CN107248152A - Textile flaw detection method based on class grid pattern and its area features - Google Patents
Textile flaw detection method based on class grid pattern and its area features Download PDFInfo
- Publication number
- CN107248152A CN107248152A CN201710350302.5A CN201710350302A CN107248152A CN 107248152 A CN107248152 A CN 107248152A CN 201710350302 A CN201710350302 A CN 201710350302A CN 107248152 A CN107248152 A CN 107248152A
- Authority
- CN
- China
- Prior art keywords
- grid pattern
- class grid
- image
- textile
- value
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0004—Industrial image inspection
- G06T7/0008—Industrial image inspection checking presence/absence
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/13—Edge detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/60—Analysis of geometric attributes
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/60—Analysis of geometric attributes
- G06T7/62—Analysis of geometric attributes of area, perimeter, diameter or volume
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
- G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
-
- 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
- G06T2207/10004—Still image; Photographic 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
- G06T2207/20—Special algorithmic details
- G06T2207/20048—Transform domain processing
- G06T2207/20061—Hough transform
-
- 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
- G06T2207/30124—Fabrics; Textile; Paper
Abstract
The present invention relates to a kind of textile flaw detection method based on class grid pattern and its area features, based on isotropism class grid pattern dividing method and areal calculation, automatic segmentation class grid pattern in textile images that can be identical from texture but different angle, the feature extraction based on areal calculation for class grid pattern is recognized with flaw, the Pixel of Digital Image half-tone information based on flat textile surface under lighting source is analyzed, textile surface flaw is automatically positioned.The present invention is especially suitable for the textile surface flaw being automatically identified in the digital picture of the textile flat surfaces that texture is identical but angle is different gathered under steady illumination light source.
Description
Technical field
The present invention relates to a kind of textile flaw detection method based on class grid pattern and its area features.
Background technology
Traditional textile flaw manual identified accuracy rate only have 60-75% (K.Srinivasan, P.H.Dastoor,
P.Radhakrishnaiah,et al..FDAS:a knowledge-based frameworkfor analysis of
Defects in woven textiles tructures, J.Text.Inst.83 (1992) 431-448.), machine is known automatically
The method of other textile flaw has practical application request.The digital picture sampling of flat textile surface (is hereinafter referred to as weaved
Product image) belong to 2 d texture, 2 d texture has been demonstrated the figure that can be defined according to 17 kinds of wallpaper groups (wallpaper group)
Case aligning method generates (K.Srinivasan, P.H.Dastoor, P.Radhakrishnaiah, et al..FDAS:
aknowledge-based framework for analysis of defects in woven textiles
Tructures, J.Text.Inst.83 (1992) 431-448.), the pattern referred to as lattice for generating 2 d texture
(lattice), the inside pattern of lattice is referred to as motif.Most textile flaw automatic detection methods can only handle p1 in wallpaper group
Textile images (H.Y.T.Ngan, G.K.H.Pang, N.H.C.Yung.Automated the fabric defect of type
Detection-A review, Imageand Vision Computing 29 (7) (2011) 442-458.), only have minority side
Method can handle the textile images beyond p1 types, such as the benchmark image difference method (wavelet- based on wavelet pretreatment
Pre-processed golden imagesubtraction, hereinafter referred to as WGIS, come from document：H.Y.T.Ngan,
G.K.H.Pang,N.H.C.Yung,etal.,Wavelet based methods on patterned fabric defect
Detection, Pattern Recognit.38 (4) (2005) 559-576), co-occurrence matrix method, cloth forest belt method
(Bollinger bands, hereinafter referred to as BB, come from document：H.Y.T.Ngan,G.K.H.Pang,Novel method for
patterned fabric inspection using Bollinger bands,Opt.Eng.45(8)(2006)087202-
1-087202-15), (regular bands, hereinafter referred to as RB, come from document to rule band method：H.Y.T.Ngan,
G.K.H.Pang,Regularity analysis for patterned texture inspection,IEEE
Trans.Autom.Sci.Eng.6 (1) (2009) 131-144), and Elo appraisal procedures (Elo rating method, hereinafter referred to as
ER, comes from document C.S.C.Tsang, H.Y.T.Ngan, G.K.H.Pang, Fabric inspection based on the
Elo rating method, Pattern Recognit.51 (2016) 378-394.) etc..Although these methods can handle p1
Textile images in addition, but pattern (the following letter of the similar lattice based on artificial selection is built upon their computational methods more
Claim class grid pattern) on.Such as WGIS requires the size and texture of artificial selection class grid pattern, and BB, RB and ER requirements are artificial fixed
The size of adopted class grid pattern.These prioris reduce the automation journey of machine recognition textile flaw to a certain extent
Degree.
The content of the invention
The technical problem to be solved in the present invention is：It is still based on manually to solve existing textile flaw automatic detection method
Selection or Manual definition, the not high deficiency of automaticity, the present invention provides a kind of based on class grid pattern and its area features
Textile flaw detection method.
The technical solution adopted for the present invention to solve the technical problems is：It is a kind of based on class grid pattern and its area features
Textile flaw detection method, comprises the following steps：
The textile images of input gray level；
Calibrate image：Ideal calibration angle is determined using Hough transform, is calibrated with ideal calibration angle rotation image
Image afterwards；
Class grid pattern is split：Image is to produce class grid pattern after segmentation calibration, and class grid pattern is met：Relative to textile figure
The row and column of picture, class grid pattern is transversely arranged by the direction of image line, and by the direction longitudinal arrangement of row；In anatomic element analysis
The textile patter cartoon composition I of methodcIn, class grid pattern has geometry and has significance difference in gray scale with background pixel
It is different；
Areal calculation：Calculate the area of class grid pattern；
Histogram analysis：The distribution of analysis classes grid pattern area, i.e. class grid pattern area histogramIfOccur different
Constant value is then to that should have free time class grid pattern；
According to the pixel index set for thering is the output of free time class grid pattern to represent flaw.
Concretely comprise the following steps：
Calibrate image step：The edge image of input picture is generated using Canny rim detections, will using Hough transform
Edge image is projected to parameter space, the slope of straight line corresponding to the peak value in the space that gets parms, the angle according to corresponding to slope
The negative value rotation input picture of degree and the transverse projection for calculating its background pixelAnd longitudinal projectionTransverse projection is calculated respectively
EntropyWith the entropy of longitudinal projectionAccording to entropy threshold exAnd eyDetermine ideal calibration angleThe angleCorresponding rotation
Image is image after calibration.
