CN108133472A - Textile fabric defect inspection method and textile fabric defect detecting device - Google Patents
Textile fabric defect inspection method and textile fabric defect detecting device Download PDFInfo
- Publication number
- CN108133472A CN108133472A CN201711217285.4A CN201711217285A CN108133472A CN 108133472 A CN108133472 A CN 108133472A CN 201711217285 A CN201711217285 A CN 201711217285A CN 108133472 A CN108133472 A CN 108133472A
- Authority
- CN
- China
- Prior art keywords
- sample
- offset
- image
- texture
- module
- 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
- G06T7/001—Industrial image inspection using an image reference approach
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Quality & Reliability (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Investigating Materials By The Use Of Optical Means Adapted For Particular Applications (AREA)
- Image Processing (AREA)
Abstract
The invention discloses a kind of textile fabric defect inspection methods, using various types of defective parts of cmos camera acquisition textile and non-defective part and the image information for acquiring sample to be tested, respectively obtain defect sample and zero defect sample, and stored；Texture enhancing is carried out to texture image using non-local mean filtering；Calculate the texture contour images for obtaining textile images；Construct the characteristic model of texture contour images；Characteristic model match comparing with zero defect sample, when the difference of offset and mean deviation amount is more than predetermined threshold value, it is fault model to determine this feature model；When the difference of offset and mean deviation amount is less than predetermined threshold value, it is qualified model to determine this feature model；Fault model and defect sample are carried out one-to-one match comparison, and to flaw sample progress classification storage.The technical program realizes digitlization defects detection, while has also carried out the classification storage of defect, convenient for the data analysis of quality testing personnel.
Description
Technical field
The present invention relates to textile technology field, more particularly to a kind of textile fabric defect inspection method and textile fabric defects detection
Device.
Background technology
Textile in process of production and enter market before have to pass through defect detection, density, yarn count, grammes per square metre,
The inspection and test of yarn twist, yarn strength, fabric construction, fabric thickness, loop length, fabric cover etc.,
Middle defect detection is part the most main.The detection of fabric defects is main or is completed by artificial vision's offline inspection, should
Method is there are detection speed is low, perching result is larger by perching personnel's subjective impact, the shortcomings of false drop rate and high omission factor.
Invention content
The technical problem to be solved in the present invention is to provide a kind of quick fault is carried out using computer image processing technology
The textile fabric defect inspection method of detection and textile fabric defect detecting device.
In order to solve the above-mentioned technical problem, the technical scheme is that：
A kind of textile fabric defect inspection method, including step：
Various types of defect parts and non-defective part using cmos camera acquisition textile, respectively obtain defect
Sample and zero defect sample, and stored；
Using the image information of cmos camera acquisition sample to be tested, and carry out image preprocessing；
Texture enhancing is carried out to texture image using non-local mean filtering；To texture, enhanced textile images carry out
Processing calculates the texture contour images for obtaining textile images；Construct the characteristic model of texture contour images；
The characteristic model match comparing with the zero defect sample, the difference of offset and mean deviation amount is big
When predetermined threshold value, it is fault model to determine this feature model；When the difference of offset and mean deviation amount is less than predetermined threshold value,
It is qualified model to determine this feature model；
The fault model and defect sample are carried out one-to-one match comparison, and to flaw sample progress classification storage.
Preferably：The process for calculating the texture contour images for obtaining textile images includes step：
First with without sampling wavelet transformation, W (x, y) is obtained；Select the wavelet transformation of i-th of subband；
Each subband wavelet coefficient, prominent each subband texture contour feature are pre-processed using the smooth method of local Gaussian；
Each subband texture profile is merged using multichannel gradient information, obtains the texture contour images of textile images.
Preferably：The step of image information of sample to be tested is acquired using cmos camera, and carries out image preprocessing packet
It includes：
Utilize the image information of cmos camera acquisition sample to be tested；
Image cutting, label processing are carried out to described image information；
Dry processing is carried out to described image information.
Preferably, the characteristic model carries out matching the process packet compared with the zero defect sample or the defect sample
It includes：
The pattern of the characteristic model is matched with the pattern in the zero defect sample；
Searching point is calculated according to matching result and corresponds to zero defect sample distance between Searching point, is denoted as offset；
When the difference of the offset and mean deviation amount is more than default bias amount threshold value, it is the flaw to determine described search point
Fault；Average value of the mean deviation amount for the offset summation of all Searching points in the pattern of the characteristic model；
The offset is compared with mean deviation amount.
