compared with normal distribution model, such as a
Fall in the range of the ε of u ± 3, then judge to meet metrics-thresholds condition, otherwise, it is determined that herein parameter does not meet metrics-thresholds.
1b) according to above-mentioned patch location and the red component of image, using the metrics-thresholds chosen, using Threshold segmentation skill
Art, is partitioned into the red area of the end pad of paster two, as flat site.
1c) all connected regions in the region that is partitioned into are marked, left side white portion is designated as 1st area, and right side is white
Zone domain is designated as 2nd area.In the horizontal and vertical coordinate in 1st area minima and maximum be designated as respectively minx1, miny1, maxx1,
maxy1;Minima and maximum are designated as respectively minx2, miny2, maxx2, maxy2 in the horizontal and vertical coordinate in 2 area.Then
Parameter t is distinguished according to formula 11、t2、t3、t4、t5、t6, and carry out with corresponding index upper limit threshold and index lower threshold
Relatively, if index is in the range of index upper limit threshold and index lower threshold, judge that patch size is correct, otherwise record
Not meet the project of metrics-thresholds.
50 width template images are sample used in this example, based on normal distribution model, the metrics-thresholds for calculating and calculating
Value such as table 3.
The metrics-thresholds of table 3
1d) according to the image upper left corner and the angular coordinate in the lower right corner, image interception is carried out, intercept method is as shown in table 1.Cut
Take bak stay image as shown in figure 14.Considered based on algorithm robustness, wherein, the angular coordinate of intercepting choose (miny1-2,
Minx1-2) and (maxy2+2, maxx2+2), intercepting region area is (maxy2-miny1+4) × (maxx2-minx1+4), figure
As long and width is recorded as y1、x1.Y in this example1It is 271, x1It is 55.
2) paster region color analysis:
2a) as shown in figure 16, from step 1d) intercept after paster image on intercept 6 block feature regions, region is designated as respectively
A, region B, region C, region D, region E, region F.The area starting point and area intercepted in this example is as shown in table 4.
Starting point and area that the image representative region of table 4 is intercepted
2b) color analysis are carried out respectively to above-mentioned 6 block feature region.
Every piece of Regional Red component meansigma methodss are calculated according to formula 2.Used in this example 50 width template images be sample, base
In normal distribution model, the red component threshold value for calculating and meansigma methodss such as table 5.
The characteristic area red component threshold value of table 5 and meansigma methodss
* note:E areas and F areas are only judged using image red component upper limit threshold in this example
As shown in Table 3, pixel red component meansigma methodss (mcol) meets the red component upper limit in every piece of region of this example
Threshold value and red component lower threshold, therefore can determine that solder joint zero defect.
3) paster type identifier:
3A. numeral identifications:After solder joint detection in paster region color analysis are completed, digital identification is carried out, numeral is known
Other flow process is as shown in Figure 8.
3a) in step 1d) in intercept after paster image on the basis of, according to the size of template image in testing image
Intercept numeric area image.Binary conversion treatment, agriculture products threshold value, the index threshold in the present embodiment are carried out to digital block area image
It is worth for 57, and a Denoising disposal is carried out to image.
Wherein, the determination method and step 1a of binary conversion treatment and metrics-thresholds) in it is identical.
Then, using the method for zone marker, by numeric area difference labelling in image.As shown in figure 17, three numerals
Zone marker is 1st ' area, 2nd ' area, 3rd ' area.Maximum and minimum value in y, the x coordinate value of each connected domain is calculated, (max1 ' is recorded as
Y, max1 ' x), (max2 ' y, max2 ' x), (max3 ' y, max3 ' x), (min1 ' y, min1 ' x), (min2 ' y, min2 ' x),
(min3 ' y, min ' 3x).Be followed successively by this example (53,113), (117,113), (182,113), (3,21), (64,18), (128,
18).Corresponding digital block is intercepted as in left comer and bottom right angular coordinate according to corresponding y, the maximum of x coordinate and minima
Domain.In view of algorithm robustness, the numeric area of intercepting is more bigger than calculated region.For the area that connected component labeling is 1 '
Domain, angular coordinate is (max1 ' y+2, max1 ' x+2) and (min1 ' y-2, min1 ' x-2);For the area that connected component labeling is 2 '
Domain, angular coordinate is (max2 ' y+2, max2 ' x+2) and (min2 ' y-2, min2 ' x-2);For the area that connected component labeling is 3 '
Domain, angular coordinate is (max3 ' y+2, max3 ' x+2) and (min3 ' y-2, min3 ' x-2);1 ' region angular coordinate is in this example
(55,115) and (1,19);2 ' region angular coordinates for (119,115) and (62,16);3 ' region angular coordinates for (184,115)
(126,16), image interception is carried out according to above-mentioned angular coordinate.