Class grid pattern segmentation step：The cartoon composition I of image after calibration is calculated using morphology component analyzing methodc, make
Use threshold value fc·max(Ic) binaryzation cartoon composition IcObtain bianry image Itc, arrange I by ranks indexed sequential respectivelytcIn it is every
The background pixel number of row and each column, i.e., laterally and longitudinally projectWithFind out respectivelyWithPeak valueWithUsing certainly
Adapt to Kmeans algorithms pairWithClustered, screened according to cluster centreWithAccording toWithCalculate and constitute
The set S of the most long continuous line index of stable line spacehAnd constitute the set S of the most long continuous column index of stable column pitchv,
And extend ShAnd SvI is included respectivelytcMost of line index and column index；With ShAnd SvIn ItcThe corresponding ranks of middle difference for point
Boundary, by ItcIt is divided into the class grid pattern of rectangleWherein, i=1,2 ... | Sh| -1, j=1,2 ... | Sv|-1。
Area calculation step：Count each class grid patternThe number a of included foreground pixeli,j, wherein, i=1,2 ...
|Sh| -1, j=1,2 ... | Sv| -1, ai,jAsArea.
Histogram analysis step：Calculate ai,jHistogramWherein, i=1,2 ... | Sh| -1, j=1,2 ... | Sv| -1,
With approximateTrue threshold value is that target adjustment is had no time the interval lower bound a of area0With upper bound a1, will be any interval not in area of having no time
[a0, a1] in ai,jCorrespondingLabeled as there is free time class grid pattern.
The beneficial effects of the invention are as follows the textile Defect Detection side of the invention based on class grid pattern and its area features
Method, based on class grid pattern dividing method can be identical but different angle from texture textile images in automatic segmentation class trrellis diagram
Case, the feature extraction based on areal calculation for class grid pattern recognizes that analysis is based on flat weaving under lighting source with flaw
The Pixel of Digital Image half-tone information on product surface, is automatically positioned textile surface flaw.The present invention is especially suitable for automatic identification
Textile in the digital picture for the textile flat surfaces that the texture gathered under steady illumination light source is identical but angle is different
Surface blemish.
Brief description of the drawings
The present invention is further described with reference to the accompanying drawings and examples.
Fig. 1 is the step schematic diagram of the present invention.
Fig. 2 is the assumed condition displaying figure of the present invention.
Fig. 3 is the textile images background pixel distribution map that texture is identical but direction is different.
Fig. 4 is the general flowchart of image calibration.
Fig. 5 is the general flowchart of image rotation.
Fig. 6 is the general principle figure of algorithm 4.
Fig. 7 is to calculate ShThe general principle figure of initial value.
Fig. 8 is treatment effect figure of the inventive method to box-shaped textile images：(a) it is box-shaped image broken ends of fractured bone scatter diagram；
(b) it is box-shaped vacancy scatter diagram；(c) it is box-shaped image reticulate pattern scatter diagram；(d) it is box-shaped image cord scatter diagram；(e)
For box-shaped image stria scatter diagram.
Fig. 9 is treatment effect figure of the inventive method to star textile images：(a) it is star-shaped image broken ends of fractured bone scatter diagram；
(b) it is star-shaped image hole scatter diagram；(c) it is star-shaped image reticulate pattern scatter diagram；(d) it is star-shaped image cord scatter diagram；(e)
For star-shaped image stria scatter diagram.
Embodiment
In conjunction with the accompanying drawings, the present invention is further explained in detail.
To make statement cheer and bright, existing centralized definition partial symbols involved in the present invention and concept.
1.Represent Positive Integer Set.
2.Expression includes zero integer set.
3.Expression includes zero arithmetic number set.
4.Expression includes zero real number set.
5.T representing matrixs or vectorial transposition.
6.Represent ratioSmall maximum integer, for example
7.{aiRepresent by index i determine by element aiThe set of composition or multiset.
8. | S | represent the element number in set S.
9.avg(S)：The average of set of computations or multiset S, S element is real number.
10.std(S)：The standard deviation of set of computations or multiset S, S element is real number.
11.med(S)：The median of set of computations or multiset S, S element is real number.
12.mod(S)：Multiset S mode is calculated, S element is real number.
13.max (S) represents to find out set or multiset S element maximum, such as max (Ic) represent IcMiddle pixel is most
High-gray level value.
14.Expression is found out qualifiedMaximum.
15.min (S) represents to find out set or multiset S element minimum value, such as min (Ic) represent IcMiddle pixel is most
Small gray value.
16.arg maxsF (s) is represented in function f domain of definition internal variable s span so that function f (s) takes
The s of maximum.
17.arg minsF (s) is represented in function f domain of definition internal variable s span so that function f (s) takes
The s of minimum value.
18.arg maxs f1(s),f2(s) represent in function f1And f2Domain of definition common factor internal variable s span in,
So that function f1And f (s)2(s) s of maximum is taken.
19.arg modi({ai) represent correspondence multiset { aiMode mod ({ ai) index.
20.dimx(I) total line number of two-dimensional image I, dim are representedy(I) I total columns is represented.
21.I (x, y) represents the pixel value that there are ranks to index (x, y) in two-dimensional image I.Line index1≤x
≤dimx(I)；Column index1≤y≤dimy(I)。
22. textile images cartoon composition Ic：To the textile images of a width gray processing, using based on Qu Bo
(curvelet) and discrete cosine transform (local discrete cosine transform, hereinafter referred to as DCT) form into
What point analysis method (morphological component analysis, hereinafter referred to as MCA) was calculated has smooth edge
The image of pattern is referred to as cartoon composition Ic, IcIt is a width gray level image.
23. threshold coefficient fc：For binaryzation IcParameter, the parameter by algorithm 3 calculate obtain.
24. binaryzation cartoon composition Itc：Use fc·max(Ic) it is used as threshold binarization IcResulting bianry image, its
In 1 represent foreground pixel, i.e. IcMiddle gray value is not less than the pixel of threshold value, and 0 represents background pixel.ItcWith IcLine number and columns
It is identical.
25. transverse projection It isMultiset, wherein
1≤k≤dimy(I), i.e.,Represent the background pixel number that line index is x.
26. longitudinal projection It isMultiset, wherein
1≤l≤dimx(I), i.e.,Represent the background pixel number that column index is y.
27.RepresentPeak value multiset,In element be referred to as peak value, peak value refers toIt is middle to meetWithElementWherein x represents line index.
28.RepresentPeak value multiset,In element be referred to as peak value, peak value refers toIt is middle to meetWithElementWherein y represents column index.
29.Expression pairMiddle element use HC algorithms obtained by cluster centre multiset.
30.Expression pairMiddle element use HC algorithms obtained by cluster centre multiset.
31.DCT sizes：MCA divides an image into not overlapping and big with fixing first in image local application DCT
Small rectangular area, then applies DCT to each region, and the size of rectangular area is referred to as DCT sizes, and unit is pixel, region
The pixel count of interior a line is referred to as the width of DCT sizes, and the pixel count of a row is referred to as the height of DCT sizes.