Preferably, it is described to match the pattern to be measured with the die plate pattern in the zero defect sample, it is specific to wrap
It includes：Searching point P1 is chosen in the new search point set formed in the pattern to be measured, and is selected in the template image
The corresponding points P2 of the P1 is taken, rotational offset Δ θ, translational offsets amount R are calculated according to selected P1 and P2 and is stretched along Y-axis
Measure K；Described search spot projection is treated into mapping to described according to rotational offset Δ θ, translational offsets amount R and along Y-axis amount of tension K
In the identical template image of case, matching result is obtained；Specifically, according to P1 and P2 computational lengths D and with horizontal direction angle theta；
D=(P1x-P2x) 2+ (P1y-P2y) 2 θ=arctanP1y-P2yP1x-P2x
Three groups of corresponding points P11, P21, P12, P22, P13 and P23 are chosen in the pattern to be measured and template image,
Then translational offsets amount R is：Ry=P11y-P21y Rx=P11x-P21x
It is along Y-axis amount of tension K：K=P22y-P21yl1cos (arctan (P12x-P11xP12y-P11y)-Δ θ) K=
P23y-P21yl2cos(arctan(P13x-P11xP13y-P11y)-Δθ)
Wherein, l1 be two point distance of P11 and P12, l2 P13, P23 twos' point distance；
L1=(P11x-P12x) 2+ (P11y-P12y) 2l2=(P11x-P13x) 2+ (P11y-P13y) 2
Then the Δ θ is：Δ θ=arctan (P22y-P21y) l2cos (θ 2)-(P23y-P21y) l1cos (θ 1)
(P23y-P21y)l1sin(θ1)-(P22y-P21y)l1sin(θ2)
Wherein, θ 1=arctan (P12x-P11xP12y-P11y) θ 2=arctan (P13x-P11xP13y-P11y).
The present invention also proposes a kind of textile fabric defect detecting device, including
Sample cmos camera：Various types of defect parts and non-defective portion using cmos camera acquisition textile
Point, defect sample and zero defect sample are respectively obtained, and stored, and acquire the image information of sample to be tested；
Image pre-processing module：Image preprocessing is carried out to the image information of the sample to be tested；
Image enhancement modeling module：Texture enhancing is carried out to texture image using non-local mean filtering, texture is enhanced
Textile images afterwards are handled, and calculate the texture contour images for obtaining textile images, construct the spy of texture contour images
Levy model；
Sample comparing module：The characteristic model match comparing with the zero defect sample, offset is with being averaged
When the difference of offset is more than predetermined threshold value, it is fault model to determine this feature model；The difference of offset and mean deviation amount
During less than predetermined threshold value, it is qualified model to determine this feature model；
Defect classification module：The fault model and defect sample are carried out one-to-one to match comparison, and to flaw sample
Carry out classification storage.
Preferably：Described image enhancing modeling module includes：
Wavelet transformation module：Using without sampling wavelet transformation, W (x, y) is obtained；Select the wavelet transformation of i-th of subband；
Wavelet coefficients preprocessing module：Each subband wavelet coefficient is pre-processed using the smooth method of local Gaussian, it is prominent each
Subband texture contour feature；
Grain table module：Each subband texture profile is merged using multichannel gradient information, obtains the line of textile images
Manage contour images.
Preferably：Described image preprocessing module includes：
Acquisition module：Utilize the image information of cmos camera acquisition sample to be tested；
Cutting module：Image cutting, label processing are carried out to described image information；
Go dry module：Dry processing is carried out to described image information.
Preferably：The sample comparing module includes：
Matching module：The pattern of the characteristic model is matched with the die plate pattern in the zero defect sample；
Offset computing module：Searching point is calculated according to matching result and corresponds to die plate pattern distance between Searching point,
It is denoted as offset；
Offset threshold calculation module：When the difference of the offset and mean deviation amount is more than default bias amount threshold value
When, it is flaw point to determine described search point；The mean deviation amount for the characteristic model pattern in all Searching points it is inclined
The average value of shifting amount summation；
Contrast module：The offset is compared with mean deviation amount.
