CN106296689A - Flaw detection method, system and device - Google Patents
Flaw detection method, system and device Download PDFInfo
- Publication number
- CN106296689A CN106296689A CN201610653137.6A CN201610653137A CN106296689A CN 106296689 A CN106296689 A CN 106296689A CN 201610653137 A CN201610653137 A CN 201610653137A CN 106296689 A CN106296689 A CN 106296689A
- Authority
- CN
- China
- Prior art keywords
- matrix
- textile images
- textile
- images
- block
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- 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
- 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 invention provides a kind of flaw detection method, system and device, the method comprises the following steps: input textile images to be detected;Calculate the template size of this textile images;According to described template size, textile images is split, and obtain multiple block;Calculate the correlation coefficient of each block and all blocks, and obtain a correlation matrix;And calculate reachability matrix according to described correlation matrix, and determine the flaw part in textile images according to this reachability matrix.The present invention can realize the detection of textile flaw.
Description
Technical field
The present invention designs a kind of Defect Detection technology, especially designs a kind of method for textile Defect Detection, system
And device.
Background technology
Textile Defect Detection is particularly important during textile production, the income that outstanding textile can bring,
Defect ware then easily causes economic loss.Existing textile Defect Detection technology is examined frequently with artificial perusal
Survey, or carry out the automatic detection of textile defect based on computer vision.Along with machine vision and the development of correlation technique, weaving
The Automatic Measurement Technique of product defect will be increasingly being applied in modern industry production.The automatic detection of textile defect at present
Technology is broadly divided into two big classes: one is that textile images is transformed to frequency domain, is analyzed the detection with defect the most again, its allusion quotation
Type algorithm is mainly based upon the method for wavelet transformation, but the method specific aim is stronger, it is impossible to be applied to various textile
In Defect Detection;Two is the method using template matching, but present template matching method cannot accomplish self adaptation, in addition it is also necessary to
Put into manpower and carry out template matching, and time cost is big, is not suitable for commercial production.
According to foregoing description, existing textile Defect Detection generally there is problems in that, (1) is ensureing detection
Detection efficiency is often ignored while accuracy;(2) detection method that spatial domain converts is related to frequency domain, with strong points, weaving
Category type can cause detection bigger deviation occur while changing;(3) flaw detection method based on template, for mould
Choosing of plate cannot realize self adaptation, needs to choose template size by hand, adds the consumption of manpower in commercial production, improve
Cost;(4) flaw detection method based on template, in order to reduce the subtle effects produced by yarn fabric image generation stretching etc.,
In most cases needing to train flawless yarn fabric image in a large number, the yarn fabric image for a few types is trained institute
The time consumed is the most acceptable, but, for the textile of numerous types, training is not the most feasible method.
Therefore, designing a kind of textile Defect Detection technology overcoming drawbacks described above is the problem needing solution badly.
Summary of the invention
The present invention is directed to the problem that prior art exists, its object is to provide a kind of faulty materials detection method, system and
Device, by detecting the flaw part in textile images, to realize the detection of textile flaw.
For achieving the above object, the invention provides a kind of flaw detection method, the method comprises the following steps: input is treated
The textile images of detection;Calculate the template size of this textile images;According to described template size, textile images is carried out
Segmentation, and obtain multiple block;Calculate the correlation coefficient of each block and all blocks, and obtain a correlation matrix;And
Calculate reachability matrix according to described correlation matrix, and determine the flaw part in textile images according to this reachability matrix.
As a further improvement on the present invention, described method is further comprising the steps of: carry out the textile images of input
Gray processing processes, so that the textile images of this input is converted to gray level image.
As a further improvement on the present invention, described template size uses statistical method to calculate, including following step
Rapid: the horizontal direction in textile images takes step-length c and takes step-length r in the vertical direction of textile images and enter textile images
Row even partition and edge cutting;By the segmentation image overlay of segmentation gained to form a three-dimensional matrice, ask for this three-dimensional square
Battle array horizontal direction and the mean variance of vertical direction;And obtain described mean variance in the horizontal direction and vertical direction minimum
Value, and the template size of textile images is asked for according to the minimum of this horizontal direction and vertical direction.