32.WillMiddle element is arranged in descending order, fromStart, calculate it poor with next element
Absolute value, if the value is not more than the height of DCT sizes, continue to calculate currentElement and next element poor absolute value and with
The height of DCT sizes compares, and continues if being not more than, and next element is if being more thanAnd terminate；If never had out
Now it is more than the high situation of DCT sizes, then willAs
33.WillMiddle element is arranged in descending order, fromStart, calculate it poor with next element
Absolute value, if the value is not more than the height of DCT sizes, continue to calculate currentElement and next element poor absolute value and with
The width of DCT sizes compares, and continues if being not more than, and next element is if being more thanAnd terminate；If never had out
Now it is more than the wide situation of DCT sizes, then willAs
34.K：The cluster number of Kmeans algorithms, the cluster number for specifying Kmeans algorithms,
35.S′h：In be not less thanElementLine index x set, S 'hMiddle x by arranging from small to large.
36.S′v：In be not less thanElementColumn index y set, S 'vMiddle y by arranging from small to large.
37.Set S 'hIn latter element and the multiset of previous element difference between adjacent two element, i.e., between line index
Away from multiset,Middle line index spacing is by the x ∈ S ' for producing the spacinghIn higher value arrange from small to large.
38.Set S 'vIn latter element and the multiset of previous element difference between adjacent two element, i.e., between column index
Away from multiset,Middle column index spacing is by the y ∈ S ' for producing the spacingvIn higher value arrange from small to large.
39. stable line space：In withThe absolute value of difference be less thanLine index spacing.I.e.In element be line index spacing, these line index spacing all withMake the difference, poor absolute value is less than
Line index spacing be referred to as stablizing line space.
40. stable column pitch：In withThe absolute value of difference be less thanColumn index spacing.I.e.In element be column index spacing, these column index spacing all withMake the difference, poor absolute value is less than
Column index spacing be referred to as stablizing column pitch.
41.The multiset that line space composition is continuously stablized in line index is met,In stable line space by produce
Two line index x ∈ S ' of the spacinghIn higher value ascending order arrangement,Line index continuously refers toMiddle arrangement
Line index x corresponding to sequentially adjacent stable line spacei∈S′hIn i values it is continuous, such as i can take 2,3,4, but can not be only
Take 2 and 4.
42.The multiset that column index continuously stablizes column pitch composition is met,In stable column pitch by produce
Two column index y ∈ S ' of the spacingvIn higher value ascending order arrangement,Column index continuously refers toMiddle arrangement is suitable
Y corresponding to the adjacent stable column pitch of sequencej∈S′vJ values it is continuous, such as j can take 2,3,4, but can not only take 2 and 4.
43.With most element numbers
44.With most element numbers
45.Sh：ShInitial value for produceLine index x ∈ S ' corresponding to middle stable line spacehSet, by calculation
Method 4 extends, ShRepresent the set of class grid pattern boundary pixel line index.
46.Sv：SvInitial value for produceColumn index y ∈ S ' corresponding to middle stable column pitchvSet, by calculation
Method 4 extends, SvRepresent the set of class grid pattern boundary pixel column index.
47. class grid pattern is demarcated：ShI corresponding to middle line indexcIn row and SvI corresponding to middle column indexcIn
Row.
48. preferable line number
49. preferable columns
50. class grid pattern：In ItcIn, according to ShThe corresponding row of middle line index and SvThe corresponding row of middle column index, by ItcPoint
Rectangular area is segmented into, rectangular area is referred to as class grid pattern, wherein ShThe corresponding row of middle line index and SvMiddle column index is corresponding to be arranged not
In class grid pattern.
51.maxx(a) represent in the set a comprising ranks index (x, y), line index x maximum.
52.minx(a) represent in the set a comprising ranks index (x, y), line index x minimum value
53.maxy(a) represent in the set a comprising ranks index (x, y), column index y maximum.
54.miny(a) represent in the set a comprising ranks index (x, y), column index y minimum value.
55.Represent according to transverse projectionComprising background pixel number calculate entropy.
56.Represent according to transverse projectionComprising background pixel number calculate entropy.
57. entropy threshold exRepresent that according to one group of class grid pattern arrangement angle be 0 ° of textile gray level image calculating of having no timeThe integer part of average.
58. entropy threshold eyRepresent that according to one group of class grid pattern arrangement angle be 0 ° of textile gray level image calculating of having no timeThe integer part of average.
59. class grid pattern area ai,j(wherein i=1,2 ... | Sh| -1, j=1,2 ... | Sv| -1) represent class grid patternInstitute
Number comprising foreground pixel.
Assuming that textile images are constituted by repeating single class grid pattern, if defining image behavior horizontal direction, i.e. angle
For 0 °, if the angle more than zero is the angle value that rotate counterclockwise is obtained since the row of image, then the repetition of class grid pattern
Process can be completed by following steps：1, according to the angle rotating Vortex class grid pattern；2, in the direction with the Vertical Square of the direction
To the class grid pattern for repeating and (replicating and mobile) to rotate through, and ensure that adjacent class grid pattern spacing is 0.Class trrellis diagram in repetitive process
The anglec of rotation of case is referred to as arrangement angle.The arrangement angle of the textile images anglec of rotation and class grid pattern shown in accompanying drawing 3 is anticipated
Justice is identical.
On the basis of being as defined above, technical scheme is now introduced.As shown in brief description of the drawings Fig. 1, the present invention
Method is made up of four parts：(1) image calibration, the segmentation of (2) class grid pattern, (3) areal calculation and (4) histogram analysis, below
Four partial contents are introduced successively by the order in accompanying drawing 1 from left to right.
The function of class grid pattern segmentation is that automatic segmentation textile images produce class grid pattern.Because textile patter is abundant
Various, corresponding class grid pattern species is various.The class grid pattern dividing method of the present invention is set up in the hypothesis to class grid pattern,
I.e.：Inventive method assumes that calibrated textile images class grid pattern has following features：Relative to the row of textile images
And row, class grid pattern is transversely arranged by the direction of image line, and presses the direction longitudinal arrangement of row；In MCA cartoon composition IcIn,
Class grid pattern has geometry and there were significant differences in gray scale with background pixel.Three kinds of situations for example shown in accompanying drawing 2, figure
A kind of situation is shown per a line in 2, often row first row is textile images, and secondary series is Ic, the 3rd row are IcThree-dimensional Mesh
Figure, the 4th row be binaryzation cartoon composition often row background pixel number distribution, the abscissa of the 4th row figure is line index, indulge sit
Mark is background pixel number.The class grid pattern of the first row textile images does not have geometry in Fig. 2, and which results in background pixel
Distribution lacks obvious periodicity；Although the class grid pattern of second row textile images has in geometry, but class grid pattern
Shape is with background in IcIn difference it is small, i.e. the most of region of corresponding Mesh figures is almost flat, and this causes background pixel
Quantity is excessive, and background pixel distribution lacks obvious periodicity；The third line textile images class grid pattern has geometry simultaneously
Difference with background in Ic is big, and the distribution of its background pixel has periodically.
