CN109187579A - Fabric defect detection method and device, computer equipment and computer-readable medium - Google Patents
Fabric defect detection method and device, computer equipment and computer-readable medium Download PDFInfo
- Publication number
- CN109187579A CN109187579A CN201811037595.2A CN201811037595A CN109187579A CN 109187579 A CN109187579 A CN 109187579A CN 201811037595 A CN201811037595 A CN 201811037595A CN 109187579 A CN109187579 A CN 109187579A
- Authority
- CN
- China
- Prior art keywords
- fabric
- network model
- fault
- neural network
- detected
- 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
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/84—Systems specially adapted for particular applications
- G01N21/88—Investigating the presence of flaws or contamination
- G01N21/95—Investigating the presence of flaws or contamination characterised by the material or shape of the object to be examined
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/84—Systems specially adapted for particular applications
- G01N21/88—Investigating the presence of flaws or contamination
- G01N21/8851—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges
Landscapes
- Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Chemical & Material Sciences (AREA)
- Analytical Chemistry (AREA)
- Biochemistry (AREA)
- General Health & Medical Sciences (AREA)
- General Physics & Mathematics (AREA)
- Immunology (AREA)
- Pathology (AREA)
- Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Signal Processing (AREA)
- Image Analysis (AREA)
Abstract
This application discloses fabric defect detection method and devices, computer equipment and computer-readable medium, this method comprises: obtaining the fabric attributes feature of image to be detected；According to the fabric attributes feature, determine that the fabric that described image to be detected is included whether there is fault.This method, device, computer equipment and computer-readable medium obtain fabric attributes feature first when detecting to image to be detected to comprehensively consider when determining fault, to eliminate interference of the fabric attributes feature to testing result, it realizes the accurate detection to fabric defects, improves accuracy in detection.
Description
Technical field
This application involves Fabric Detection technical fields, in particular to fabric defect detection method and device, computer equipment
And computer-readable medium.
Background technique
On the production line of the fabrics such as woven fabric, looped fabric, non-woven cloth, need whether to detect produced fabric
There are faults, for example, whether having spot, broken hole, fluffing etc. on fabric.
Current detection method mainly is found to knit before perching equipment by testing staff station in such a way that naked eyes detect
Object fault is simultaneously marked fault or records.In the case where the yield of fabric is very big, being detected by testing staff will be very
Take manpower, moreover, testing staff is easy fatigue after a period of time that works, to there is a possibility that erroneous detection occurs.Therefore,
The overall fault detection efficiency detected by testing staff is not high and accuracy in detection is not sufficiently stable.
In the related technology, perching is carried out using computer mainly to realize by machine vision and fault classification method,
That is, cloth image to be detected is obtained by photography technology, by presetting multiple classifications, using detection model according to this
The feature of image determines the probability for each classification that the image is belonging respectively in multiple classification, and the classification of maximum probability is true
It is set to classification described in the image, to obtain the classification of fault described in the cloth image.But it is tested currently with computer
The method of cloth does not all take in the attributive character of fabric itself (such as stamp, texture etc.), often by the category of fabric
Property feature erroneous detection be fault, cause detection accuracy rate it is very low.
Summary of the invention
In view of problem above, the embodiment of the present invention provides a kind of fabric defect detection method and device, computer equipment
And computer-readable medium, it can solve the technical issues of above-mentioned background technology part is mentioned.
The fabric defect detection method of embodiment according to the invention, comprising: the fabric attributes for obtaining image to be detected are special
Sign；According to the fabric attributes feature, determine that the fabric that described image to be detected is included whether there is fault.
The fabric defects detection device of embodiment according to the invention, comprising: module is obtained, for obtaining image to be detected
Fabric attributes feature；Determining module, for according to the fabric attributes feature, determining that described image to be detected included knits
Object whether there is fault.
The computer equipment of embodiment according to the invention, including processor；And memory, it is stored thereon with executable
Instruction, wherein the executable instruction makes the processor execute method above-mentioned upon being performed.
The computer-readable medium of embodiment according to the invention, is stored thereon with executable instruction, wherein described to hold
Row instruction makes computer execute method above-mentioned upon being performed.
It can be seen from the above that the scheme of the embodiment of the present invention obtains first when detecting to image to be detected
Fabric attributes feature does testing result to eliminate fabric attributes feature with being comprehensively considered when determining fault
It disturbs, realizes the accurate detection to fabric defects, improve accuracy in detection.
Detailed description of the invention
Fig. 1 is the flow chart of the fabric defect detection method of one embodiment according to the invention；
Fig. 2 is the process of the method for the fabric attributes feature of acquisition image to be detected of one embodiment according to the invention
Figure；
Fig. 2 a is the structural schematic diagram of a typical RNN model；
Fig. 2 b is the schematic diagram that RNN model is unfolded in time；
Fig. 2 c is a neural network basic unit of the RNN-ResNet model of one embodiment according to the invention
Structural schematic diagram；
Fig. 2 d is a neural network basic unit of the LSTM-ResNet model of one embodiment according to the invention
Structural schematic diagram；
Fig. 3 is one embodiment according to the invention according to the fabric attributes feature, determines described image to be detected
Flow chart of the fabric for being included with the presence or absence of the method for fault；
Fig. 3 a is the knot with attribute/fault discriminant classification network CNN model of one embodiment according to the invention
Structure schematic diagram；
Fig. 3 b is attribute/fault discriminant classification network structural schematic diagram of one embodiment according to the invention；
Fig. 3 c is the structural schematic diagram of the target classification Recurrent networks of one embodiment according to the invention；
Fig. 4 is the flow chart of the method for model training of one embodiment according to the invention；
Fig. 5 is the overview flow chart of the fabric defect detection method of one embodiment according to the invention；
Fig. 6 is the schematic diagram of the fabric defects detection device of one embodiment according to the invention；
Fig. 7 is the schematic diagram of the computer equipment of one embodiment according to the invention.