Upper left, lower-left, upper right, the white portion of bottom right of image are taken, A ', B ', C ', D ' is respectively labeled as, each region
White pixel point quantity is denoted as respectively NA ', NB ', NC ', ND '.Intercept rule such as table 2.In the present embodiment, paster includes three numbers
Word, y2Respectively 52,53,54;x2Respectively 94,95,95.The intercepting starting point coordinate of each digital picture and area such as table 6-
8。
The image interception starting point coordinate and area of the first digit of table 6 " 0 "
The image interception starting point coordinate and area of the second digit of table 7 " 0 "
The image interception starting point coordinate and area of the third digit of table 8 " 2 "
From the feature of printing digital, the possibility shape at four angles of image is divided into three species as shown in Figure 10
Type:A types, b types and c types, sort by the quantity of white pixel point, and at most, b types take second place c types, and a types are minimum.Different types of angle point
Between white pixel point quantity difference it is larger, and the white pixel quantitative difference between the angle point of same type is less.The present embodiment
In, each digital corresponding white pixel point number such as table 9.
The white pixel point quantity of the different digital white portion of table 9
Then, individual digit image is identified.Connected domain quantity is as index in image with numeric area as background.
In this example, paster numeral is respectively " 0 ", " 0 " and " 2 ", therefore connected domain n is respectively 2,2 and 1.
When connected domain quantity is 2, numeral is determined whether according to flow process Figure 11.
50 width template images are sample used in this example, based on normal distribution model, the template threshold value for calculating such as table 10
(only calculating required unilateral threshold value).
Template threshold value when the connected domain quantity of table 10 is 2
From formula 3, formula 4 and formula 5, white pixel point quantity (NA ', NB ', NC ', ND ') of white portion with
The relation of template threshold value is unsatisfactory for formula 3 and formula 4, meets formula 5, then can determine whether that digital picture is " 0 ".
When connected domain quantity is 1, numeral is determined whether according to Figure 12.50 width template images are sample used in this example
This, based on normal distribution model, the template threshold value for calculating such as table 11 (only calculates required unilateral threshold value).
Template threshold value when the connected domain quantity of table 11 is 1
From formula 6-12, the picture traverse of white portion be unsatisfactory for formula 6, white pixel point quantity (NA ', NB ',
NC ', ND ') formula 7 and formula 10 are unsatisfactory for the relation of template threshold value, meet formula 11, then can determine whether that digital picture is
“2”。
According to above-mentioned flow process, each repetition of figures is recognized three times, respectively to three numeral identifications of Figure 17, then with template
Digital threshold is compared, and judges whether numeral is correct, so as to whether decision element model is correct.
Finally, it is real through paster positioning and Size calculation, paster region color analysis, three steps of paster type identifier
Now to the detection of paster.