Preferably：It is described to match the pattern to be measured with the pattern in the zero defect sample, it specifically includes：
Searching point P1 is chosen in the new search point set formed in the pattern to be measured, and in the template image described in selection
The corresponding points P2 of P1 calculates rotational offset Δ θ, translational offsets amount R and along Y-axis amount of tension K according to selected P1 and P2；Root
According to rotational offset Δ θ, translational offsets amount R and along Y-axis amount of tension K by described search spot projection to identical with the pattern to be measured
Template image in, obtain matching result；Specifically, according to P1 and P2 computational lengths D and with horizontal direction angle theta；
D=(P1x-P2x) 2+ (P1y-P2y) 2 θ=arctanP1y-P2yP1x-P2x
Three groups of corresponding points P11, P21, P12, P22, P13 and P23 are chosen in the pattern to be measured and template image,
Then translational offsets amount R is：Ry=P11y-P21y Rx=P11x-P21x
It is along Y-axis amount of tension K：K=P22y-P21yl1cos (arctan (P12x-P11xP12y-P11y)-Δ θ) K=
P23y-P21yl2cos(arctan(P13x-P11xP13y-P11y)-Δθ)
Wherein, l1 be two point distance of P11 and P12, l2 P13, P23 twos' point distance；
L1=(P11x-P12x) 2+ (P11y-P12y) 2l2=(P11x-P13x) 2+ (P11y-P13y) 2
Then the Δ θ is：Δ θ=arctan (P22y-P21y) l2cos (θ 2)-(P23y-P21y) l1cos (θ 1)
(P23y-P21y)l1sin(θ1)-(P22y-P21y)l1sin(θ2)
Wherein, θ 1=arctan (P12x-P11xP12y-P11y) θ 2=arctan (P13x-P11xP13y-P11y).
Using above-mentioned technical proposal, as a result of image recognition technology, increased by the texture that image is carried out to textile fabric
By force, defect sample and zero defect sample are respectively constituted, by the way that the sample image of detection is compared with two kinds of models, to sample
Product image is classified, difference model whether be qualified model, and to confirm the defects of version.The technical program can be fast
The quality of speed detection textile fabric, while the technical staff for being additionally favorable for production establishes the databases of faulty materials, convenient for data analysis, with
Improving The Quality of Products.
Description of the drawings
Fig. 1 is the flow chart of textile fabric defect inspection method of the present invention；
Fig. 2 is the flow chart of step S30 in Fig. 1；
Fig. 3 is the flow chart of step S20 in Fig. 1；
Fig. 4 is the flow chart of step S40 in Fig. 1；
Fig. 5 is the schematic diagram of textile fabric defect detecting device of the present invention；
Fig. 6 is the schematic diagram of 300 modules in Fig. 5；
Fig. 7 is the schematic diagram of 200 modules in Fig. 5；
Fig. 8 is the schematic diagram of 400 modules in Fig. 5.
In figure, 100- sample cmos cameras, 200- image pre-processing modules, 300- image enhancement modeling modules, 400-
Sample comparing module, 500- defect classification modules, 210- acquisition modules, 220- cutting modules, 230- remove dry module, 310- small echos
Conversion module, 320- wavelet coefficients preprocessing modules, 330- grain table modules, 410- matching modules, 420- offsets calculate
Module, 430- offset threshold calculation modules, 440- contrast modules.
Specific embodiment
The specific embodiment of the present invention is described further below in conjunction with the accompanying drawings.It should be noted that for
The explanation of these embodiments is used to help understand the present invention, but do not form limitation of the invention.It is in addition, disclosed below
The each embodiment of the present invention in involved technical characteristic can be combined with each other as long as they do not conflict with each other.
With reference to Fig. 1, the present invention proposes a kind of textile fabric defect inspection method, including step：
S10：Various types of defect parts and non-defective part using cmos camera acquisition textile, respectively obtain
Defect sample and zero defect sample, and stored；
S20：Using the image information of cmos camera acquisition sample to be tested, and carry out image preprocessing；
S30：:Texture enhancing is carried out to texture image using non-local mean filtering；The enhanced textile figure to texture
As being handled, the texture contour images for obtaining textile images are calculated；Construct the characteristic model of texture contour images；
S40：The characteristic model match comparing with the zero defect sample, the difference of offset and mean deviation amount
When value is more than predetermined threshold value, it is fault model to determine this feature model；The difference of offset and mean deviation amount is less than default threshold
During value, it is qualified model to determine this feature model；
S50：The fault model and defect sample are carried out one-to-one match comparison, and flaw sample classify and is deposited
Storage.
Using above-mentioned technical proposal, as a result of image recognition technology, increased by the texture that image is carried out to textile fabric
By force, defect sample and zero defect sample are respectively constituted, by the way that the sample image of detection is compared with two kinds of models, to sample
Product image is classified, difference model whether be qualified model, and to confirm the defects of version.The technical program can be fast
The quality of speed detection textile fabric, while the technical staff for being additionally favorable for production establishes the databases of faulty materials, convenient for data analysis, with
Improving The Quality of Products.