Specific embodiment
Theme described herein is discussed referring now to example embodiment.It should be understood that discussing these embodiments only
It is in order to enable those skilled in the art can better understand that being not to claim to realize theme described herein
Protection scope, applicability or the exemplary limitation illustrated in book.It can be in the protection scope for not departing from present disclosure
In the case of, the function and arrangement of the element discussed are changed.Each example can according to need, omit, substitute or
Add various processes or component.For example, described method can be executed according to described order in a different order, with
And each step can be added, omits or combine.In addition, feature described in relatively some examples is in other examples
It can be combined.
As used in this article, term " includes " and its modification indicate open term, are meant that " including but not limited to ".
Term "based" indicates " being based at least partially on ".Term " one embodiment " and " embodiment " expression " at least one implementation
Example ".Term " another embodiment " expression " at least one other embodiment ".Term " first ", " second " etc. may refer to not
Same or identical object.Here may include other definition, either specific or implicit.Unless bright in context
It really indicates, otherwise the definition of a term is consistent throughout the specification.
Fig. 1 shows the flow chart of the fabric defect detection method of one embodiment according to the invention.Referring to Fig.1, should
Method 100 may include:
S102 obtains the fabric attributes feature of image to be detected.
Wherein, fabric attributes can refer to the perceptual property that the normal fabric surface of not fault is presented, specifically can be with
The fabric attributes having including fabrics such as stamp, texture, jacquard weave, pattern itself.
It when obtaining the fabric attributes feature of image to be detected, can be such as, but not limited to, extract image to be detected and meet
The feature of fabric attributes parameter is as fabric attributes feature.
The fabric attributes feature for obtaining image to be detected can use the fabric attributes Feature Selection Model constructed in advance, knit
Object attributive character, which extracts model, can obtain fabric attributes Feature Selection Model parameter according to normal fabric sample image, according to mould
Shape parameter obtains fabric attributes feature.
Obtaining fabric attributes feature can be using such as, but not limited to: histogram, gray level co-occurrence matrixes, Markov are random
Field model (MRF-Markov Random Field), autoregression texture model (simultaneous auto-regressive,
SAR), Fourier's shape description symbols (Fourier shape deors), wavelet descriptor (Wavelet Deor), neural network
The methods of model.Preferably, timing neural network model can be used in neural network model, or, timing-residual error neural network
Model, wherein timing-residual error neural network model is each basic unit addition residual error network in timing neural network
The neural network model of composition, the residual error network is by the output weighted superposition of the basic unit last moment to described basic
In the output at unit current time.
Optionally, timing neural network model includes Recognition with Recurrent Neural Network (RNN:Recurrent Neural
Network) model, long short-term memory (LSTM:Long Short-Term Memory) model or gating cycle unit (GRU:
Gated Recurrent Unit).Timing-residual error neural network model includes circulation-residual error neural network (RNN-
ResNet:Recurrent Neural Network-Residual Network) model, long short-term memory-residual error (LSTM-
ResNet:Long Short-Term Memory-Residual Network) model or gating cycle unit-residual error (GRU-
ResNet:Gated Recurrent Unit-Residual Network) model.
S104 determines that the fabric that described image to be detected is included whether there is fault according to the fabric attributes feature.
Wherein it is determined that fabric, which can use the fault constructed in advance with the presence or absence of fault, determines model, determine described to be checked
The fabric that altimetric image is included specifically can be with the presence or absence of fault and match image to be detected with fabric attributes feature, really
Determine region identical and different with fabric attributes feature in image to be detected, different region then determines to be detected if it exists
There are faults for the fabric that image is included, and the fabric that different region then determines that image to be detected is included if it does not exist is not deposited
In fault.
Wherein, the determination of fault can be using such as, but not limited to neural network model, Normalized Grey Level relevant matches, most
Small two multiply Image Matching, geometric graphic element method, fourier shape description method etc..Preferably, neural network model can be using convolution mind
Through network (CNN:Convolutional Neural Network) model.
It can be seen from the above that the scheme of the embodiment of the present invention obtains first when detecting to image to be detected
Fabric attributes feature, for being comprehensively considered when determining fault, to eliminate fabric attributes feature to testing result
Interference realizes the accurate detection to fabric defects, improves accuracy in detection.
Fig. 2 shows fabric attributes feature (the i.e. steps of acquisition image to be detected of one embodiment according to the invention
S102 the flow chart of method).Referring to Fig. 2, this method 200 may include:
S202: the feature vector of one or more candidate regions of described image to be detected is obtained.
Wherein, it can identify that positioning and image dividing processing are one or more of to obtain by executing to image to be detected
Candidate region.Identification positioning and image dividing processing are known technologies, omit descriptions thereof herein.
Here, the characteristics of image of feature vector characterization candidate region, the feature vector for obtaining candidate region can use example
Such as, but not limited to, histogram, gray level co-occurrence matrixes, Markov random field model (MRF-Markov Random Field), from
Return texture model (simultaneous auto-regressive, SAR), Fourier's shape description symbols (Fourier shape
Deors), wavelet descriptor (Wavelet Deor), neural network model or other kinds of feature extractor.
S204: it according to described eigenvector, is obtained corresponding to the candidate region pair using first nerves network model
The fabric attributes feature of elephant.Object corresponding to candidate region may include fault, fabric attributes, and first nerves network model obtains
Take the fabric attributes feature for wherein belonging to the object of fabric attributes.