As a further improvement on the present invention, the template size of textile images is asked for by following steps: according to level
The value of step-length r and step-length c is traveled through to ask for the optimum of r and c as template size by the minimum of direction and vertical direction;
Or make step-length c carry out solving of r equal to the horizontal direction image length of textile images, then solve c, or make step-length r equal to spinning
The vertical direction picture traverse of fabric image carries out solving of c, then solves r, and obtains the optimum of r and c as template size.
As a further improvement on the present invention, multiple block is obtained by following steps: according to template size and textile
Length and the width of image carry out cutting to textile images;According to template size, the textile average mark after cutting is segmented into many
The block that individual size is identical;And each block is numbered.
As a further improvement on the present invention, the flaw part in textile images is determined by following steps: calculate each
The interior minima with other block correlation coefficienies of block four neighborhood, stores this calculating gained minima to minima matrix;
Median according to this minima matrix determines adaptive threshold T;According to this adaptive threshold T in described correlation matrix
Each element carry out threshold decision, and obtain the correlation matrix after threshold decision;To the phase relation after this threshold decision
Matrix number carries out logical operations to obtain reachability matrix;Realize textile images according to this reachability matrix to have no time partly and flaw portion
The classification divided;And obtain the numbering corresponding to flaw part, and determine the flaw part in textile images.
As a further improvement on the present invention, by following steps, each element in described correlation matrix is carried out
Threshold decision: the value of each element in correlation matrix is compared with described adaptive threshold T;Value at this element
During less than or equal to described adaptive threshold T, the value of this element is designated as 0;The big described adaptive threshold T of value at this element
Time, the value of this element is designated as 1;And the correlation matrix being worth to after threshold decision according to described each element.
As a further improvement on the present invention, by following steps, the correlation matrix after this threshold decision is patrolled
Collect computing: by formula Pk=logical (TCk) carry out logical operations, k=0,1,2 ..., (number of blocks of horizontal direction ×
The number of blocks of vertical direction);TC is the correlation matrix after described threshold decision, TCkFor element warp each in matrix TC
Cross k step and redirect the result obtained;Matrix TC is become logic matrix by logical computing, the numerical value being more than 0 is all remembered in matrix TC
Being 1, the numerical value equal to 0 keeps constant;And work as Pk+1=PkTime the P that tries to achievekFor described reachability matrix.
For achieving the above object, a kind of Defect Detection system, described system is used for: input textile images to be detected;
Calculate the template size of this textile images;According to described template size, textile images is split, and obtain multiple district
Block;Calculate the correlation coefficient of each block and all blocks, and obtain a correlation matrix;And according to described correlation coefficient square
Battle array calculates reachability matrix, and determines the flaw part in textile images according to this reachability matrix.
For achieving the above object, a kind of Defect Detection device, described device includes microprocessor, and this microprocessor is used for:
Input textile images to be detected;Calculate the template size of this textile images;According to described template size to textile figure
As splitting, and obtain multiple block;Calculate the correlation coefficient of each block and all blocks, and obtain a correlation coefficient square
Battle array;And calculate reachability matrix according to described correlation matrix, and determine the flaw in textile images according to this reachability matrix
Part.
Utilize flaw detection method of the present invention, system and device, textile images can be asked for template, and root
Cut textile images according to this template asked for and realize the detection of flaw part, decreasing artificial participation, improve weaving
The ability of product Defect Detection automatization, has saved cost, and detection method strong adaptability, and the time simultaneously decreasing detection is multiple
Miscellaneous degree.
Accompanying drawing explanation
Fig. 1 is the running environment figure of present pre-ferred embodiments Defect Detection system;
Fig. 2 is the FB(flow block) of present pre-ferred embodiments flaw detection method.
Fig. 3 be present pre-ferred embodiments be the textile images of cycle variation law.
Fig. 4 is the mean variance curve chart of present pre-ferred embodiments horizontal direction and vertical direction.
Fig. 5 is the textile images of gained after present pre-ferred embodiments is split.
Fig. 6 is the schematic diagram of present pre-ferred embodiments four neighborhood.
Fig. 7 is the pilot process figure of present pre-ferred embodiments Defect Detection.
Fig. 8 is the flaw schematic diagram of the different textile images that present pre-ferred embodiments is obtained.