Reference Fig. 2, specifically：The process for calculating the texture contour images for obtaining textile images includes step：
S31：First with without sampling wavelet transformation, W (x, y) is obtained；Select the wavelet transformation of i-th of subband；
S32：Each subband wavelet coefficient is pre-processed using the smooth method of local Gaussian, prominent each subband texture profile is special
Sign；
S33：Each subband texture profile is merged using multichannel gradient information, obtains the texture profile diagram of textile images
Picture.
Reference Fig. 3, specifically：Using the image information of cmos camera acquisition sample to be tested, and carry out image preprocessing
The step of include：
S21：Utilize the image information of cmos camera acquisition sample to be tested；
S22：Image cutting, label processing are carried out to described image information；
S23：Dry processing is carried out to described image information.
With reference to Fig. 4, specifically, the characteristic model match comparing with the zero defect sample or the defect sample
Process include：
S41：The pattern of the characteristic model is matched with the pattern in the zero defect sample；
S42 corresponds to the distance between Searching point according to matching result calculating Searching point with zero defect sample, is denoted as offset；
S43：When the difference of the offset and mean deviation amount is more than default bias amount threshold value, described search point is determined
For flaw point；Average value of the mean deviation amount for the offset summation of all Searching points in the pattern of the characteristic model；
S44：The offset is compared with mean deviation amount.
Specifically, it is described to match the pattern to be measured with the die plate pattern in the zero defect sample, it is specific to wrap
It includes：Searching point P1 is chosen in the new search point set formed in the pattern to be measured, and is selected in the template image
The corresponding points P2 of the P1 is taken, rotational offset Δ θ, translational offsets amount R are calculated according to selected P1 and P2 and is stretched along Y-axis
Measure K；Described search spot projection is treated into mapping to described according to rotational offset Δ θ, translational offsets amount R and along Y-axis amount of tension K
In the identical template image of case, matching result is obtained；Specifically, according to P1 and P2 computational lengths D and with horizontal direction angle theta；
D=(P1x-P2x) 2+ (P1y-P2y) 2 θ=arctanP1y-P2yP1x-P2x
Three groups of corresponding points P11, P21, P12, P22, P13 and P23 are chosen in the pattern to be measured and template image,
Then translational offsets amount R is：Ry=P11y-P21y Rx=P11x-P21x
It is along Y-axis amount of tension K：K=P22y-P21yl1cos (arctan (P12x-P11xP12y-P11y)-Δ θ) K=
P23y-P21yl2cos(arctan(P13x-P11xP13y-P11y)-Δθ)
Wherein, l1 be two point distance of P11 and P12, l2 P13, P23 twos' point distance；
L1=(P11x-P12x) 2+ (P11y-P12y) 2l2=(P11x-P13x) 2+ (P11y-P13y) 2
Then the Δ θ is：Δ θ=arctan (P22y-P21y) l2cos (θ 2)-(P23y-P21y) l1cos (θ 1)
(P23y-P21y)l1sin(θ1)-(P22y-P21y)l1sin(θ2)
Wherein, θ 1=arctan (P12x-P11xP12y-P11y) θ 2=arctan (P13x-P11xP13y-P11y).
With reference to Fig. 5, the present invention also proposes a kind of textile fabric defect detecting device, including
Sample cmos camera：Various types of defect parts and non-defective portion using cmos camera acquisition textile
Point, defect sample and zero defect sample are respectively obtained, and stored, and acquire the image information of sample to be tested；
Image pre-processing module 200：Image preprocessing is carried out to the image information of the sample to be tested；
Image enhancement modeling module 300：Texture enhancing is carried out to texture image using non-local mean filtering, texture is increased
Textile images after strong are handled, and calculate the texture contour images for obtaining textile images, construction texture contour images
Characteristic model；
Sample comparing module：The characteristic model match comparing with the zero defect sample, offset is with being averaged
When the difference of offset is more than predetermined threshold value, it is fault model to determine this feature model；The difference of offset and mean deviation amount
During less than predetermined threshold value, it is qualified model to determine this feature model；
Defect classification module 500：The fault model and defect sample are carried out one-to-one to match comparison, and to flaw sample
This progress classification storage.