Claims (3)
1. a kind of printed circuit board (PCB) paster defect inspection method, it is characterised in that:
Methods described comprises the steps:
1) paster positioning and Size calculation:
1a) according to PCB patch location information, by framing technology, the overall patch location of positioning simultaneously intercepts paster image;Point
From image red channel, the red channel image to isolating carries out binarization operation, index for selection threshold value, and image is carried out
Denoising disposal;
1b) according to above-mentioned patch location and the red component of image, using the metrics-thresholds chosen, using Threshold sementation,
It is partitioned into the red area of the end pad of paster two, as flat site;
1c) all connected regions in the region that is partitioned into are marked, left side white portion is designated as 1st area, right side white area
Domain is designated as 2nd area;Minima and maximum are designated as respectively minx1, miny1, maxx1, maxy1 in the horizontal and vertical coordinate in 1 area;
Minima and maximum are designated as respectively minx2, miny2, maxx2, maxy2 in the horizontal and vertical coordinate in 2 area;According to formula 1
Difference parameter t1、t2、t3、t4、t5、t6, and be compared with corresponding index upper limit threshold and index lower threshold;
If index is in the range of index upper limit threshold and index lower threshold, judge that patch size is correct, otherwise record
Not meet the project of metrics-thresholds bound;
1d) according to the image upper left corner and the angular coordinate in the lower right corner, image interception is carried out, intercept method is:With the image upper left corner
Used as coordinate origin, abscissa is y directions, and the paster y directions length after cutting is y1;Vertical coordinate is x directions, after cutting
Paster x directions length is x1;Wherein, the angular coordinate of intercepting chooses (miny1-2, minx1-2) and (maxy2+2, maxx2+
2) it is (maxy2-miny1+4) × (maxx2-minx1+4), to intercept region area;
2) paster region color analysis:
2a) from step 1d) intercept after paster image on intercept 6 block feature regions, be designated as respectively region A, region B, region C,
Region D, region E, region F, record the area starting point coordinate and area for intercepting;
2b) to step 2a) in 6 block feature regions carry out color analysis respectively, using in region pixel red component it is average
Value (mcol) is used as index, and formula is as follows:
Wherein, col is each pixel red color component value in region;N is pixel number in region;
Every piece of Regional Red component meansigma methodss are calculated according to formula 2;
When pixel red component meansigma methodss (mcol) meet under red component upper limit threshold and red component in every piece of region
During limit threshold value, then solder joint zero defect is judged;If there is not meeting red component upper limit threshold and red component lower threshold
Situation, then judge that solder joint is defective, and records defect area;
3) paster type identifier:
3A. numeral identifications:After solder joint detection in paster region color analysis are completed, digital identification is carried out;
3a) in step 1d) in intercept after paster image on the basis of, intercepted in testing image according to the size of template image
Numeric area image;Binary conversion treatment, agriculture products threshold value are carried out to digital block area image, and a denoising is carried out to image
Process;
3b) using the method for zone marker, three numeric areas in image are respectively labeled as into 1st ' area, 2nd ' area, 3rd ' area.Calculate every
Maximum and minimum value in y, the x coordinate value of individual connected domain, be designated as (max1 ' y, max1 ' x), (max2 ' y, max2 ' x), (max3 '
Y, max3 ' x), (min1 ' y, min1 ' are x), (min2 ' y, min2 ' are x), (min3 ' y, min3 ' is x);According to corresponding y, x coordinate
Maximum and minima intercept corresponding numeric area as the upper left corner and bottom right angular coordinate;The numeric area of intercepting is more than calculating
The region for obtaining;For connected component labeling for 1 ' region, angular coordinate be (max1 ' y+2, max1 ' x+2) and (min1 ' y-2,