Wherein, first nerves network model is trained in advance for extracting the deep learning mould of fabric attributes feature
Type, preferably first nerves network model are timing neural network model.In a kind of embodiment, first nerves network model can
To be RNN model.Fig. 2 a show the structural schematic diagram of a typical RNN model, and RNN model includes input unit (Input
Units), output unit (Output Units) and hidden unit (Hidden Units).In present embodiment, by the time of extraction
Input set of the feature vector of favored area as input unit, the output of output unit integrate as fabric attributes feature.Fabric attributes
The fabric attributes feature of object corresponding to characteristic present candidate region is different since fabric attributes have certain regularity
Candidate region corresponding to object may be that there are the attributes of certain rule connection, can be with using the memory function of RNN model
Input by the output of a upper candidate region as current candidate region, to know to the fabric attributes with regularity
Not.It should be noted that first nerves network model can be multilayer check configuration or multi-layer biaxially oriented structure, each layer of nerve net
Multiple basic units can be used in network.Fig. 2 b show the schematic diagram that RNN model is unfolded in time, wherein U, V, W are network ginseng
Number, neural network basic unit calculation formula are as follows:
st=f (Uxt+Wst-1)
ot=SOFTMAX (Vst)
Wherein, xtFor external world's input of t moment, stOutput is remembered for the RNN neural network unit of t moment, and U, V, W are net
Network parameter, f can be the functions such as tanh, otFor the output of t moment.When it is implemented, first nerves network model can be exported
Fabric attributes information coding be fabric attributes feature, to obtain the fabric attributes feature of the corresponding object in the candidate region.
If the RNN model number of plies used is more, due to increasing with the number of plies, led when using back-propagation method calculating
When number, the range value of the gradient (initial several layers of from output layer to network) of backpropagation can sharp reduce, and as a result make
It is very small relative to the derivative of initially several layers of weights at whole loss function, in this way, when using gradient descent method
It waits, initially several layers of weight variations is very slow, so that they can not effectively be learnt from training sample, thus
There is the phenomenon that gradient disperse (diffusion of gradients).Based on this, in another embodiment, first nerves net
Network model combines by using RNN and residual error network (ResNet:Residual Network) and forms circulation-residual error neural network
The mode of model solves the problem above-mentioned.In the present embodiment, ResNet is added in RNN to connect to form RNN-ResNet mould
Type, wherein ResNet can be by the output weighted superposition of RNN last moment to currently exporting, so that deeper neural network is easy
In training.
In the present embodiment, RNN can be common recognition sequence network, it is to be understood that add the essence of ResNet
Process is that RNN basic unit is added, and Fig. 2 c is that the neural network of RNN-ResNet provided in this embodiment is substantially single
The structural schematic diagram of member, the neural network basic unit calculation formula after addition are as follows:
st=f (Uxt+Wst-1)+α·st-1
ot=SOFTMAX (Vst)
Wherein, xtFor external world's input of t moment, stIt is exported for the RNN-ResNet neural network unit memory of t moment, U,
V, W is network parameter, and f can be the functions such as tanh, otFor the output of t moment, α is residual error coefficient.When it is implemented, can be by
The fabric attributes information coding of one neural network model output is fabric attributes feature, corresponding right to obtain the candidate region
The fabric attributes feature of elephant.
It is understood that residual error coefficient α is added in RNN basic unit, so that the memory of RNN basic unit
Export stItem increases α st-1, it will be in the output weighted superposition of RNN last moment to current output.When α is 0, as
Common RNN basic unit, the f (Ux when α is 1, in RNN basic unitt+Wst-1) it is equivalent to study st-st-1, that is, introduce residual
Poor mechanism is the compromise proposal of two kinds of situations as 0 < α < 1.
Fig. 2 d shows the structural representation of a neural network basic unit of the LSTM-ResNet model of one embodiment
Figure.Identical as the principle of above-mentioned RNN-ResNet, as shown in Figure 2 d, the substantive process of addition ResNet is basic for LSTM
Unit is added, so that basic unit is in output stIn increase α st-1, it is defeated by LSTM unit last moment
It is weighted in the output at current time out.The principle of GRU-ResNet model is same as described above, omits descriptions thereof herein.
Fig. 3 show one embodiment according to the invention according to the fabric attributes feature, determine described to be detected
Flow chart of the fabric that image is included with the presence or absence of the method for fault (i.e. step S104).Referring to Fig. 3, this method 300 includes:
S302 obtains image to be detected one or more candidate region.
Wherein, it can identify that positioning and image dividing processing are one or more of to obtain by executing to image to be detected
Candidate region.Identification positioning and image dividing processing are known technologies, omit descriptions thereof herein.
S304, according to the fabric attributes feature and the candidate region, nervus opticus network model that utilization has been trained
It determines in the candidate region with the presence or absence of fault effective coverage, whether is deposited with the fabric that the described image to be detected of determination is included
In fault.
Wherein, nervus opticus network model is to be trained in advance for determining whether is fabric that image to be detected is included
There are the deep learning models of fault.In a kind of embodiment, nervus opticus network model be can be with attribute/fault classification
The CNN model for differentiating network, is to increase by one attribute/fault discriminant classification network nerve in original convolutional neural networks
Network model, CNN model described in present embodiment may include most basic CNN model, or, R-CNN, Fast R-CNN,
The improved model on CNN such as Faster R-CNN.Fig. 3 a is that the present embodiment has attribute/fault discriminant classification network CNN
The structural schematic diagram figure of model, the present embodiment increase an attribute/fault discriminant classification network in original CNN model, lead to
It crosses the attribute/fault discriminant classification network and constitutes two classifiers, the input of attribute/fault discriminant classification network is fabric category
Property feature and candidate region, export as fault effective coverage.Attribute/fault discriminant classification network is according to fabric attributes feature to time
Favored area carries out two discriminant classification of attribute/fault, then excludes the attribute inactive area for being identified as fabric attributes, retains quilt
It is determined as the fault effective coverage of fault.