Reference Fig. 6, specifically：Described image enhancing modeling module 300 includes：
Wavelet transformation module 310：Using without sampling wavelet transformation, W (x, y) is obtained；The small echo of i-th of subband is selected to become
It changes；
Wavelet coefficients preprocessing module 320：Each subband wavelet coefficient is pre-processed using the smooth method of local Gaussian, it is prominent
Each subband texture contour feature；
Grain table module 330：Each subband texture profile is merged using multichannel gradient information, obtains textile images
Texture contour images.
Reference Fig. 7, specifically：Described image preprocessing module 200 includes：
Acquisition module 210：Utilize the image information of cmos camera acquisition sample to be tested；
Cutting module 220：Image cutting, label processing are carried out to described image information；
Go dry module 230：Dry processing is carried out to described image information.
Reference 8, specifically：The sample comparing module includes：
Matching module 410：The pattern of the characteristic model is matched with the die plate pattern in the zero defect sample；
Offset computing module 420：According to matching result calculate Searching point correspond to die plate pattern between Searching point away from
From being denoted as offset；
Offset threshold calculation module 430：When the difference of the offset and mean deviation amount is more than default bias amount threshold
During value, it is flaw point to determine described search point；The mean deviation amount is all Searching points in the pattern of the characteristic model
The average value of offset summation；
Contrast module 440：The offset is compared with mean deviation amount.
Specifically：It is described to match the pattern to be measured with the pattern in the zero defect sample, it specifically includes：
Searching point P1 is chosen in the new search point set formed in the pattern to be measured, and in the template image described in selection
The corresponding points P2 of P1 calculates rotational offset Δ θ, translational offsets amount R and along Y-axis amount of tension K according to selected P1 and P2；Root
According to rotational offset Δ θ, translational offsets amount R and along Y-axis amount of tension K by described search spot projection to identical with the pattern to be measured
Template image in, obtain matching result；Specifically, according to P1 and P2 computational lengths D and with horizontal direction angle theta；
D=(P1x-P2x) 2+ (P1y-P2y) 2 θ=arctanP1y-P2yP1x-P2x
Three groups of corresponding points P11, P21, P12, P22, P13 and P23 are chosen in the pattern to be measured and template image,
Then translational offsets amount R is：Ry=P11y-P21y Rx=P11x-P21x
It is along Y-axis amount of tension K：K=P22y-P21yl1cos (arctan (P12x-P11xP12y-P11y)-Δ θ) K=
P23y-P21yl2cos(arctan(P13x-P11xP13y-P11y)-Δθ)
Wherein, l1 be two point distance of P11 and P12, l2 P13, P23 twos' point distance；
L1=(P11x-P12x) 2+ (P11y-P12y) 2l2=(P11x-P13x) 2+ (P11y-P13y) 2
Then the Δ θ is：Δ θ=arctan (P22y-P21y) l2cos (θ 2)-(P23y-P21y) l1cos (θ 1)
(P23y-P21y)l1sin(θ1)-(P22y-P21y)l1sin(θ2)
Wherein, θ 1=arctan (P12x-P11xP12y-P11y) θ 2=arctan (P13x-P11xP13y-P11y).
The present invention acquires various types of defect parts of textile and non-defective part by cmos camera, establishes sample
This, the image collected is carried out carrying out texture protrusion, goes the processing such as dry.Digitlization defects detection is realized, while is also carried out
The classification storage of defect, convenient for the data analysis of quality testing personnel.
Embodiments of the present invention are explained in detail above in association with attached drawing, but the present invention is not limited to described implementations
Mode.For a person skilled in the art, in the case where not departing from the principle of the invention and spirit, to these embodiments
A variety of change, modification, replacement and modification are carried out, are still fallen in protection scope of the present invention.
Claims (10)
1. a kind of textile fabric defect inspection method, it is characterised in that：Including step：
Various types of defect parts and non-defective part using cmos camera acquisition textile, respectively obtain defect sample
With zero defect sample, and stored；
Using the image information of cmos camera acquisition sample to be tested, and carry out image preprocessing；
Texture enhancing is carried out to texture image using non-local mean filtering；To texture at enhanced textile images
Reason calculates the texture contour images for obtaining textile images；Construct the characteristic model of texture contour images；
The characteristic model match comparing with the zero defect sample, the difference of offset and mean deviation amount is more than in advance
If during threshold value, it is fault model to determine this feature model；When the difference of offset and mean deviation amount is less than predetermined threshold value, determine
This feature model is qualified model；
The fault model and defect sample are carried out one-to-one match comparison, and to flaw sample progress classification storage.