(1) image calibration
Image calibration is set up in the analysis that background pixel is distributed in binaryzation textile images cartoon composition, such as accompanying drawing
Shown in 3.In fig. 3, the textile images positioned at center every 15 ° of rotations once, have rotated altogether 7 times (i.e. since 0 °
0 °, 15 °, 30 ° ...), the image of rotation is arranged around original image in the direction of the clock, the transverse projection of each imageWith it is vertical
To projectionShow to graphically, basis respectively is labeled with below figureWithThe entropy of calculatingWithObserve the institute of accompanying drawing 3
ShowWithIt can be found that the entropy of 0 ° and 90 ° image exceedes the entropy of other rotation images.If there is class grid pattern arrangement angles
Several textile images of having no time for 0 ° are spent, then can calculate each imageWithTake the integer portion of the average of these entropys
Get two threshold value exAnd ey, can be used for judging the arrangement angle of class grid pattern whether close to 0 ° or 90 ° of multiple.exWith
eyCalculating process false code description see algorithm 1.
For the textile images I that class grid pattern arrangement angle is unknown, I is rotated into θ=1 ° successively, 2 °, 3 ° ... 360 ° simultaneously
Calculate correspondingWithIf there is correspondence in angle, θWithMaximumSoReferred to as ideal calibration angle
Degree,Corresponding rotation image is then as calibration result.It is defined as follows.
Above-mentioned calculatingProcess due to θ value it is excessive, computational efficiency is not high, and this law is using flow as shown in Figure 4
CalculateApproximation.As shown in Figure 4, for the unknown textile images I of class grid pattern arrangement angle, the present invention is used
Canny edge detection methods calculate I edge, using Hough transform by edge projection into parameter space, take in parameter space
Preceding nθThe angle, θ of straight slope corresponding to individual peak value, rotates I according to taken θ, obtains nθIndividual rotation image, according to each rotation
ImageWithCalculateWithTakeWithThe corresponding angle of middle maximum, it is approximate with the angle It is right
The rotation image answered is then as final calibration result.Algorithm 2 is shown in the false code description of calibration flow.
It is sky, example by the corner parts of the image of affine transformation although image rotation can be completed by affine transformation
As shown in Figure 5.Number has been arranged in order 5 width images from left to right in accompanying drawing 5, and the first width shows class grid pattern by 37 ° of arrangements
Textile images I, the second pair is shown rotates minus 37 ° of obtained image (original Is using affine transformationr), go out in the image
4 delta-shaped regions of (pixel value is 0) of leaving a blank are showedIf not handling these regions of leaving a blank, then these regions are not only
It can influenceWithCalculating, it is also possible to flaw is identified as by subsequent step.For fillingThis invention takes found outHypotenuseEnd points and parallel toLonger right-angle side straight line, by symmetry axis of the straight line by original IrPixel answer
Make toAlgorithm is shown in 3rd width of number and the fourth officer image vision process from left to right in accompanying drawing 5, its false code description
3.The rightmost of accompanying drawing 5 indicates " final Ir" image be algorithm 3 rotation results, although region of leaving a blank is filled, but some
Artifact (artifacts) is appeared in rotation results, and the class grid pattern arrangement in such as lower left corner occurs in that dislocation.
(2) class grid pattern is split
As shown in Figure 6, the textile images given for a width, class grid pattern segmentation (algorithm 4) calculates I using MCAc
With texture composition, obtained threshold coefficient f is calculated according to algorithm 6c, use threshold value fc·max(Ic) binaryzation IcObtain Itc.Fig. 6
In show IcMesh figure, IcIn two-dimensional pattern three-dimensional " mountain peak ", binaryzation I are shown as in Mesh figurescEquivalent to one
The pixel corresponding to part that individual gray plane is blocked above mountain peak, mountain peak plane saves as the part institute below 1, mountain peak plane
Corresponding pixel saves as 0, and this binaryzation result is Itc, i.e. Fig. 6 lower rights arrow " use threshold binarization IcObtain Itc”
Signified pattern.
Assuming that textile images are at least made up of 4 class grid patterns, then ItcThe object size of middle correspondence class grid pattern should
, should less than the half of picture size, so if occurring in that oversized situation, then this object is not then class grid pattern
From ItcIt is middle to delete oversized object, i.e.,：By Moore-Neighbor track algorithms (Moore-Neighbor tracing
Algorithm I) is obtainedtcThe closure edge of middle object.For each object with closure edge, the object ranks rope is found out
The extreme value drawn, if the absolute value of the difference of the object line index extreme value is more than 0.75dimx(Itc), or column index extreme value difference
Absolute value more than 0.75dimy(Itc), then from ItcMiddle deletion object, i.e., be set to 0 by the pixel of oversized object.
The geometry of textile images class grid pattern is by ItcIn two value objects described by, class grid pattern it is rich and varied
Result in the distribution of background pixel between the diversity of two-value object geometry, but two value objects is influenceed small by its shape, i.e.,：
Two value object of different shapes, if its arrangement in same direction is identical, then the distribution phase of background pixel in this direction
Seemingly.As shown in Figure 6, the background pixel number that statistics binaryzation cartoon composition is often gone with each column, order difference in rows and columns
It is to constitute the transverse projection of background pixel to arrange background pixel numberAnd longitudinal projection WithPeak value be designated as respectively
MultisetWithClose to label " transverse projection during transverse projection and longitudinal projection illustrate i.e. in accompanying drawing 6" throw longitudinal direction
ShadowDark dot, these peak values reflect ItcThe boundary of middle class grid pattern.
Other peak values are filtered to obtain these peak values, it is rightWithClustered respectively using Kmeans algorithms, first
Optimum cluster number K is estimated according to silhouette coefficient, to initialize Kmeans algorithms and right with the K of largest contours coefficientWithCluster, i.e.,：It is 1,2 to calculate K ...Kmeans algorithms pairThe silhouette coefficient of classification, with correspondence largest contours system
Several K initializes Kmeans algorithms and rightClassified, the cluster centre obtained by classification saves as multisetCalculating K is
1,2 ...Kmeans algorithms pairThe silhouette coefficient of classification, Kmeans is initialized with the K of correspondence largest contours coefficient
Algorithm is simultaneously rightClassified, the cluster centre obtained by classification saves as multisetThe false code description of Kmeans algorithms is shown in
Algorithm 5.