Specifically, having attribute/fault discriminant classification network CNN model includes convolution feature extraction network, attribute/defect
Point discriminant classification network and target classification Recurrent networks.Wherein, convolution feature extraction network is used to extract the feature of candidate region
Figure, specific convolution feature extraction network can be using AlexNet, ZFnet, GoogleNet, VGG as convolution feature extraction
Network.Wherein, attribute/fault discriminant classification network input layer is a pond layer, and output layer is one Softmax layers, in
Between be several hidden layers, all layers of cascade connection, when it is implemented, attribute/fault discriminant classification network can be using including one
Pond layer, three full articulamentums (hidden layer) and one Softmax layers.Wherein, target classification Recurrent networks are according to attribute/fault point
Class differentiates the fault effective coverage of network output, extracts provincial characteristics from the characteristic pattern that convolution feature extraction network extracts, from
And it determines in fault effective coverage and returns amendment with the presence or absence of fault and target fault bounding box.If image to be detected whole area
Domain is identified as inactive area, it is determined that the fabric that image to be detected is included be not present fault, if differentiate result there are
Imitate region, it is determined that there are faults for the fabric that image to be detected is included.
Fig. 3 b shows attribute/fault discriminant classification network structural schematic diagram of one embodiment according to the invention.
Attribute/fault discriminant classification network may include sequentially connected ROI region pond layer, three full articulamentums (hidden layer) and
One softmax layers, when it is implemented, attribute/fault discriminant classification network network parameter can be used as shown in table 1.
Table 1, the present embodiment attribute/fault discriminant classification network parameter
Fig. 3 c shows the structural schematic diagram of the target classification Recurrent networks of one embodiment according to the invention.It is specific real
The network parameter of Shi Shi, target classification Recurrent networks can be using as shown in table 2.
Table 2, the present embodiment target classification Recurrent networks parameter
From the above, it can be seen that scheme provided in an embodiment of the present invention, using with attribute/fault discriminant classification net
The CNN model of network can exclude the inactive area for being identified as fabric attributes, retain the effective coverage for being identified as fault, thus
Interference of the fabric attributes feature to testing result is eliminated, the accurate detection to fabric defects is realized, improves accuracy in detection.
In one embodiment, detect fabric that image to be detected is included there are after fault, above-mentioned method
Further include the following contents: described image to be detected being carried out using classifier to classify and determine that described image to be detected is included
The fault type of fabric surface.Wherein, classifier can use the disaggregated model constructed in advance, and classifier may include Bayes
Classifier, Nearest Neighbor Classifier, softmax classifier, SVM classifier one of which.
Wherein, fault type be such as, but not limited to spot, yarn defect, float, printing and dyeing fault, side defect, fold, skew of weft, broken hole,
Hook silk, sanding unevenness, blur, fluffing, scratch, roll line, stop Mark.Wherein, spot include greasy dirt, rust spot, color dot, spot,
Mildew, auxiliary agent spot；Yarn defect includes dead cotton, slubbing, flyings, thick young yarn, soiled yarn, the dry unevenness of item；Float includes broken yarn, knot, leakage
Needle, rotten needle, try to stop people from fighting each other it is elastic, disconnected try to stop people from fighting each other, spacing is unstable, cloth cover plays snake, filling is shown up, color fibre, needle path, mistake yarn, yarn trace, dew of trying to stop people from fighting each other
Bottom tries to stop people from fighting each other and shows up；Printing and dyeing fault include bite, stamp displacement, stamp staining, stamp cross bottom, stamp is bad, dyeing flower, two tone colour,
Lose colour, difference；When defect includes pin hole, double needle hole, crimping, rotten side, narrow envelope, wealthy envelope；Fold includes intermediate catcher mark, cloth cover
Wrinkle, folding line；Skew of weft includes twill, arch.
Fig. 4 shows the flow chart of the method for model training of one embodiment according to the invention.It is shown in Fig. 4
Method 400 corresponds to the training stage, obtains the RNN-ResNet model for fabric defects detection using training data, has
Attribute/fault discriminant classification network CNN model and SVM classifier.Method 400 shown in Fig. 4 can by computer or other
Suitably there is the electronic equipment of computing capability to realize.
As shown in figure 4, S402, receives the image of multiple original shootings.Wherein, the image of multiple original shooting includes more
The normal picture of a no fault and it is multiple have the problem of fault image, the multiple normal picture include have it is successional multiple
Image and do not have successional multiple images.At least one fabric attributes is spliced to form with successional multiple images to follow
Ring, such as institutional framework circulation, stamp circulation.
S404 executes image labeling (Image Annotation) processing to the image of multiple original shooting, to obtain
First sample image set SP1.Wherein, each of first sample image set SP1 sample image is to multiple original shooting
The one of image of image execute image labeling and handle.Image labeling processing is known technology, is omitted herein
Descriptions thereof.Each image can contain the markup information of one or more attributes, such as the markup information about attribute, pass
In the markup information of fault.
S406 executes gray processing processing to first sample image set SP1, will be each in first sample image set SP1
Sample image is converted to gray level image.
S408 chooses some or all sample images as kind from the first sample image set SP1 that gray processing is handled
Subgraph.
S410 executes one or many angularly rotations, mirror image and/or other suitable behaviour to each drawing of seeds picture
Make, with from obtaining one or more images derived from each drawing of seeds picture.Wherein, the first sample image of gray processing processing
Collect the sample image in SP1 and is together to form the second sample graph image set SP2 from the image obtained derived from each drawing of seeds picture.