Main element symbol description
Following detailed description of the invention will further illustrate the present invention in conjunction with above-mentioned accompanying drawing.
Detailed description of the invention
Describe the present invention below with reference to each embodiment shown in the drawings.But these embodiments are not
Limit the present invention, structure that those of ordinary skill in the art is made or conversion functionally according to these embodiments all to wrap
Containing within the scope of the present invention.
The running environment figure of present pre-ferred embodiments Defect Detection system it is shown refering to Fig. 1.This Defect Detection system
For the flaw part in textile images is detected, to realize the detection of textile flaw.
Described Defect Detection system 10 runs in calculating device 100, and this Defect Detection system 10 includes computerization journey
Sequence instructs.This calculating device 100 can be the terminal units such as computer, notebook computer or server.Described calculating device 100
Also include processor 20, memory element 30, input block 40 and display unit 50.
Described processor 20 is for performing the programmed instruction in calculating device 100.Described memory element 30 is based on storing
Calculate the data of device 100.This memory element 30 can be to be built in storage device (the such as internal memory and hard calculated in device 100
Dish etc.), it is also possible to for external storage device (such as portable hard drive etc.).This preferred embodiment is with memory element 30 for being built in
It is introduced as a example by calculating the storage device in device 100.Described Defect Detection system 10 is stored in memory element 30, and transports
Row, on processor 20, is performed its computerization programmed instruction by processor 20.
Described input block 40 can be the input equipment such as mouse, keyboard, for receiving the to be detected of operator's input
Textile images.Described display unit 50 can be the display devices such as display screen, for show textile images to be detected and
Image etc. corresponding to testing result.
Described Defect Detection system 10 is for inputting textile images to be detected.This textile images to be detected can be by
Operator inputs through input block 40, or chooses in the textile images to be detected of storage from memory element 30.
Described Defect Detection system 10 is for carrying out pretreatment operation to the textile images of input.This pretreatment operation can
Think that gray processing processes, so that the textile images of this input is converted to gray level image.Described pretreatment operation can reduce weaving
The time complexity of the calculating of product pictures subsequent and process etc..
Described Defect Detection system 10 is for calculating the template size of the textile images of preprocessed operation.At this preferably
In embodiment, described Defect Detection system 10 uses statistical method calculation template size.
Described statistical method calculation template size specifically comprise the following steps that the horizontal direction in textile images takes step-length
C and take step-length r in the vertical direction of textile images textile images is carried out even partition and edge cutting;Gained will be split
Segmentation image overlay to form a three-dimensional matrice, ask for this three-dimensional matrice horizontal direction and the mean variance of vertical direction;
Obtain described mean variance in the horizontal direction and the minimum of vertical direction, and minimum according to this horizontal direction and vertical direction
Value asks for the template size of textile images.When the integral multiple that step-length c is the horizontal direction cycle, mean variance obtains in level
The minimum in direction, when the integral multiple that step-length r is the vertical direction cycle, mean variance obtains the minimum in vertical direction.?
When asking for template size, according to the minimum of above-mentioned horizontal direction and vertical direction, the value of r and c can be traveled through to ask for
Good value is as template size;The horizontal direction image length of c=textile images can also be made, first carry out solving of r, the most again
Solve c, or make the vertical direction picture traverse of r=textile images, first carry out solving of c, then solve r, and obtain r and
The optimum of c is as template size.
Refering to shown in Fig. 3, textile images presents obvious cycle variation law.One is advised containing cyclically-varying
Textile images X of rule, image is sized to Row × Col, Row and represents textile images vertical direction picture traverse, Col table
Showing textile images horizontal direction image length, the template size of this textile images is unknown.Choosing vertical direction step-length is r,
Horizontal direction step-length is c, and according to the template of r × c size, image is carried out even partition and edge cutting, wherein vertical direction
It is divided intoBlock, horizontal direction is divided intoBlock, then obtain the segmentation image of I × J block r × c size.To obtain
Segmentation image overlay, obtain the three-dimensional matrice of a r × c × (I × J), and put it in three dimensions, this three dimensions
Initial point be set to O.In rOc plane, the variance of each point vertical direction is represented by S2(i, j), wherein 1≤i≤r, 1≤j≤c,
The formula of the mean variance calculating three-dimensional matrice horizontal direction and vertical direction is as follows:
Wherein,
When step-length that and if only if is the positive integer times in cycle, formula (1) obtains minimum.It is to say, when step-length c is level
During the integral multiple in direction cycle, mean variance obtains minimum in the horizontal direction, when the integer that step-length r is the vertical direction cycle
Times time mean variance obtain in the minimum of vertical direction.For solving of formula (1), it may be considered that the value traversal of r and c is asked
Take optimum as template size;C=Col can also be made first to carry out solving of r, solve c the most again, or make r=Row advanced
Row c solves, and then solves r.In practical operation, second method is in hgher efficiency, and speed is faster.