Due to the randomness of data,WithIt is possible to include multiple close cluster centres, these close cluster centres
In minimum value elected as threshold value respectivelyWithI.e.：WillDescending is arranged, fromStart, two before and after calculating
The absolute value of the difference of element, it is high that difference is more than DCT sizesFirst element beSimilarly, willDescending is arranged,
FromStart, the absolute value of the difference of two elements before and after calculating, it is wide that difference is more than DCT sizesFirst element be In be not less thanPeak value press corresponding line index, be designated as S 'h；In be not less thanPeak value press institute
Corresponding column index, is designated as S 'v。
Due to the interference of the factors such as flaw, S 'hWith S 'vIn ranks index not necessarily accurately reflect the boundary of class grid pattern.
It is therefore necessary to evaluate S 'hIn whether there is with stablizing the line index of line space, and S 'vIn whether there is and have between stable row
Away from column index, these ranks index is as the boundary of class grid pattern to split class grid pattern.For S 'h, by S 'hMiddle element is done
Ascending order is arranged, and the multiset of line index spacing is defined asHave
The multiset of line space is continuously stablized in line indexIt is defined as follows.
Wherein xi∈S′h, i is the continuous positive integer of numerical value, and such as i can take 2,3,4, but can not only take 2 and 4.Due to can
There can be multiple meetThe continuous spacing of definition, it is possible that existing multipleWherein with most element numbers
(it is denoted as) corresponding to line index be used as class grid pattern boundary set ShInitial value, it is defined as follows.
Accompanying drawing 7 is shown with a secondary textile images transverse projectionBased on calculate ShThe process of initial value, left side shows
ShowPeak valueRepresented with dark dot, it is rightClustered, obtainedPass through cluster in accompanying drawing 77 are obtained
The cluster centre of class, according toCluster centre calculate obtain threshold valueIndicated in figureStraight line on fork.Root
According toScreeningElement, delete compared with small leak, i.e. diagram in the middle of accompanying drawing 7, according to the row rope of the continuous peak value of line index
Draw and calculate its line space, then obtain with different length (element number)Wherein with most elements(i.e. accompanying drawing
Indicated in 7 " maximum") beShInitial value beAs shown in the rightmost side of accompanying drawing 7.Similarly, can be with
Calculate And SvInitial value, be defined as follows.
Because ShWith SvCorresponding to initial valueWithInclude respectively and stablize line space and stablize column pitch, because
ThisWithMedian is defined as preferable line numberWith preferable columns
ShWith SvInitial value example see accompanying drawing 6, as shown in fig. 6, textile images only subregion is by ShAnd SvInitial value cover simultaneously
Lid, i.e., Fig. 6 indicates " ShAnd SvInitial value " diagram in grid, the region of only vertical straight line only has SvCovering.ShAnd Sv's
Extension is based onWithCarry out.Due to ShIn line index by ascending order arrange, from min (Sh) start, with step-lengthTo ItcRow rope
Draw minimum value 1 to move, that is, calculateCheck forIt is full
FootIf there is x ', then x ' is added to Sh, x is otherwise added to ShAnd keep ShMiddle element ascending order row
Row, are calculated againAnd repeat the above steps；Similarly, from max (Sh) start, with step-lengthTo ItcOK
Index maximum dimx(Ic) mobile, that is, calculateCheck forMeetIf there is x ', then x ' is added to Sh, otherwise x is added
Add Sh, calculate againAnd repeat the above steps.To SvWith step-lengthSimilar extension is done, algorithm is referred to
4。
Extend obtained ShWith SvThe most of region of textile images is substantially covers, as shown in Figure 6.According to ShWith Sv
The row and column index included respectively, can be split row and columns of the Ic as where these indexes, the region obtained by segmentation is determined
Justice is class grid pattern, and it is defined as follows.
Wherein1≤i≤|Sh| -1 and 1≤j≤| Sv| -1, therefore IcBe partitioned into (| Sh|-1)·(|Sv
| -1) it is individualAnd i and j are respectivelyRow and column index in units of class grid pattern.One of class grid pattern segmentation is important
Parameter is fc, as shown in fig. 6, IcBinaryzation be based on threshold value fc·max(Ic) complete, and the threshold value depends on fc.If deposited
In multiple indefectible textile images, applied for each image and be based on different fcThe algorithm 1 of value can obtain multipleWith
Obtained by calculatingWithHistogram, wherein occurrence number is mostWithAnd its corresponding fcValue pair determines fc's
Final value has reference significance.Assuming that there are n indefectible textile images I1,I2…InWith m fcSelectable value c1,c2…
cm, orderWithIt is I to represent input respectivelyk, k=1,2 ... n and fc=cl, l=1,2 ... m algorithm 1 calculates
ObtainWithFor each IkAll there are mWith mOn IkMultisetAnd multisetF corresponding to the middle most elements of occurrence numbercValue is designated as c respectivelyh(k, l) and
cv(k, l), is defined as follows.
For IkIf, ch(k, l) and cv(k, l) is identical, then respective index (k, l) is stored in setIn,Definition
It is as follows.
For eachI.e. for IkIf at least there is a l ' and cause ch(k,l′)≡cv(k, l ') is set up, then
With med ({ cl) closest ch(k, l ') participates in fcCalculating, fcIt is defined as follows.
Wherein chThe index of (k, l) Expression, which takes, to be metK values in definition.Calculate fcFalse code
Algorithm 6 is shown in description.
(3) areal calculation
Assuming that textile images I binaryzation cartoon compositions ItcBackground pixel value be 0, then class grid patternArea ai,j
It is defined asThe number of included foreground pixel, whereinI=1,2 ... | Sh| -1 and j=1,2 ... | Sv| -1, i.e.,：
The area of multiple class grid patterns of having no time (the class grid pattern for not including flaw) is often similar, therefore has no time in the presence of one
The interval of class grid pattern area, this interval is referred to as area interval of having no time.If the area of a class grid pattern is not or not face of having no time
In product is interval, then such grid pattern is probably to have free time class grid pattern (the class grid pattern for including flaw).If it is identical to there is n width textures
And class grid pattern orientation identical is had no time textile images Ik, k=1,2 ... n, to each image IkCarry out class grid pattern point
Cut and calculate class grid pattern area, the minimum value and maximum of gained area are designated as a respectively0And a (k)1(k), then area of having no time
Interval lower bound a0With upper bound a1It is defined by the formula.