min1’x-2);For connected component labeling for 2 ' region, angular coordinate be (max2 ' y+2, max2 ' x+2) and (min2 ' y-2,
min2’x-2);For connected component labeling for 3 ' region, angular coordinate be (max3 ' y+2, max3 ' x+2) and (min3 ' y-2,
min3’x-2);
Upper left, lower-left, upper right, the white portion of bottom right of image are taken, A ', B ', C ', D ' is respectively labeled as, each region white
Pixel quantity is denoted as respectively NA ', NB ', NC ', ND ';
Then, individual digit image is identified;Connected domain quantity is index to figure in image with numeric area as background
As the numeral that may be characterized is classified, sorting technique is as follows:
If 3b1) connected domain quantity is 3, the numeral of characterization image is 8;
If 3b2) connected domain quantity is 2, the numeral of characterization image is 9,6, one of 0, now carries out further numeral identification, will
NA ', NB ', NC ', ND ', so as to judge the numeral of characterization image, specifically judge flow process compared with the template threshold value of formula 3-5
For:When NA ', NB ', NC ', ND ' coincidence formulas 3, then judge numeral as 9;Work as NA ', NB ', NC ', ND ' do not meet formula 3 and
During coincidence formula 4, then judge numeral as 6;When NA ', NB ', NC ', ND ' do not meet formula 3 and 4 and coincidence formula 5, then judge
Numeral is 0;
Wherein, formula 3,4,5 is as follows:
(NA’+NC’+ND’)/3-NB’>φ1Formula 3
(NA’+NB’+ND’)/3-NC’>φ2Formula 4
|NA’-(NA’+NB’+NC’+ND’)|+|NB’-(NA’+NB’+NC’+ND’)|+|NC’-(NA’+NB’+NC’+ND’)|+|
ND’-(NA’+NB’+NC’+ND’)|>φ3
Formula 5
Wherein, φ1、φ2、φ3The respectively template threshold value of formula 3,4,5;
According to above-mentioned flow process and formula, pass sequentially through template matching and judge whether numeral is 9,6,0;If above-mentioned formula is discontented with
Foot, then process decision chart is as None- identified;
If 3b3) connected domain quantity is 1, the numeral of characterization image is 1,2,3,4,5, one of 7, now carries out further numeral
Identification, picture traverse and NA ', NB ', NC ', ND ' is compared with the template threshold value of formula 6-12, so as to judge characterization image
Numeral, specifically judge that flow process is:
When picture traverse coincidence formula 6, then judge numeral as 1;When picture traverse does not meet formula 6, according to NA ', NB ',
NC ', ND ' proceed to judge;When NC ' coincidence formulas 7, continuation judges NA ', NC ', ND ' whether coincidence formula 8;Work as NA ',
When NC ', ND ' coincidence formulas 8, then judge numeral as 4;When NA ', NC ', ND ' do not meet formula 8 and coincidence formula 9, then judge
Numeral is 7;When NC ' coincidence formulas 7 and NA ', NC ', ND ' do not meet formula 8 and formula 9, then image fails identification;
When NC ' does not meet formula 7, continuation judges NA ', NC ', ND ' whether coincidence formula 10;During coincidence formula 10, then judge
Numeral is 5;When NA ', NC ', ND ' do not meet formula 10 and NA ', NB ', NC ', ND ' coincidence formulas 11, then judge numeral as 2;
When NA ', NB ', NC ', ND ' do not meet formula 10 and formula 11 and coincidence formula 12, then judge numeral as 3;Work as NA ', NB ',
When NC ', ND ' do not meet formula 7, formula 10, formula 11 and formula 12, then image fails identification;
Wherein formula 6-12 is as follows:
width<η1Formula 6
NC’<η2Formula 7
|NA’-(NA’+NC’+ND’)/3|<η3Formula 8
|NC’-(NA’+ND’)/2|>η4Formula 9
|NA’-(NA’+NC’+ND’)/3|<η5Formula 10
NB’+NC’-NA’-ND’>η6Formula 11
NB’+ND’-NA’-NC’>η7Formula 12
Wherein, width is picture traverse;η1、η2、η3、η4、η5、η6、η7Respectively formula 6 arrives the template threshold value of formula 12;
According to above-mentioned flow process and formula, pass sequentially through the steps such as picture traverse calculating, template matching, judge numeral be whether 1,2,
3、4、5、7;If above-mentioned formula is unsatisfactory for, process decision chart is as None- identified;
3B. compares with template numeral, judges whether numeral is correct, so as to whether decision element model is correct.
2. a kind of printed circuit board (PCB) paster defect inspection method as claimed in claim 1, it is characterised in that:The paster model
Numeral identification number of times is recognized three times for each repetition of figures in identification step.