Refering to shown in Fig. 4, for textile images in the horizontal direction and vertical direction mean variance.Variance as can be known from Fig. 3
It is 21,42,63,84 that average obtains minimizing point in the horizontal direction;Mean variance obtains minimizing point in vertical direction
16,33,50,67,83, the template size asking for gained textile images is 17 × 21.
Described Defect Detection system 10 is for splitting textile images according to described template size, and obtains multiple
Block.In this preferred embodiment, Defect Detection system 10 according to template size and the length of textile images and width to spinning
Fabric image carries out cutting, and according to template size, the textile average mark after cutting is segmented into the block that multiple size is identical,
And each block is numbered.Such as, textile images size is Row × Col, and template size is r × c, orderAccording to template size and I, J value cutting textile images;By the textile images after cutting,
Horizontal direction is averagely split according to c pixel size, and vertical direction is averagely split according to r pixel size, dividing the image into as I ×
The block of J block formed objects, is designated as B (1,1), B (1,2) ..., B (I, J);To each block number, be designated as No.1,
No.2......The textile images of gained after splitting it is shown in present pre-ferred embodiments refering to Fig. 5.
Described Defect Detection system 10 is for calculating the correlation coefficient of each block and all blocks, and obtains a phase relation
Matrix number.In this preferred embodiment, with the block of acquisition, as B, (1,1), B (1,2) ..., B (I, J) are introduced, the described flaw
Defect detecting system 10 calculates B (1,1) and B (1,1), B (1,2) ..., B (I, J), B (1,2) and B (1,1), B (1,2) ..., B
(I, J) ..., B (I, J) and B (1,1), B (1,2) ..., the correlation coefficient of B (I, J) are also saved in a matrix.Relevant
The computational methods of coefficient such as formula (2):
In formula (2), Re represents that correlation coefficient, A, B represent that two sizes are the matrix of m × n,In representing matrix
The average of element:
After Calculation of correlation factor completes, described Defect Detection system 10 obtain the matrix of one (I × J) × (I × J) with
Preserve correlation coefficient, be designated as correlation matrix C, in C each element representation be C (x, y), wherein 1≤x≤(I × J), 1≤
y≤(I×J)。
Described Defect Detection system 10 is used for calculating reachability matrix according to described correlation matrix, and according to this up to square
Battle array determines the flaw part in textile images.Specifically comprising the following steps that of described acquisition flaw part calculates each block four neighborhood
In with the minima of other block correlation coefficienies, this calculating gained minima is stored to minima matrix, and according to this
The median of little value matrix determines adaptive threshold T;According to this adaptive threshold T to each unit in described correlation matrix
Element carries out threshold decision, and obtains the correlation matrix after threshold decision;Correlation matrix after this threshold decision is entered
Row logical operations to obtain reachability matrix, according to this reachability matrix realize textile images have no time part with flaw part point
Class;Obtain the numbering corresponding to flaw part, and determine the flaw part in textile images.