WhereinWithA is represented respectively0And a (k)1(k) linear order collection (linearly ordered set), wherein k
=1,2 ... n, i.e.,WithIn element according to value size ascending order arrangement,WithRespectivelyWithElement index,
(4) histogram analysis
For a secondary textile gray level image I, by image calibration, class grid pattern is split and calculates class grid pattern area,
With reference to the interval lower bound a of area that has no time0With upper bound a1The distribution of analysis classes grid pattern area, i.e. class grid pattern area histogram Exceptional value (outlier) often to that should have free time class grid pattern.Because a0Calculating be based on indefectible textile images,
For specific a textile images I, a0There is the true threshold value t of the flaw and flawless class grid pattern in the image with being precisely separating*May
Have differences, so the present invention uses notch value t ' and cliff of displacement value t " for t*Approximation.T is made to representTransverse axis scale, i.e.,Span, h (t) representLongitudinal axis scale, i.e. value is t's in INumber, t ' expressions a0<H (t) during t
" first breach ", t " represents a0<H (t) " first cliff of displacement " during t, it is defined as follows.
For I, any class grid pattern area is less than a0Or t*Approximation and class grid pattern area are more than a1Class grid pattern quilt
Labeled as there is free time class grid pattern, algorithm 7 is shown in the false code description of the process.
The Defect Detection recruitment evaluation of the inventive method is tested with Hong Kong University's Electrical and Electronic engineering department industrial automation
Based on the 106 width pixel sizes that room is provided is 256 × 256 24 color textile product images, there is free time textile figure to each
As generating 10 width rotation image using affine transformation with random angles, the image that there is serious artifact is then deleted, is finally given
490 class grid patterns have free time textile images and 51 not rotated textile images of having no time by what random direction was arranged,
The image of these in experiment (490+51=541 width) is converted into the gray level image of 8.541 width images include two kinds of patterns：Box-shaped
Image and star-shaped image, wherein box-shaped image include 26 it is indefectible and 251 width have flaw image；Star-shaped image include 25 width without
Flaw and 239 width have flaw image.Two kinds of patterns have the flaw image to include 5 kinds of flaw types：The broken ends of fractured bone (broken end), hole
Hole (hole), reticulate pattern (netting multiple), cord (thick bar) and stria (thin bar), every kind of flaw
The particular number of type refers to the first row of table 1 and table 2.All flaw images have size identical and class grid pattern orientation
Identical flaw reference map (ground-truth image), flaw reference map is 2 value images, wherein 1 represents flaw, 0 represents
Background.
Algorithm for comparing includes WGIS, BB, RB and ER.Index for assessment includes true positives (true
Positive, hereinafter referred to as TP), false positive (false positive, hereinafter referred to as FP), True Positive Rate (true positive
Rate, hereinafter referred to as TPR), false positive rate (false positive rate, hereinafter referred to as FPR), positive predictive value
(positive predictive value, hereinafter referred to as PPV) and negative predictive value negative predictive value,
Hereinafter referred to as NPV).
TPR, which is weighed, represents that the pixel of flaw is correctly demarcated as the ratio of flaw by algorithm in flaw reference map, FPR weighs the flaw
Represent that the pixel of background is demarcated as the flaw in the ratio of flaw, the flaw of PPV measure algorithms output by algorithmic error in defect reference map
Background proportion in flaw proportion in defect reference map, the background of NPV measure algorithms output in flaw reference map.It is right
In TPR, PPV and NPV, desired value is the bigger the better, then the smaller the better for FPR.Related mathematical definition can be in document
(M.K.Ng,H.Y.T.Ngan,X.Yuan,et al.,Patterned fabric inspection and
visualization by the method of image decomposition,IEEE
Trans.Autom.Sci.Eng.11 (3) (2014) 943-947) in find.Experimental Hardware platform is Intel containing processor
CoreTMThe notebook computer of i7-3610QM 230-GHz and 8.00GB internal memories, software is Windows 10 and Maltab8.4.
BB and RB outputs bianry image (wherein 1 represents flaw, and 0 represents background), although size is consistent with input picture, but
Elongated zones in output image close to edge are not processed, and the pixel value in these regions is set as 0.Although ER and WGIS are also defeated
Go out bianry image (wherein 1 represents flaw, and 0 represents background), but size is smaller than input picture, therefore its result arest neighbors
Size conversion is input image size by interpolation.The output image of BB, RB, ER and WGIS after treatment can pass through logic
Computing is directly compared with flaw reference map.These four algorithms, which are required, is manually entered parameter, and wherein ER and WGIS require defeated
Enter a width pattern masterplate.For WGIS, the upper left corner for the piece image that pattern masterplate is sorted by name from indefectible image is cut
Take, the pixel size of the pattern masterplate of box-shaped image is 27 × 25, the pixel size of the pattern masterplate of star-shaped image is 22 × 18.
For ER, pattern masterplate pixel size perseverance is 28 × 26, represents that the parameter of match number of times is set to 15.For RB, rule band pixel
Size perseverance is 25.For BB, the pixel size of row band (row band) and row band (column band) is respectively 15 and 25, mark
Quasi- difference amount is 2.
Because the inventive method does not export bianry image, but output indexes the defect areas that form is represented with ranks,
Therefore output result directly can not be compared with flaw reference map.For Calculation Estimation index, generation binary map is taken in evaluation process
Defect areas set is converted to bianry image (wherein 1 represents flaw, and 0 represents background) by the method for picture, and specific method is by the flaw
The copied part covered in defect reference map by defect areas is all 0 and size and flaw figure identical bianry image to a width pixel
Same position, the output that synthesized bianry image participates in assessing as the inventive method.Because the inventive method is to image
Calibrated, flaw reference map has also been carried out by respective alignment according to collimation angle, to reach flaw reference map direction of rotation
With collimation angle identical purpose.For box-shaped image, the parameter f of the inventive methodcFor 0.5, a0For 155, a1For 262, exWith
eyIt is 5；For star-shaped image, the parameter f of the inventive methodcFor 0.488, a0For 90, a1For 147, exAnd eyIt is 6.
Tables 1 and 2 shows the testing result for several algorithms for participating in assessing, and where each row (is removed and indicates " overview " most
The five-element afterwards) all represent a kind of index average value of algorithm in specific flaw species, first row represents flaw type (except indicating
The last five-element of " overview "), the numeral in its bracket represents the textile images quantity of the type, and secondary series represents algorithm mark
Know the pixel average amount for flaw, the 3rd row to the 6th row represent TPR average values, and FPR average values, PPV average values and NPV are flat
Average (average value or par are average value of the algorithm to all image detection results of specific flaw type), last row
Algorithm title is shown, the optimal value in each row is shown in overstriking font form.Tables 1 and 2 indicates the last five-element of " overview "
The index average value to box-shaped image and all textile images of star-shaped image is represented respectively.Table 1 shows the inspection of box-shaped image
Result is surveyed, the aggregate performance of the inventive method is close with WGIS, but TPR is lower than WGIS.The TPR of the inventive method is in hole type
Reach maximum.