2. textile fabric defect inspection method according to claim 1, it is characterised in that：Calculate the line for obtaining textile images
The process for managing contour images includes step：
First with without sampling wavelet transformation, W (x, y) is obtained；Select the wavelet transformation of i-th of subband；
Each subband wavelet coefficient, prominent each subband texture contour feature are pre-processed using the smooth method of local Gaussian；
Each subband texture profile is merged using multichannel gradient information, obtains the texture contour images of textile images.
3. textile fabric defect inspection method according to claim 1, it is characterised in that：It is acquired using cmos camera to be measured
The image information of sample, and the step of carrying out image preprocessing include：
Utilize the image information of cmos camera acquisition sample to be tested；
Image cutting, label processing are carried out to described image information；
Dry processing is carried out to described image information.
4. the defects of image according to claim 1 detection method, which is characterized in that the characteristic model with it is described intact
The process that sunken sample or the defect sample carry out matching comparison includes：
The pattern of the characteristic model is matched with the pattern in the zero defect sample；
Searching point is calculated according to matching result and corresponds to zero defect sample distance between Searching point, is denoted as offset；
When the difference of the offset and mean deviation amount is more than default bias amount threshold value, it is flaw to determine described search point
Point；Average value of the mean deviation amount for the offset summation of all Searching points in the pattern of the characteristic model；
The offset is compared with mean deviation amount.
5. the defects of image according to claim 4 detection method, which is characterized in that described by the pattern to be measured and institute
The die plate pattern stated in zero defect sample is matched, and is specifically included：The new Searching point formed in the pattern to be measured
Searching point P1 is chosen in set, and the corresponding points P2 of the P1 is chosen in the template image, according to selected P1 and P2
Calculate rotational offset Δ θ, translational offsets amount R and along Y-axis amount of tension K；According to rotational offset Δ θ, translational offsets amount R and edge
Y-axis amount of tension K obtains matching result by described search spot projection to the template image identical with the pattern to be measured；Specifically
, according to P1 and P2 computational lengths D and with horizontal direction angle theta；
D=(P1x-P2x) 2+ (P1y-P2y) 2 θ=arctanP1y-P2yP1x-P2x
Three groups of corresponding points P11, P21, P12, P22, P13 and P23 are chosen in the pattern to be measured and template image,
Then translational offsets amount R is：Ry=P11y-P21y Rx=P11x-P21x
It is along Y-axis amount of tension K：K=P22y-P21yl1cos (arctan (P12x-P11xP12y-P11y)-Δ θ) K=P23y-
P21yl2cos(arctan(P13x-P11xP13y-P11y)-Δθ)
Wherein, l1 be two point distance of P11 and P12, l2 P13, P23 twos' point distance；
L1=(P11x-P12x) 2+ (P11y-P12y) 2l2=(P11x-P13x) 2+ (P11y-P13y) 2
Then the Δ θ is：Δ θ=arctan (P22y-P21y) l2cos (θ 2)-(P23 y-P21y) l1cos (θ 1) (P23y-
P21y)l1sin(θ1)-(P22y-P21y)l1sin(θ2)
Wherein, θ 1=arctan (P12x-P11xP12y-P11y) θ 2=arctan (P13 x-P11xP13y-P11y).
6. a kind of textile fabric defect detecting device, it is characterised in that：Including
Sample cmos camera：Various types of defect parts and non-defective part using cmos camera acquisition textile,
Defect sample and zero defect sample are respectively obtained, and is stored, and acquires the image information of sample to be tested；
Image pre-processing module：Image preprocessing is carried out to the image information of the sample to be tested；
Image enhancement modeling module：Texture enhancing is carried out to texture image using non-local mean filtering, it is enhanced to texture
Textile images are handled, and calculate the texture contour images for obtaining textile images, construct the character modules of texture contour images
Type；
Sample comparing module：The characteristic model match comparing with the zero defect sample, offset and mean deviation
When the difference of amount is more than predetermined threshold value, it is fault model to determine this feature model；The difference of offset and mean deviation amount is less than
During predetermined threshold value, it is qualified model to determine this feature model；
Defect classification module：The fault model and defect sample are carried out one-to-one match comparison, and to the progress of flaw sample
Classification storage.
7. textile fabric defect detecting device according to claim 6, it is characterised in that：Described image enhances modeling module packet
It includes：
Wavelet transformation module：Using without sampling wavelet transformation, W (x, y) is obtained；Select the wavelet transformation of i-th of subband；
Wavelet coefficients preprocessing module：Each subband wavelet coefficient, prominent each subband are pre-processed using the smooth method of local Gaussian
Texture contour feature；
Grain table module：Each subband texture profile is merged using multichannel gradient information, obtains the texture wheel of textile images
Wide image.