In this preferred embodiment, with the block of acquisition, as B, (1,1), B (1,2) ..., B (I, J) are introduced.Observe
Textile images after segmentation, it is found that the trickle deviation of some pixels can occur during splitting, causes
The reason of this situation is that textile both horizontally and vertically there occurs stretching, in order to reduce what end product was caused by stretching
Deviation, often considers the neighborhood impact on block.In described Defect Detection system 10 calculates each block four neighborhood and other districts
The minima of block correlation coefficient, and this calculating gained minima is stored to minima matrix R, its size is I × J, and takes
Matrix median is as adaptive threshold T.The position of four neighborhoods is refering to shown in Fig. 6.Assume current block be B (i, j), wherein 2
≤ i≤I-1,2≤j≤J-1, calculate this block B (i, j) with B (i-1, j), B (i+1, j), B (i, j-1), the phase of B (i, j+1)
Closing coefficient and be designated as Re1, Re2, Re3, Re4 respectively, (i, j)=min{Re1, Re2, Re3, Re4}, computation rule needs basis to take R
The change of i, j changes, and has calculated and has obtained a minima matrix R afterwards, and size is I × J, takes matrix median conduct
Adaptive threshold T.
Each element in described correlation matrix is entered by described Defect Detection system 10 according to described adaptive threshold T
Row threshold decision.The step of described threshold decision is as follows: for element each in correlation matrix C, if C (then will by x, y)≤T
This element is designated as 0, otherwise then this element is designated as 1, it is thus achieved that the correlation matrix TC after threshold decision.This matrix TC is to relevant
Coefficient matrix C is the same, is all the matrix of (I × J) × (I × J) size, and this matrix TC is the matrix of only 0 and 1 element,
And diagonal entry is 1.
Described Defect Detection system 10 according to the correlation matrix after threshold decision carry out logical operations with obtain up to
Matrix, and realize textile images according to this reachability matrix and have no time part and the classification of flaw part.The calculating side of reachability matrix
Method is as follows: the correlation matrix TC after threshold decision carries out logical operations:
Pk=logical (TCk) (3),
K=0,1,2 ..., (I × J), work as Pk+1=PkTime terminate, PkIt is required reachability matrix.(3) in formula
Logical computing represents and matrix TC is become logic matrix, numerical value more than 0 will all be designated as 1 in matrix TC, the number equal to 0
Value keeps constant;TCkIn representing matrix TC, each element redirects, through k step, the result obtained.The indefectible part when k value becomes big
Value can be fixed as 1, the value of flaw part can be fixed as 0, thus realizes the classification of flawless part and flaw part.
Showing the pilot process figure for Fig. 5 detection refering to Fig. 7, what the intermediate image of Fig. 7 represented is by matrix TC
The reachability matrix asked for, have no time number of times is higher when part and the effect of flaw part classifying, the linear part of black represents
Cannot be introduced into the part of apoplexy due to endogenous wind, namely flaw part.It is corresponding that what the right image of Fig. 7 represented is horizontally oriented black part
Numbering, the numbering that namely flaw part place block is corresponding, the position of flaw part is i.e. can determine that according to this numbering.Inhomogeneity
The Defect Detection design sketch of type textile is refering to shown in Fig. 8, and this Defect Detection design sketch is shown on reality unit 50.
The FB(flow block) of present pre-ferred embodiments flaw detection method it is shown refering to Fig. 2.
Step S10, described Defect Detection system 10 inputs textile images to be detected, and the textile images to input
Carry out pretreatment operation.This pretreatment operation can be that gray processing processes, so that the textile images of this input is converted to gray scale
Image.
Step S20, described Defect Detection system 10 calculates the template size of the textile images of preprocessed operation.At this
In preferred embodiment, described Defect Detection system 10 uses statistical method calculation template size.
Described statistical method calculation template size specifically comprise the following steps that the horizontal direction in textile images takes step-length
C and take step-length r in the vertical direction of textile images textile images is carried out even partition and edge cutting;Gained will be split
Segmentation image overlay to form a three-dimensional matrice, ask for this three-dimensional matrice horizontal direction and the mean variance of vertical direction;
Obtain described mean variance in the horizontal direction and the minimum of vertical direction, and minimum according to this horizontal direction and vertical direction
Value asks for the template size of textile images.When the integral multiple that step-length c is the horizontal direction cycle, mean variance obtains in level
The minimum in direction, when the integral multiple that step-length r is the vertical direction cycle, mean variance obtains the minimum in vertical direction.?