By the operation of S408 and S410, the quantity of training sample can be increased (for example, can be by 2500 sample images
The sample image more than 50000 or even 100000 is obtained after treatment), and with the increase of training samples, finally instruct
The neural network model got has higher accuracy in detection.
S406-S410 constitutes the image preprocessing process (Image Preprocessing) of method 400.
S412 obtains the property parameters of each gray level image in the second sample graph image set SP2, wherein the property parameters
The including but not limited to length of image, width etc..
S414, from each rule chosen in the second sample graph image set SP2 in its property parameters the first rule set of satisfaction
Multiple images, as training the third sample graph image set SP3 of RNN-ResNet.Wherein, first rule set is for defining
The condition that sample image suitable for training RNN-ResNet needs to meet.For example, the definition of the first rule set is suitable for RNN-
Length limitation, width limitation that the sample image of ResNet needs to meet etc..Wherein, third sample graph image set SP3 includes having
Successional multiple images.
Under normal conditions, the part of the surface attribute of fabric has the rule of loop cycle, by with successional multiple
Image trains RNN-ResNet, can utilize its memory function, carry out detection training to the fabric attributes with regularity.
S416, chosen from the second sample graph image set SP2 its property parameters meet Second Rule concentration it is each rule
Multiple images, as the 4th sample graph image set SP4 for training with attribute/fault discriminant classification network CNN model.Its
In, which, which is used to define, is suitable for the sample image that training has attribute/fault discriminant classification network CNN model
The condition for needing to meet.For example, Second Rule collection, which defines, is suitable for training with attribute/fault discriminant classification network CNN mould
Length limitation, width limitation that the sample image of type needs to meet etc..4th sample graph image set SP4 include multiple normal pictures and
Multiple problem images.
S418, chosen from the second sample graph image set SP2 its property parameters meet in third rule set it is each rule
Multiple images, as training the 5th sample graph image set SP5 of SVM classifier.Wherein, the third rule set is suitable for defining
Condition for training the sample image of SVM classifier to need to meet.For example, the definition of third rule set is suitable for SVM points of training
Length limitation, width limitation that the sample image of class device needs to meet etc..5th sample graph image set SP5 includes multiple a variety of faults
The problem of classification image.
S420, uses third sample graph image set SP3 as training data, and training obtains RNN-ResNet.
S422, uses the 4th sample graph image set SP4 as training data, and training is obtained with attribute/fault discriminant classification
The CNN model of network.
S424, uses the 5th sample graph image set SP5 as training data, and training obtains SVM classifier.
Fig. 5 shows the overview flow chart of the fabric defect detection method of one embodiment according to the invention.Shown in Fig. 5
Method 500 correspond to the actually detected stage, application method 400 training obtain RNN-ResNet model, have attribute/defect
The CNN model and SVM classifier of point discriminant classification network, which carry out the fabric surface that detection image is included, whether there is fault.Method
500 can for example by computer or other suitably there is the electronic equipment of computing capability to realize.Referring to Fig. 5, this method 500
Include:
S502 obtains image to be detected T.
Optionally, image to be detected T can be acquired by CCD industrial camera, wherein above-mentioned CCD (Charge Coupled
Device, photosensitive coupling component) it is in digital camera for recording the semiconductor subassembly of light variation.
S504 executes pretreatment to image to be detected T, such as, but not limited to, image T is converted to gray level image etc..
S506 executes identification positioning and image dividing processing to pretreated image T, obtains one or more candidate regions
Domain.
S508 obtains the fabric attributes feature of object corresponding to the candidate region using RNN-ResNet model.
S510, according to the fabric attributes feature, using with attribute/fault discriminant classification network CNN model inspection
It whether there is fault effective coverage in the candidate region.
S512, if it does not exist fault effective coverage, it is determined that fault is not present in the fabric that image T is included, then process
Terminate.
S514, fault effective coverage if it exists, then using classifier to fault object corresponding to fault effective coverage into
Row classification.
S516 determines the fault type for the fabric surface that image T is included and exports the information of fault type, then process
Terminate.
Other modifications
It will be understood by those skilled in the art that although in the above embodiments, method 500 includes executing to image to be detected
Pretreated step S504, however, the present invention is not limited thereto.In other embodiments of the invention, such as but do not limit to
In in the case where being suitable under the original state of image T using model to be detected, method 500 can not also include
Pretreated step S504 is executed to image to be detected.
It will be understood by those skilled in the art that although in the above embodiments, method 400 includes to the received figure of institute
Step S404 as executing image labeling processing, however, the present invention is not limited thereto.In other embodiments of the invention,
Such as, but not limited to, in the case where the received images of step S402 institute have executed image labeling processing, method 400 can also be with
Do not include the steps that executing institute's received image image labeling processing S404.
Although it will be understood by those skilled in the art that in the above embodiments, method 400 include step S406-S408 with
Derivative more sample images, however, the present invention is not limited thereto.In other embodiments of the invention, such as but not office
It is limited to, in the case where the quantity of existing sample image is enough, method 400 can not also include step S406-S408.
Although it will be understood by those skilled in the art that in the above embodiments, method 400 include step S408-S410 with
The quantity of equilibrium problem sample image and normal sample image and more sample images are obtained, however, the present invention not office
It is limited to this.In other embodiments of the invention, such as, but not limited to, the problem sample in the received image of step S402 institute
The quantity of image and normal sample image be balance and quantity it is enough in the case where, method 400 can not also include
Step S408-S410.
Although method 400 includes step S406 with by sample it will be understood by those skilled in the art that in the above embodiments
Image is converted into gray level image, however, the present invention is not limited thereto.In other embodiments of the invention, such as but not office
It is limited to, in the case where the received image of step S402 institute has been gray level image, method 400 can not also include step S406.