Table 1
Table 2 enumerates the testing result of star-shaped image, and the TPR average values of the inventive method reach in all flaw types
Optimal, FPR average values are optimal in cord and the broken ends of fractured bone, and NPV average values are optimal in all flaw types.Indicating
In the last five-element of " overview ", the overall TPR average values and NPV average values of the inventive method are optimal.
Table 2
Fig. 8 and Fig. 9 are the scatter diagrams (scatter plot) for assessing each algorithm TPR and FPR being related to, and two width figures are all shown
The coordinate points that the TPR and FPR calculated according to the every width textile images testing result of different flaw types is constituted, such as box-shaped figure
The flaw type broken ends of fractured bone of picture has 49 width images, then in the legend that " the box-shaped image broken ends of fractured bone " is indicated in Fig. 8 and Fig. 9, each algorithm exists
There are 49 points by coordinate value of TPR and FPR in TPR-FPR coordinate systems, TPR the and FPR coordinates each put represent to calculate respectively
Method is to TPR and FPR desired value of the 1 width flaw type for the box-shaped image detection result of the broken ends of fractured bone.Exist in some scatter diagrams indivedual
The coordinate points of algorithm are less than the situation of image number, and this is due to that some images are not identified as the image containing flaw by the algorithm,
Therefore in the absence of corresponding desired value.TPR and FPR ideal value is 1.00 and 0.00, a left side for correspondence TPR-FPR coordinate systems respectively
Upper angle.Fig. 8 shows TPR the and FPR values of each algorithm testing result of box-shaped image, wherein the inventive method in the broken ends of fractured bone, thick bar
The Detection results of line and stria preferably, that is, correspond to the point of testing result of the present invention close to the coordinate system upper left corner, although WGIS has
Higher TPR but FPR is very high.For star-shaped image, i.e., shown in Fig. 9, the TPR-FPR points of present invention side are close to coordinate system upper left
Angle, illustrates that the inventive method has preferable star-shaped image Detection results.
Using the above-mentioned desirable embodiment according to the present invention as enlightenment, by above-mentioned description, relevant staff is complete
Various changes and amendments can be carried out without departing from the scope of the technological thought of the present invention' entirely.The technology of this invention
Property scope is not limited to the content on specification, it is necessary to its technical scope is determined according to right.
Claims (2)
1. a kind of textile flaw detection method based on class grid pattern and its area features, it is characterised in that including following step
Suddenly：
The textile images of input gray level；
Calibrate image：Ideal calibration angle is determined using Hough transform, is schemed after being calibrated with ideal calibration angle rotation image
Picture；
Class grid pattern is split：Image is to produce class grid pattern after segmentation calibration, and class grid pattern is met：Relative to textile images
Row and column, class grid pattern is transversely arranged by the direction of image line, and by the direction longitudinal arrangement of row；In anatomic element analysis method
Textile patter cartoon composition IcIn, class grid pattern has geometry and there were significant differences in gray scale with background pixel；
Areal calculation：Calculate the area of class grid pattern；
Histogram analysis：The distribution of analysis classes grid pattern area, i.e. class grid pattern area histogramIfThere is exceptional value
Then to that should have free time class grid pattern；
According to the pixel index set for thering is the output of free time class grid pattern to represent flaw.
2. the textile flaw detection method as claimed in claim 1 based on class grid pattern and its area features, its feature exists
In concretely comprising the following steps：
Calibrate image step：The edge image of input picture is generated using Canny rim detections, using Hough transform by edge
Image projection is to parameter space, the slope of straight line, the angle according to corresponding to slope corresponding to the peak value in the space that gets parms
Negative value rotates input picture and calculates the transverse projection of its background pixelAnd longitudinal projectionThe entropy of transverse projection is calculated respectivelyWith the entropy of longitudinal projectionAccording to entropy threshold exAnd eyDetermine ideal calibration angleThe angleCorresponding rotation image
Image after as calibrating；
Class grid pattern segmentation step：The cartoon composition I of image after calibration is calculated using morphology component analyzing methodc, use threshold value
fc·max(Ic) binaryzation cartoon composition IcObtain bianry image Itc, arrange I by ranks indexed sequential respectivelytcIn every row and every
The background pixel number of row, i.e., laterally and longitudinally projectWithFind out respectivelyWithPeak valueWithUsing adaptive
Kmeans algorithms pairWithClustered, screened according to cluster centreWithAccording toWithCalculate and constitute stable row
The set S of the most long continuous line index of spacinghAnd constitute the set S of the most long continuous column index of stable column pitchv, and extend
ShAnd SvI is included respectivelytcMost of line index and column index；With ShAnd SvIn ItcThe corresponding ranks of middle difference are boundary, will
ItcIt is divided into the class grid pattern of rectangleWherein, i=1,2 ... | Sh| -1, j=1,2 ... | Sv|-1；
Area calculation step：Count each class grid patternThe number a of included foreground pixeli,j, wherein, i=1,2 ... | Sh|-
1, j=1,2 ... | Sv| -1, ai,jAsArea；
Histogram analysis step：Calculate ai,jHistogramWherein, i=1,2 ... | Sh| -1, j=1,2 ... | Sv| -1, with approximateTrue threshold value is that target adjustment is had no time the interval lower bound a of area0With upper bound a1, by any not in the area interval [a that has no time0, a1]
Interior ai,jCorrespondingLabeled as there is free time class grid pattern.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710350302.5A CN107248152A (en) | 2017-05-18 | 2017-05-18 | Textile flaw detection method based on class grid pattern and its area features |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710350302.5A CN107248152A (en) | 2017-05-18 | 2017-05-18 | Textile flaw detection method based on class grid pattern and its area features |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN107248152A true CN107248152A (en) | 2017-10-13 |
Family
ID=60017533
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710350302.5A Pending CN107248152A (en) | 2017-05-18 | 2017-05-18 | Textile flaw detection method based on class grid pattern and its area features |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107248152A (en) |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107895371A (en) * | 2017-11-24 | 2018-04-10 | 常州大学 | Textile flaw detection method based on peak value coverage values and Gabor characteristic |