8. textile fabric defect detecting device according to claim 6, it is characterised in that：Described image preprocessing module packet
It includes：
Acquisition module：Utilize the image information of cmos camera acquisition sample to be tested；
Cutting module：Image cutting, label processing are carried out to described image information；
Go dry module：Dry processing is carried out to described image information.
9. textile fabric defect detecting device according to claim 6, it is characterised in that：The sample comparing module includes：
Matching module：The pattern of the characteristic model is matched with the die plate pattern in the zero defect sample；
Offset computing module：Searching point is calculated according to matching result and corresponds to die plate pattern distance between Searching point, is denoted as
Offset；
Offset threshold calculation module：When the difference of the offset and mean deviation amount is more than default bias amount threshold value, really
Described search point is determined for flaw point；The mean deviation amount is total for the offset of all Searching points in the pattern of the characteristic model
The average value of sum；
Contrast module：The offset is compared with mean deviation amount.
10. textile fabric defect detecting device according to claim 6, it is characterised in that：It is described by the pattern to be measured with
Pattern in the zero defect sample is matched, and is specifically included：The new search point set formed in the pattern to be measured
Searching point P1 is chosen in conjunction, and the corresponding points P2 of the P1 is chosen in the template image, is counted according to selected P1 and P2
Calculate rotational offset Δ θ, translational offsets amount R and along Y-axis amount of tension K；According to rotational offset Δ θ, translational offsets amount R and along Y
Axis amount of tension K obtains matching result by described search spot projection to the template image identical with the pattern to be measured；Specifically
, according to P1 and P2 computational lengths D and with horizontal direction angle theta；
D=(P1x-P2x) 2+ (P1y-P2y) 2 θ=arctanP1y-P2yP1x-P2x
Three groups of corresponding points P11, P21, P12, P22, P13 and P23 are chosen in the pattern to be measured and template image,
Then translational offsets amount R is：Ry=P11y-P21y Rx=P11x-P21x
It is along Y-axis amount of tension K：K=P22y-P21yl1cos (arctan (P12x-P11xP12y-P11y)-Δ θ) K=P23y-
P21yl2cos(arctan(P13x-P11xP13y-P11y)-Δθ)
Wherein, l1 be two point distance of P11 and P12, l2 P13, P23 twos' point distance；
L1=(P11x-P12x) 2+ (P11y-P12y) 2l2=(P11x-P13x) 2+ (P11y-P13y) 2
Then the Δ θ is：Δ θ=arctan (P22y-P21y) l2cos (θ 2)-(P23 y-P21y) l1cos (θ 1) (P23y-
P21y)l1sin(θ1)-(P22y-P21y)l1sin(θ2)
Wherein, θ 1=arctan (P12x-P11xP12y-P11y) θ 2=arctan (P13 x-P11xP13y-P11y).
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711217285.4A CN108133472A (en) | 2017-11-28 | 2017-11-28 | Textile fabric defect inspection method and textile fabric defect detecting device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711217285.4A CN108133472A (en) | 2017-11-28 | 2017-11-28 | Textile fabric defect inspection method and textile fabric defect detecting device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108133472A true CN108133472A (en) | 2018-06-08 |
Family
ID=62389789
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201711217285.4A Pending CN108133472A (en) | 2017-11-28 | 2017-11-28 | Textile fabric defect inspection method and textile fabric defect detecting device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108133472A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116862917A (en) * | 2023-09-05 | 2023-10-10 | 微山县振龙纺织品有限公司 | Textile surface quality detection method and system |
| CN117893467A (en) * | 2023-12-06 | 2024-04-16 | 江苏新丝路纺织科技有限公司 | Textile defect type identification method |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPH0843318A (en) * | 1994-08-01 | 1996-02-16 | Kanebo Ltd | Method and device for detecting defect of texture |
| CN101187640A (en) * | 2006-11-15 | 2008-05-28 | 上海市纤维检验所 | Method and device for detecting textile structure and dying defect using infrared laser |
| CN102331425B (en) * | 2011-06-28 | 2013-04-03 | 合肥工业大学 | Textile defect detection method based on defect enhancement |
| CN103413314A (en) * | 2013-08-19 | 2013-11-27 | 合肥师范学院 | Textile defect detecting algorithm based on texture gradients |
| CN103954627A (en) * | 2014-04-21 | 2014-07-30 | 杭州电子科技大学 | Electronic component surface defect detection method based on sample library dictionary |