Although it will be understood by those skilled in the art that in the above embodiments, method 400 include step S412-S418 with
It chooses and is suitable for training RNN-ResNet, with attribute/CNN model of fault sorter network and the sample image of SVM classifier,
However, the present invention is not limited thereto.In other embodiments of the invention, method 400 can not also include step S412-
S418。
Although it will be understood by those skilled in the art that in the above embodiments, being trained to for obtaining fabric attributes feature
Neural network model be RNN-ResNet model, however, the present invention is not limited thereto.In the other embodiment of the present invention
In, it is trained to can also be for the neural network model for obtaining fabric attributes feature including RNN model, LSTM model or GRU
Model can also be GRU-ResNet mould of LSTM-ResNet model or GRU of the LSTM in conjunction with ResNet in conjunction with ResNet
Type.
Although it will be understood by those skilled in the art that in the above embodiments, being trained to for detecting fault effective coverage
Neural network model be that there is attribute/fault discriminant classification network CNN model, however, the present invention is not limited thereto.?
In the other embodiment of the present invention, it is trained to can also be that other have for the neural network model for detecting fault effective coverage
The neural network model of discriminant classification function.
Although it will be understood by those skilled in the art that in the above embodiments, being trained to for fault effective coverage institute
The classifier that corresponding fault object is classified is SVM classifier, however, the present invention is not limited thereto.Of the invention
In other embodiments, it is trained to for can also be to the classifier that fault object is classified corresponding to fault effective coverage
Bayes classifier, Nearest Neighbor Classifier, softmax classifier or other kinds of classifier.
Fig. 6 shows the schematic diagram of the fabric defects detection device of one embodiment according to the invention.Dress shown in fig. 6
Setting 600 can use the mode of software, hardware or software and hardware combining to realize.Device 600 for example may be mounted at computer or
Other suitably have in the electronic equipment of computing capability.Device 600 is corresponding with above-mentioned fabrics defect detection method, due to dress
The embodiment for setting 600 is substantially similar to the embodiment of method, so describing fairly simple, related place is referring to embodiment of the method
Part explanation.
As shown in fig. 6, device 600 may include obtaining module 602 and determining module 604.Module 602 is obtained for obtaining
The fabric attributes feature of image to be detected.Determining module 604 is used to determine the mapping to be checked according to the fabric attributes feature
As the fabric for being included is with the presence or absence of fault.
In one aspect, module 602 is obtained to be further used for using described in the first nerves network model acquisition trained
The fabric attributes feature of object corresponding to one or more candidate regions of image to be detected.
Optionally, the first nerves network model is timing neural network model, or, timing-residual error nerve net
Network model；Wherein, the timing-residual error neural network model is residual in the addition of each basic unit of timing neural network
The neural network model that poor network is constituted, the residual error network is by the output weighted superposition of the basic unit last moment to institute
In the output for stating basic unit current time.
Further alternative, the timing neural network model includes Recognition with Recurrent Neural Network model, long short-term memory mould
Type or gating cycle model of element；The timing-residual error neural network model includes circulation-residual error neural network model, length
Short-term memory-residual error neural network model or gating cycle unit-residual error neural network model.
On the other hand, determining module 604 is further used for according to the fabric attributes feature and the mapping to be checked
One or more candidate regions of picture, being determined in the candidate region using the nervus opticus network model trained whether there is
Fault effective coverage, the fabric for being included with the described image to be detected of determination is with the presence or absence of fault.
Optionally, the nervus opticus network model is to increase by one attribute/fault point in original convolutional neural networks
Class differentiates the neural network model of network, the attribute/fault discriminant classification network for differentiate in the candidate region whether
There are fault effective coverages.
It is further alternative, if the attribute/fault discriminant classification network include sequentially connected ROI region pond layer,
Do hidden layer and softmax layers.
In yet another aspect, device 600 further includes categorization module.Wherein, if categorization module is used for the mapping to be checked
There are faults for the fabric that picture is included, then classify to described image to be detected using classifier and determine the mapping to be checked
As the fault type for the fabric surface for being included.
Optionally, the classifier include Bayes classifier, Nearest Neighbor Classifier, linear classifier, SVM classifier its
Middle one kind.
Fig. 7 shows the schematic diagram of the computer equipment of one embodiment according to the invention.As shown in fig. 7, equipment 700
It may include processor 702 and memory 704, wherein be stored with executable instruction on memory 704, wherein described executable
Instruction makes processor 702 execute method 100 shown in FIG. 1, method shown in Fig. 2 200, side shown in Fig. 3 upon being performed
Method 300, method shown in Fig. 4 400 or method shown in fig. 5 500.
As shown in fig. 7, equipment 700 can be realized in the form of universal computing device.Equipment 700 can also include that connection is different
The bus 706 of system component (including processor 702 and memory 704).Bus 706 indicate one of a few class bus structures or
It is a variety of, including memory bus or Memory Controller, peripheral bus, graphics acceleration port, processor or using a variety of
The local bus of any bus structures in bus structures.For example, these architectures include but is not limited to industrial standard
Architecture (ISA) bus, microchannel architecture (MAC) bus, enhanced isa bus, Video Electronics Standards Association
(VESA) local bus and peripheral component interconnection (PCI) bus.
Equipment 700 typically comprises a variety of computer system readable media.These media can be it is any can be by equipment
The usable medium of 700 access, including volatile and non-volatile media, moveable and immovable medium.