3. a kind of printed circuit board (PCB) paster defect inspection method as claimed in claim 1, it is characterised in that:Described binaryzation
Operation adopts Two-peak method, metrics-thresholds to adopt normal distribution method with template Threshold.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710005017.XA CN106651857B (en) | 2017-01-04 | 2017-01-04 | A kind of printed circuit board patch defect inspection method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710005017.XA CN106651857B (en) | 2017-01-04 | 2017-01-04 | A kind of printed circuit board patch defect inspection method |
Publications (2)
Publication Number | Publication Date |
---|---|
CN106651857A true CN106651857A (en) | 2017-05-10 |
CN106651857B CN106651857B (en) | 2019-06-04 |
Family
ID=58843679
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710005017.XA Active CN106651857B (en) | 2017-01-04 | 2017-01-04 | A kind of printed circuit board patch defect inspection method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106651857B (en) |
Cited By (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107945184A (en) * | 2017-11-21 | 2018-04-20 | 安徽工业大学 | A kind of mount components detection method positioned based on color images and gradient projection |
CN108665439A (en) * | 2017-08-22 | 2018-10-16 | 深圳安博电子有限公司 | Method of testing substrate and terminal device |
CN109358070A (en) * | 2018-08-31 | 2019-02-19 | 广州超音速自动化科技股份有限公司 | Pole piece detection method, electronic equipment, storage medium and system |
CN111429444A (en) * | 2020-04-02 | 2020-07-17 | 苏州杰锐思智能科技股份有限公司 | Chip appearance detection method and device, electronic equipment and storage medium |
CN111465312A (en) * | 2020-04-14 | 2020-07-28 | 杭州洛微科技有限公司 | Photoelectric product packaging production method based on periodic array arrangement |
CN111862057A (en) * | 2020-07-23 | 2020-10-30 | 中山佳维电子有限公司 | Picture labeling method and device, sensor quality detection method and electronic equipment |
CN112579810A (en) * | 2019-09-30 | 2021-03-30 | 深圳市嘉立创科技发展有限公司 | Printed circuit board classification method and device, computer equipment and storage medium |
CN112579540A (en) * | 2020-11-03 | 2021-03-30 | 珠海越亚半导体股份有限公司 | Component mounting position identification method, mounting control method, device and medium |
CN117078620A (en) * | 2023-08-14 | 2023-11-17 | 正泰集团研发中心(上海)有限公司 | PCB welding spot defect detection method and device, electronic equipment and storage medium |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US5517234A (en) * | 1993-10-26 | 1996-05-14 | Gerber Systems Corporation | Automatic optical inspection system having a weighted transition database |
US7142708B2 (en) * | 2001-06-22 | 2006-11-28 | Hitachi, Ltd. | Defect detection method and its apparatus |
CN101477066A (en) * | 2009-01-09 | 2009-07-08 | 华南理工大学 | Circuit board element mounting/welding quality detection method and system based on super-resolution image reconstruction |
-
2017
- 2017-01-04 CN CN201710005017.XA patent/CN106651857B/en active Active
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US5517234A (en) * | 1993-10-26 | 1996-05-14 | Gerber Systems Corporation | Automatic optical inspection system having a weighted transition database |
US7142708B2 (en) * | 2001-06-22 | 2006-11-28 | Hitachi, Ltd. | Defect detection method and its apparatus |
CN101477066A (en) * | 2009-01-09 | 2009-07-08 | 华南理工大学 | Circuit board element mounting/welding quality detection method and system based on super-resolution image reconstruction |
Non-Patent Citations (5)
Title |
---|
QINGXIANG WANG 等: "Unsupervised Defect Detection of Flexible Printed Circuit Board Gold Surfaces Based on Wavelet Packet Frame", 《2010 2ND INTERNATIONAL CONFERENCE ON INDUSTRIAL AND INFORMATION SYSTEMS》 * |
张纪铃: "电路板板载元器件检测***研究", 《中国优秀博硕士学位论文全文数据库(硕士)信息科技辑》 * |
曹亮: "基于机器视觉的电路板检测***与方法的研究", 《中国优秀硕士学位论文全文数据库信息科技辑》 * |
薛骏: "基于机器视觉的表面贴装元件检测技术研究", 《中国优秀硕士学位论文全文数据库信息科技辑》 * |
陈臣: "印刷电路板的自动光学检测***的设计与研究", 《中国优秀硕士学位论文全文数据库信息科技辑》 * |
Cited By (14)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108665439A (en) * | 2017-08-22 | 2018-10-16 | 深圳安博电子有限公司 | Method of testing substrate and terminal device |
CN107945184B (en) * | 2017-11-21 | 2020-10-09 | 安徽工业大学 | Surface-mounted component detection method based on color image segmentation and gradient projection positioning |