| CN104483320A (en) * | 2014-10-27 | 2015-04-01 | 中国计量学院 | Digitized defect detection device and detection method of industrial denitration catalyst |
| CN106780393A (en) * | 2016-12-28 | 2017-05-31 | 辽宁师范大学 | Image de-noising method based on image set |
| CN106815830A (en) * | 2016-12-13 | 2017-06-09 | 中国科学院自动化研究所 | The defect inspection method of image |
-
2017
- 2017-11-28 CN CN201711217285.4A patent/CN108133472A/en active Pending
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPH0843318A (en) * | 1994-08-01 | 1996-02-16 | Kanebo Ltd | Method and device for detecting defect of texture |
| CN101187640A (en) * | 2006-11-15 | 2008-05-28 | 上海市纤维检验所 | Method and device for detecting textile structure and dying defect using infrared laser |
| CN102331425B (en) * | 2011-06-28 | 2013-04-03 | 合肥工业大学 | Textile defect detection method based on defect enhancement |
| CN103413314A (en) * | 2013-08-19 | 2013-11-27 | 合肥师范学院 | Textile defect detecting algorithm based on texture gradients |
| CN103954627A (en) * | 2014-04-21 | 2014-07-30 | 杭州电子科技大学 | Electronic component surface defect detection method based on sample library dictionary |
| CN104483320A (en) * | 2014-10-27 | 2015-04-01 | 中国计量学院 | Digitized defect detection device and detection method of industrial denitration catalyst |
| CN106815830A (en) * | 2016-12-13 | 2017-06-09 | 中国科学院自动化研究所 | The defect inspection method of image |
| CN106780393A (en) * | 2016-12-28 | 2017-05-31 | 辽宁师范大学 | Image de-noising method based on image set |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116862917A (en) * | 2023-09-05 | 2023-10-10 | 微山县振龙纺织品有限公司 | Textile surface quality detection method and system |
| CN116862917B (en) * | 2023-09-05 | 2023-11-24 | 微山县振龙纺织品有限公司 | Textile surface quality detection method and system |
| CN117893467A (en) * | 2023-12-06 | 2024-04-16 | 江苏新丝路纺织科技有限公司 | Textile defect type identification method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108520274A (en) | High reflecting surface defect inspection method based on image procossing and neural network classification | |
| KR102536011B1 (en) | System, method for training and applying a defect classifier on wafers with deeply stacked layers | |
| CN111402226A (en) | Surface defect detection method based on cascade convolution neural network | |
| CN105069790B (en) | A kind of gear open defect fast image detection method | |
| CN102854191B (en) | Real-time visual detection and identification method for high speed rail surface defect | |
| CN111325713A (en) | Wood defect detection method, system and storage medium based on neural network | |
| CN110490874A (en) | Weaving cloth surface flaw detecting method based on YOLO neural network | |
| CN111223093A (en) | AOI defect detection method | |
| CN104101600B (en) | Cross Section of CC Billet testing of small cracks method | |
| TW201928869A (en) | An automated optical inspection method using deep learning, and an optical inspection apparatus, computer program, computer readable medium, and deep learning system | |
| CN111982916A (en) | Welding seam surface defect detection method and system based on machine vision | |
| CN114170161A (en) | Cigarette appearance defect detection method and system | |
| CN109919934A (en) | A kind of liquid crystal display panel defect inspection method based on the study of multi-source domain depth migration | |
| CN112991271B (en) | Aluminum profile surface defect visual detection method based on improved yolov3 | |
| CN110689524B (en) | No-reference online image definition evaluation method and system | |
| CN106373123A (en) | K_tSL central clustering algorithm-based industrial component surface defect detection method | |
| CN111189837A (en) | Cigarette appearance online detection method and device | |
| CN110403232A (en) | A kind of cigarette quality detection method based on second level algorithm | |
| CN111724372A (en) | Method, terminal and storage medium for detecting cloth defects based on antagonistic neural network | |
| CN103221807A (en) | Rapid processing and detection of non-uniformities in web-based materials | |
| CN115482195B (en) | Train part deformation detection method based on three-dimensional point cloud | |
| CN110298840A (en) | A kind of yarn faults detection method based on image | |
| CN112150460A (en) | Detection method, detection system, device, and medium | |
| CN109285140A (en) | A kind of printed circuit board image registration appraisal procedure | |
| CN112200790B (en) | Cloth defect detection method, device and medium |
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: 20180608 |
|
| RJ01 | Rejection of invention patent application after publication |