Memory 704 may include the computer system readable media of form of volatile memory, such as arbitrary access is deposited
Reservoir (RAM) 708 and and/or cache memory 710.Equipment 700 may further include other removable/nonremovable
, volatile/non-volatile computer system storage medium.Only as an example, storage system 712 can be used for reading and writing not removable
Dynamic, non-volatile magnetic media (Fig. 7 do not show, commonly referred to as " hard disk drive ").Although being not shown in Fig. 7, can provide
Disc driver for being read and write to removable non-volatile magnetic disk (such as " floppy disk "), and to removable anonvolatile optical disk
The CD drive of (such as CD-ROM, DVD-ROM or other optical mediums) read-write.In these cases, each driver can
To be connected by one or more data media interfaces with bus 706.Memory 704 may include that at least one program produces
Product, the program product have one group of (for example, at least one) program module, these program modules are configured to perform in the present invention
State the function of the embodiment of Fig. 1,2,3,4 or 5.
Program/utility 714 with one group of (at least one) program module 716, can store in such as memory
In 704, such program module 716 includes but is not limited to operating system, one or more application program, other program modules
And program data, it may include the realization of network environment in each of these examples or certain combination.Program module 716
Usually execute the function and/or method in the embodiment of above-mentioned Fig. 1,2,3,4 or 5 described in the invention.
Equipment 700 can also be logical with one or more external equipments 800 (such as keyboard, sensing equipment, display 900 etc.)
Letter, can also be enabled a user to one or more equipment interact with the equipment 700 communicate, and/or with make the equipment 700
Any equipment (such as network interface card, modem etc.) that can be communicated with one or more of the other calculating equipment communicates.This
Kind communication can be carried out by input/output (I/O) interface 718.Also, equipment 700 can also by network adapter 720 with
One or more network (such as local area network (LAN), wide area network (WAN) and/or public network, such as internet) communication.Such as
Shown in figure, network adapter 720 is communicated by bus 706 with other modules of equipment 700.It should be understood that although not showing in figure
Out, other hardware and/or software module can be used with bonding apparatus 700, including but not limited to: microcode, device driver, superfluous
Remaining processor, external disk drive array, RAID system, tape drive and data backup storage system etc..
Program of the processor 702 by operation storage in memory 704, thereby executing various function application and data
Processing, such as realize neural network model compression method shown in above-described embodiment.
The embodiment of the present invention also provides a kind of computer-readable medium, is stored thereon with executable instruction, wherein described
Executable instruction makes computer execute method 100 shown in FIG. 1, method shown in Fig. 2 200, shown in Fig. 3 upon being performed
Method 300, method shown in Fig. 4 400 or method shown in fig. 5 500.
The computer-readable medium of the present embodiment may include in the memory 704 in above-mentioned embodiment illustrated in fig. 7
RAM708, and/or cache memory 710, and/or storage system 712.
With the development of science and technology, the route of transmission of computer program is no longer limited by tangible medium, it can also be directly from net
Network downloading, or obtained using other modes.Therefore, the computer-readable medium in the present embodiment not only may include tangible
Medium can also include invisible medium.
It will be understood by those skilled in the art that the embodiment of the present invention can provide as method, apparatus or computer program production
Product.Therefore, in terms of the embodiment of the present invention can be used complete hardware embodiment, complete software embodiment or combine software and hardware
Embodiment form.Moreover, it wherein includes computer available programs generation that the embodiment of the present invention, which can be used in one or more,
The meter implemented in the computer-usable storage medium (including but not limited to magnetic disk storage, CD-ROM, optical memory etc.) of code
The form of calculation machine program product.
The embodiment of the present invention be referring to according to the method for the embodiment of the present invention, the process of device and computer program product
Figure and/or block diagram describe.It should be understood that every one stream in flowchart and/or the block diagram can be realized by computer program instructions
The combination of process and/or box in journey and/or box and flowchart and/or the block diagram.It can provide these computer programs
Instruct the processor of general purpose computer, special purpose computer, Embedded Processor or other programmable data processing terminal devices
To generate a machine, so that being produced by the instruction that computer or the processor of other programmable data processing terminal devices execute
Life is for realizing the function of specifying in one or more flows of the flowchart and/or one or more blocks of the block diagram
Device.
The specific embodiment illustrated above in conjunction with attached drawing describes exemplary embodiment, it is not intended that may be implemented
Or fall into all embodiments of the protection scope of claims." exemplary " meaning of the term used in entire this specification
Taste " be used as example, example or illustration ", be not meant to than other embodiments " preferably " or " there is advantage ".For offer pair
The purpose of the understanding of described technology, specific embodiment include detail.However, it is possible in these no details
In the case of implement these technologies.In some instances, public in order to avoid the concept to described embodiment causes indigestion
The construction and device known is shown in block diagram form.
The foregoing description of present disclosure is provided so that any those of ordinary skill in this field can be realized or make
Use present disclosure.To those skilled in the art, the various modifications carried out to present disclosure are apparent
, also, can also answer generic principles defined herein in the case where not departing from the protection scope of present disclosure
For other modifications.Therefore, present disclosure is not limited to examples described herein and design, but disclosed herein with meeting
Principle and novel features widest scope it is consistent.
Claims (20)
1. fabric defect detection method, comprising:
Obtain the fabric attributes feature of image to be detected；
According to the fabric attributes feature, determine that the fabric that described image to be detected is included whether there is fault.
2. according to the method described in claim 1, wherein, the fabric attributes feature for obtaining image to be detected includes:
It is obtained corresponding to one or more candidate regions of described image to be detected using the first nerves network model trained
Object fabric attributes feature.
3. according to the method described in claim 2, wherein,
The first nerves network model is timing neural network model, or, timing-residual error neural network model；Wherein,
The timing-residual error neural network model is each basic unit addition residual error network composition in timing neural network
Neural network model, the residual error network work as the output weighted superposition of the basic unit last moment to the basic unit
In the output at preceding moment.
4. according to the method described in claim 3, wherein,
The timing neural network model includes Recognition with Recurrent Neural Network model, long memory models or gating cycle unit mould in short-term
Type；The timing-residual error neural network model includes circulation-residual error neural network model, long short-term memory-residual error nerve net
Network model or gating cycle unit-residual error neural network model.