When asking for template size, according to the minimum of above-mentioned horizontal direction and vertical direction, the value of r and c can be traveled through to ask for
Good value is as template size;The horizontal direction image length of c=textile images can also be made, first carry out solving of r, the most again
Solve c, or make the vertical direction picture traverse of r=textile images, first carry out solving of c, then solve r, and obtain r and
The optimum of c is as template size.
Step S30, textile images is split according to described template size, and is obtained by described Defect Detection system 10
Multiple blocks.In this preferred embodiment, Defect Detection system 10 is according to template size and the length of textile images and width
Textile images is carried out cutting, and according to template size, the textile average mark after cutting is segmented into the district that multiple size is identical
Block, and each block is numbered.
Step S40, described Defect Detection system 10 calculates the correlation coefficient of each block and all blocks, and obtains a phase
Close coefficient matrix.
Step S50, described Defect Detection system 10 calculates reachability matrix according to described correlation matrix, and can according to this
Reach the flaw part that matrix determines in textile images.Specifically comprising the following steps that of described acquisition flaw part calculates each block four
The interior minima with other block correlation coefficienies of neighborhood, stores this calculating gained minima to minima matrix, and according to
The median of this minima matrix determines adaptive threshold T;According to every in described correlation matrix of this adaptive threshold T
Individual element carries out threshold decision, and obtains the correlation matrix after threshold decision;To the correlation coefficient square after this threshold decision
Battle array carries out logical operations to obtain reachability matrix, realizes textile images according to this reachability matrix and has no time partly and flaw part
Classification;Obtain the numbering corresponding to flaw part, and determine the flaw part in textile images.
It is to be understood that, although this specification is been described by according to embodiment, but the most each embodiment only comprises one
Individual independent technical scheme, this narrating mode of description is only that for clarity sake those skilled in the art should will say
Bright book is as an entirety, and the technical scheme in each embodiment can also be through appropriately combined, and forming those skilled in the art can
With other embodiments understood.
The a series of detailed description of those listed above is only for the feasibility embodiment of the present invention specifically
Bright, they also are not used to limit the scope of the invention, all equivalent implementations made without departing from skill of the present invention spirit
Or change should be included within the scope of the present invention.
Claims (10)
1. a flaw detection method, it is characterised in that the method comprises the following steps:
Input textile images to be detected;
Calculate the template size of this textile images;
According to described template size, textile images is split, and obtain multiple block;
Calculate the correlation coefficient of each block and all blocks, and obtain a correlation matrix;And
Calculate reachability matrix according to described correlation matrix, and determine the flaw portion in textile images according to this reachability matrix
Point.
Flaw detection method the most according to claim 1, it is characterised in that described method is further comprising the steps of:
The textile images of input is carried out gray processing process, so that the textile images of this input is converted to gray level image.
Flaw detection method the most according to claim 1, it is characterised in that described template size uses statistical method to enter
Row calculates, and comprises the following steps:
Horizontal direction in textile images takes step-length c and takes step-length r to textile images in the vertical direction of textile images
Carry out even partition and edge cutting;
By the segmentation image overlay of segmentation gained to form a three-dimensional matrice, ask for this three-dimensional matrice horizontal direction and Vertical Square
To mean variance;And
Obtain described mean variance in the horizontal direction and the minimum of vertical direction, and according to this horizontal direction and vertical direction
Minimum asks for the template size of textile images.
Flaw detection method the most according to claim 3, it is characterised in that ask for textile images by following steps
Template size:
The value of step-length r and step-length c is traveled through with the optimum asking for r and c by the minimum according to horizontal direction and vertical direction
As template size;Or
Make step-length c carry out solving of r equal to the horizontal direction image length of textile images, then solve c, or make step-length r be equal to
The vertical direction picture traverse of textile images carries out solving of c, then solves r, and the optimum obtaining r and c is big as template
Little.
Flaw detection method the most according to claim 1, it is characterised in that obtain multiple block by following steps:
According to template size and the length of textile images and width, textile images is carried out cutting;
According to template size, the textile average mark after cutting is segmented into the block that multiple size is identical;And
Each block is numbered.