| CN107945164A (en) * | 2017-11-24 | 2018-04-20 | 常州大学 | Textile flaw detection method based on peak threshold, rotational alignment and composite character |
| CN107945165A (en) * | 2017-11-24 | 2018-04-20 | 常州大学 | Textile flaw detection method based on peak value coverage values and areal calculation |
| CN107967680A (en) * | 2017-11-24 | 2018-04-27 | 常州大学 | Textile flaw detection method based on peak threshold and composite character |
| CN107977961A (en) * | 2017-11-24 | 2018-05-01 | 常州大学 | Textile flaw detection method based on peak value coverage values and composite character |
| CN109518446A (en) * | 2018-12-21 | 2019-03-26 | 季华实验室 | A kind of intelligent method of cutting out of cutter |
| CN111223109A (en) * | 2020-01-03 | 2020-06-02 | 四川新网银行股份有限公司 | Complex form image analysis method |
| CN111257346A (en) * | 2020-02-20 | 2020-06-09 | 清华大学 | PCB positioning device and method based on projection filtering |
| US11788918B2 (en) | 2020-06-18 | 2023-10-17 | Trevillyan Labs, Llc | Fluid detection fabric |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101866427A (en) * | 2010-07-06 | 2010-10-20 | 西安电子科技大学 | Method for detecting and classifying fabric defects |
| CN102706881A (en) * | 2012-03-19 | 2012-10-03 | 天津工业大学 | Cloth defect detecting method based on machine vision |
| CN102879401A (en) * | 2012-09-07 | 2013-01-16 | 西安工程大学 | Method for automatically detecting and classifying textile flaws based on pattern recognition and image processing |
| CN103234976A (en) * | 2013-04-03 | 2013-08-07 | 江南大学 | Warp knitting machine cloth flaw on-line visual inspection method based on Gabor transformation |
-
2017
- 2017-05-18 CN CN201710350302.5A patent/CN107248152A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101866427A (en) * | 2010-07-06 | 2010-10-20 | 西安电子科技大学 | Method for detecting and classifying fabric defects |
| CN102706881A (en) * | 2012-03-19 | 2012-10-03 | 天津工业大学 | Cloth defect detecting method based on machine vision |
| CN102879401A (en) * | 2012-09-07 | 2013-01-16 | 西安工程大学 | Method for automatically detecting and classifying textile flaws based on pattern recognition and image processing |
| CN103234976A (en) * | 2013-04-03 | 2013-08-07 | 江南大学 | Warp knitting machine cloth flaw on-line visual inspection method based on Gabor transformation |
Non-Patent Citations (2)
| Title |
|---|
| LIANG JIA ET AL.: "Fabric defect inspection based on lattice segmentation and Gabor filtering", 《NEUROCOMPUTING》 * |
| 林如意: "基于图像处理技术的纺织品瑕疵检测方法", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107967680B (en) * | 2017-11-24 | 2019-07-09 | 常州大学 | Textile flaw detection method based on peak threshold and composite character |
| CN107977961B (en) * | 2017-11-24 | 2019-10-11 | 常州大学 | Textile flaw detection method based on peak value coverage values and composite character |
| CN107945165A (en) * | 2017-11-24 | 2018-04-20 | 常州大学 | Textile flaw detection method based on peak value coverage values and areal calculation |
| CN107967680A (en) * | 2017-11-24 | 2018-04-27 | 常州大学 | Textile flaw detection method based on peak threshold and composite character |
| CN107977961A (en) * | 2017-11-24 | 2018-05-01 | 常州大学 | Textile flaw detection method based on peak value coverage values and composite character |
| CN107895371B (en) * | 2017-11-24 | 2021-10-01 | 常州大学 | Textile flaw detection method based on peak coverage value and Gabor characteristics |
| CN107945164A (en) * | 2017-11-24 | 2018-04-20 | 常州大学 | Textile flaw detection method based on peak threshold, rotational alignment and composite character |
| CN107945164B (en) * | 2017-11-24 | 2019-07-26 | 常州大学 | Textile flaw detection method based on peak threshold, rotational alignment and composite character |
| CN107895371A (en) * | 2017-11-24 | 2018-04-10 | 常州大学 | Textile flaw detection method based on peak value coverage values and Gabor characteristic |
| CN109518446A (en) * | 2018-12-21 | 2019-03-26 | 季华实验室 | A kind of intelligent method of cutting out of cutter |
| CN111223109A (en) * | 2020-01-03 | 2020-06-02 | 四川新网银行股份有限公司 | Complex form image analysis method |
| CN111223109B (en) * | 2020-01-03 | 2023-06-06 | 四川新网银行股份有限公司 | Complex form image analysis method |
| CN111257346A (en) * | 2020-02-20 | 2020-06-09 | 清华大学 | PCB positioning device and method based on projection filtering |
| CN111257346B (en) * | 2020-02-20 | 2022-02-25 | 清华大学 | PCB positioning device and method based on projection filtering |
| US11788918B2 (en) | 2020-06-18 | 2023-10-17 | Trevillyan Labs, Llc | Fluid detection fabric |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107248152A (en) | Textile flaw detection method based on class grid pattern and its area features | |
| CN107274385A (en) | Textile flaw detection method based on class grid pattern and its Gabor characteristic | |
| CN106770323B (en) | The textile flaw detection method filtered based on hierarchical clustering and Gabor | |
| Ng et al. | Patterned fabric inspection and visualization by the method of image decomposition | |
| CN105550660B (en) | A kind of recognition methods of woven fabric structure structure species | |
| DE102016100101A1 (en) | SYSTEM FOR MACHINE SEALING AND ANALYTICAL METHOD FOR SEGMENTING PLANAR SURFACES | |
| CN104346818B (en) | A kind of threads per unit length method for automatic measurement | |
| CN107945164B (en) | Textile flaw detection method based on peak threshold, rotational alignment and composite character | |
| CN104282008B (en) | The method and apparatus that Texture Segmentation is carried out to image | |
| CN110245593A (en) | A kind of images of gestures extraction method of key frame based on image similarity | |
| CN107895371A (en) | Textile flaw detection method based on peak value coverage values and Gabor characteristic | |
| Techniková et al. | Weaving density evaluation with the aid of image analysis | |
| CN111862027B (en) | Textile flaw detection method based on low-rank sparse matrix decomposition | |
| CN115131348A (en) | Method and system for detecting textile surface defects | |
| CN115100206A (en) | Printing defect identification method for textile with periodic pattern | |
| CN114862855A (en) | Textile defect detection method and system based on template matching | |
| CN107977961B (en) | Textile flaw detection method based on peak value coverage values and composite character | |
| CN107967680B (en) | Textile flaw detection method based on peak threshold and composite character | |
| CN112488135A (en) | Method for classifying BIM three-dimensional image features of transformer substation | |
| Georgieva et al. | Identification of surface leather defects. | |
| CN113256571B (en) | Vector graphic feature point extraction method based on direction feature and local uniqueness | |
| CN114663416A (en) | Computer vision-based copper bar cold shut defect detection method and system | |
| CN107945165B (en) | Textile flaw detection method based on peak value coverage values and areal calculation | |
| Botha et al. | Image processing towards the automated identification of nanoparticles in SEM images | |
| KR20230006822A (en) | Classification of abnormal wafer images |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20171013 |
|
| RJ01 | Rejection of invention patent application after publication |