CN107945184A (en) * | 2017-11-21 | 2018-04-20 | 安徽工业大学 | A kind of mount components detection method positioned based on color images and gradient projection |
CN109358070A (en) * | 2018-08-31 | 2019-02-19 | 广州超音速自动化科技股份有限公司 | Pole piece detection method, electronic equipment, storage medium and system |
CN112579810B (en) * | 2019-09-30 | 2023-10-27 | 深圳市嘉立创科技发展有限公司 | Printed circuit board classification method, device, computer equipment and storage medium |
CN112579810A (en) * | 2019-09-30 | 2021-03-30 | 深圳市嘉立创科技发展有限公司 | Printed circuit board classification method and device, computer equipment and storage medium |
CN111429444A (en) * | 2020-04-02 | 2020-07-17 | 苏州杰锐思智能科技股份有限公司 | Chip appearance detection method and device, electronic equipment and storage medium |
CN111429444B (en) * | 2020-04-02 | 2024-03-26 | 苏州杰锐思智能科技股份有限公司 | Chip appearance detection method and device, electronic equipment and storage medium |
CN111465312A (en) * | 2020-04-14 | 2020-07-28 | 杭州洛微科技有限公司 | Photoelectric product packaging production method based on periodic array arrangement |
CN111862057A (en) * | 2020-07-23 | 2020-10-30 | 中山佳维电子有限公司 | Picture labeling method and device, sensor quality detection method and electronic equipment |
CN111862057B (en) * | 2020-07-23 | 2021-10-22 | 中山佳维电子有限公司 | Picture labeling method and device, sensor quality detection method and electronic equipment |
CN112579540A (en) * | 2020-11-03 | 2021-03-30 | 珠海越亚半导体股份有限公司 | Component mounting position identification method, mounting control method, device and medium |
CN117078620A (en) * | 2023-08-14 | 2023-11-17 | 正泰集团研发中心(上海)有限公司 | PCB welding spot defect detection method and device, electronic equipment and storage medium |
CN117078620B (en) * | 2023-08-14 | 2024-02-23 | 正泰集团研发中心(上海)有限公司 | PCB welding spot defect detection method and device, electronic equipment and storage medium |
Also Published As
Publication number | Publication date |
---|---|
CN106651857B (en) | 2019-06-04 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106651857B (en) | A kind of printed circuit board patch defect inspection method | |
CN107945184B (en) | Surface-mounted component detection method based on color image segmentation and gradient projection positioning | |
CN109900711A (en) | Workpiece, defect detection method based on machine vision | |
CN106501272B (en) | Machine vision soldering tin positioning detection system | |
KR940003829B1 (en) | Position recognizing method | |
CN115205290B (en) | Online detection method and system for PCB production process | |
CN106651802B (en) | Machine vision scolding tin position finding and detection method | |
CN110108712A (en) | Multifunctional visual sense defect detecting system | |
KR20180095972A (en) | High-speed automated optical inspection apparatus supporting double scan approach | |
CN112884743B (en) | Detection method and device, detection equipment and storage medium | |
CN111861979A (en) | Positioning method, positioning equipment and computer readable storage medium | |
CN105136818B (en) | The image detection method of printed base plate | |
CN111275674A (en) | Chip burning detection method and system and computer readable storage medium | |
US6633663B1 (en) | Method and system for determining component dimensional information | |
CN117589770A (en) | PCB patch board detection method, device, equipment and medium | |
CN117649404A (en) | Medicine packaging box quality detection method and system based on image data analysis | |
CN110322395B (en) | Part outline shape detection method and device based on image processing and affine transformation | |
CN109060799A (en) | A kind of assembling line finished product detection determination method | |
CN115452844B (en) | Injection molding part detection method and system based on machine vision | |
CN114092448B (en) | Plug-in electrolytic capacitor mixed detection method based on deep learning | |
CN111507949A (en) | Chip identification method based on vision | |
CN116465908A (en) | Optical detection method for printed circuit board production | |
CN112837285B (en) | Edge detection method and device for panel image | |
CN206470210U (en) | Machine vision scolding tin position detecting system | |
JPH08272078A (en) | Method and apparatus for inspecting pattern |
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 |