Flaw detection method the most according to claim 1, it is characterised in that determined in textile images by following steps
Flaw part:
Calculate the interior minima with other block correlation coefficienies of each block four neighborhood, this calculating gained minima is stored to minimum
In value matrix;
Median according to this minima matrix determines adaptive threshold T;
According to this adaptive threshold T each element in described correlation matrix carried out threshold decision, and obtain threshold value and sentence
The correlation matrix having no progeny;
Correlation matrix after this threshold decision is carried out logical operations to obtain reachability matrix;
Realize textile images according to this reachability matrix to have no time part and the classification of flaw part;And
Obtain the numbering corresponding to flaw part, and determine the flaw part in textile images.
Flaw detection method the most according to claim 6, it is characterised in that by following steps to described correlation coefficient square
Each element in Zhen carries out threshold decision:
The value of each element in correlation matrix is compared with described adaptive threshold T;
When the value of this element is less than or equal to described adaptive threshold T, the value of this element is designated as 0;
During the big described adaptive threshold T of value at this element, the value of this element is designated as 1;And
The correlation matrix being worth to after threshold decision according to described each element.
Flaw detection method the most according to claim 6, it is characterised in that by following steps to this threshold decision after
Correlation matrix carries out logical operations:
By formula Pk=logical (TCk) carry out logical operations, k=0,1,2 ..., (the number of blocks of horizontal direction × hang down
Nogata to number of blocks);
TC is the correlation matrix after described threshold decision, TCkRedirect, through k step, the knot obtained for element each in matrix TC
Really;
Matrix TC is become logic matrix by logical computing, the numerical value being more than 0 is all designated as 1, the numerical value equal to 0 in matrix TC
Keep constant;And
Work as Pk+1=PkTime the P that tries to achievekFor described reachability matrix.
9. a Defect Detection system, it is characterised in that described system is used for:
Input textile images to be detected;
Calculate the template size of this textile images;
According to described template size, textile images is split, and obtain multiple block;
Calculate the correlation coefficient of each block and all blocks, and obtain a correlation matrix;And
Calculate reachability matrix according to described correlation matrix, and determine the flaw portion in textile images according to this reachability matrix
Point.
10. a Defect Detection device, it is characterised in that described device includes microprocessor, this microprocessor is used for:
Input textile images to be detected;
Calculate the template size of this textile images;
According to described template size, textile images is split, and obtain multiple block;
Calculate the correlation coefficient of each block and all blocks, and obtain a correlation matrix;And
Calculate reachability matrix according to described correlation matrix, and determine the flaw portion in textile images according to this reachability matrix
Point.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610653137.6A CN106296689B (en) | 2016-08-10 | 2016-08-10 | Flaw detection method, system and device |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610653137.6A CN106296689B (en) | 2016-08-10 | 2016-08-10 | Flaw detection method, system and device |
Publications (2)
Publication Number | Publication Date |
---|---|
CN106296689A true CN106296689A (en) | 2017-01-04 |
CN106296689B CN106296689B (en) | 2019-03-22 |
Family
ID=57668054
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201610653137.6A Expired - Fee Related CN106296689B (en) | 2016-08-10 | 2016-08-10 | Flaw detection method, system and device |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106296689B (en) |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107895363A (en) * | 2017-10-31 | 2018-04-10 | 常州大学 | Textile flaw detection method based on template characteristic |
CN107966444A (en) * | 2017-10-12 | 2018-04-27 | 常州信息职业技术学院 | Textile flaw detection method based on template |
CN114972359A (en) * | 2022-08-03 | 2022-08-30 | 江苏万喜登家居科技有限公司 | Mesh fabric defect rapid detection method based on symmetry analysis |
CN115201201A (en) * | 2022-06-24 | 2022-10-18 | 广东工业大学 | Forming mesh surface quality detection method based on machine vision |
CN116385375A (en) * | 2023-03-17 | 2023-07-04 | 银河航天(北京)网络技术有限公司 | Forest defect area detection method and device based on remote sensing image and storage medium |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101216435A (en) * | 2008-01-03 | 2008-07-09 | 东华大学 | Fabric flaw automatic detection method based on multi-fractal characteristic parameter |
CN102221559A (en) * | 2011-03-05 | 2011-10-19 | 河海大学常州校区 | Online automatic detection method of fabric defects based on machine vision and device thereof |
-
2016
- 2016-08-10 CN CN201610653137.6A patent/CN106296689B/en not_active Expired - Fee Related
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101216435A (en) * | 2008-01-03 | 2008-07-09 | 东华大学 | Fabric flaw automatic detection method based on multi-fractal characteristic parameter |
CN102221559A (en) * | 2011-03-05 | 2011-10-19 | 河海大学常州校区 | Online automatic detection method of fabric defects based on machine vision and device thereof |
Non-Patent Citations (2)
Title |
---|
COLIN S.C.TSANG等: "Fabric inspection based on the Elo rating method", 《PATTERN RECOGNITION》 * |
杨晓波: "基于GMRF模型的统计特征畸变织物疵点识别", 《纺织学报》 * |
Cited By (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107966444A (en) * | 2017-10-12 | 2018-04-27 | 常州信息职业技术学院 | Textile flaw detection method based on template |
CN107895363A (en) * | 2017-10-31 | 2018-04-10 | 常州大学 | Textile flaw detection method based on template characteristic |
CN115201201A (en) * | 2022-06-24 | 2022-10-18 | 广东工业大学 | Forming mesh surface quality detection method based on machine vision |
CN114972359A (en) * | 2022-08-03 | 2022-08-30 | 江苏万喜登家居科技有限公司 | Mesh fabric defect rapid detection method based on symmetry analysis |
CN114972359B (en) * | 2022-08-03 | 2022-10-21 | 江苏万喜登家居科技有限公司 | Mesh fabric defect rapid detection method based on symmetry analysis |
CN116385375A (en) * | 2023-03-17 | 2023-07-04 | 银河航天(北京)网络技术有限公司 | Forest defect area detection method and device based on remote sensing image and storage medium |
CN116385375B (en) * | 2023-03-17 | 2023-10-20 | 银河航天(北京)网络技术有限公司 | Forest defect area detection method and device based on remote sensing image and storage medium |
Also Published As
Publication number | Publication date |
---|---|
CN106296689B (en) | 2019-03-22 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106296689A (en) | Flaw detection method, system and device | |
CN108230359B (en) | Object detection method and apparatus, training method, electronic device, program, and medium | |
Çelik et al. | Development of a machine vision system: real-time fabric defect detection and classification with neural networks | |
CN107870172A (en) | A kind of Fabric Defects Inspection detection method based on image procossing | |
Ghosh et al. | Pattern classification of fabric defects using support vector machines | |
CN107561738A (en) | TFT LCD surface defect quick determination methods based on FCN | |
CN109598287A (en) | The apparent flaws detection method that confrontation network sample generates is generated based on depth convolution | |
Meng et al. | A multi-task and multi-scale convolutional neural network for automatic recognition of woven fabric pattern | |
CN104268505A (en) | Automatic cloth defect point detection and recognition device and method based on machine vision | |
CN108960255A (en) | Conspicuousness fabric defect detection method based on color similarity and position aggregation | |
CN104063713B (en) | A kind of semi-autonomous on-line study method based on random fern grader | |
CN110889838A (en) | Fabric defect detection method and device | |
CN114842019B (en) | Battery plate surface defect detection method, system, storage medium and equipment | |
CN110197208A (en) | A kind of textile flaw intelligent measurement classification method and device | |
Zhou et al. | Exploring faster RCNN for fabric defect detection | |
CN114372968B (en) | Defect detection method combining attention mechanism and adaptive memory fusion network | |
US10727025B2 (en) | System and method of analyzing a crystal defect | |
US20220076404A1 (en) | Defect management apparatus, method and non-transitory computer readable medium | |
Shi et al. | Fabric defect detection using local contrast deviations | |
CN105184777A (en) | Painted design fabric defect detection method based on image decomposition | |
CN113591948A (en) | Defect pattern recognition method and device, electronic equipment and storage medium | |
CN107169498A (en) | It is a kind of to merge local and global sparse image significance detection method | |
CN107945164A (en) | Textile flaw detection method based on peak threshold, rotational alignment and composite character | |
Bumrungkun | Defect detection in textile fabrics with snake active contour and support vector machines | |
Rippel et al. | Anomaly Detection for Automated Visual Inspection: A Review |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20190322 Termination date: 20190810 |
|
CF01 | Termination of patent right due to non-payment of annual fee |