5. according to the described in any item methods of claim 2-4, wherein it is described according to the fabric attributes feature, determine described in
The fabric that image to be detected is included includes: with the presence or absence of fault
According to one or more candidate regions of the fabric attributes feature and described image to be detected, utilization trained second
Neural network model determines in the candidate region with the presence or absence of fault effective coverage, is included with the described image to be detected of determination
Fabric whether there is fault.
6. according to the method described in claim 5, wherein,
The nervus opticus network model is to increase by one attribute/fault discriminant classification network in original convolutional neural networks
Neural network model, the attribute/fault discriminant classification network are effective with the presence or absence of fault in the candidate region for differentiating
Region.
7. according to the method described in claim 6, wherein,
The attribute/fault discriminant classification network includes sequentially connected ROI region pond layer, several hidden layers and softmax layers.
8. according to the method described in claim 1, wherein, the method also includes:
If there are faults for the fabric that described image to be detected is included, described image to be detected is divided using classifier
Class and the fault type for determining the fabric surface that described image to be detected is included.
9. according to the method described in claim 8, wherein,
The classifier include Bayes classifier, Nearest Neighbor Classifier, linear classifier, SVM classifier one of which.
10. fabric defects detection device, comprising:
Module is obtained, for obtaining the fabric attributes feature of image to be detected；
Determining module, for according to the fabric attributes feature, the fabric for determining that described image to be detected is included to whether there is
Fault.
11. device according to claim 10, wherein the acquisition module is further used for:
It is obtained corresponding to one or more candidate regions of described image to be detected using the first nerves network model trained
Object fabric attributes feature.
12. device according to claim 11, wherein
The first nerves network model is timing neural network model, or, timing-residual error neural network model；Wherein,
The timing-residual error neural network model is each basic unit addition residual error network composition in timing neural network
Neural network model, the residual error network work as the output weighted superposition of the basic unit last moment to the basic unit
In the output at preceding moment.
13. device according to claim 12, wherein
The timing neural network model includes Recognition with Recurrent Neural Network model, long memory models or gating cycle unit mould in short-term
Type；The timing-residual error neural network model includes circulation-residual error neural network model, long short-term memory-residual error nerve net
Network model or gating cycle unit-residual error neural network model.
14. the described in any item devices of 1-13 according to claim 1, wherein the determining module is further used for:
According to one or more candidate regions of the fabric attributes feature and described image to be detected, utilization trained second
Neural network model determines in the candidate region with the presence or absence of fault effective coverage, is included with the described image to be detected of determination
Fabric whether there is fault.
15. device according to claim 14, wherein
The nervus opticus network model is to increase by one attribute/fault discriminant classification network in original convolutional neural networks
Neural network model, the attribute/fault discriminant classification network are effective with the presence or absence of fault in the candidate region for differentiating
Region.
16. device according to claim 15, wherein
The attribute/fault discriminant classification network includes sequentially connected ROI region pond layer, several hidden layers and softmax layers.
17. device according to claim 10, wherein described device further include:
Categorization module, if the fabric for being included for described image to be detected there are fault, using classifier to it is described to
Detection image carries out the fault type for classifying and determining the fabric surface that described image to be detected is included.
18. device according to claim 17, wherein
The classifier include Bayes classifier, Nearest Neighbor Classifier, linear classifier, SVM classifier one of which.
19. computer equipment, comprising:
Processor, and
Memory is stored thereon with executable instruction, wherein the executable instruction holds the processor
Any one of method of row claim 1-9.
20. computer-readable medium is stored thereon with executable instruction, wherein the executable instruction makes upon being performed
Any one of method of computer perform claim requirement 1-9.
Priority Applications (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811037595.2A CN109187579A (en) | 2018-09-05 | 2018-09-05 | Fabric defect detection method and device, computer equipment and computer-readable medium |
| PCT/CN2019/096973 WO2020048248A1 (en) | 2018-09-05 | 2019-07-22 | Textile defect detection method and apparatus, and computer device and computer-readable medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811037595.2A CN109187579A (en) | 2018-09-05 | 2018-09-05 | Fabric defect detection method and device, computer equipment and computer-readable medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109187579A true CN109187579A (en) | 2019-01-11 |
Family
ID=64915093
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811037595.2A Pending CN109187579A (en) | 2018-09-05 | 2018-09-05 | Fabric defect detection method and device, computer equipment and computer-readable medium |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN109187579A (en) |
| WO (1) | WO2020048248A1 (en) |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110188833A (en) * | 2019-06-04 | 2019-08-30 | 北京字节跳动网络技术有限公司 | Method and apparatus for training pattern |
| CN110570397A (en) * | 2019-08-13 | 2019-12-13 | 创新奇智(重庆)科技有限公司 | Method for detecting ready-made clothes printing defects based on deep learning template matching algorithm |
| WO2020048248A1 (en) * | 2018-09-05 | 2020-03-12 | 深圳灵图慧视科技有限公司 | Textile defect detection method and apparatus, and computer device and computer-readable medium |
| CN110889838A (en) * | 2019-11-26 | 2020-03-17 | 武汉纺织大学 | Fabric defect detection method and device |
| CN111402226A (en) * | 2020-03-13 | 2020-07-10 | 浙江工业大学 | Surface defect detection method based on cascade convolution neural network |
| CN112907521A (en) * | 2021-02-05 | 2021-06-04 | 上海展湾信息科技有限公司 | SSD and GLCM based welding air hole target detection method |
| CN113723527A (en) * | 2021-09-01 | 2021-11-30 | 重庆赛迪奇智人工智能科技有限公司 | Classification model training method, flaw detection device and electronic equipment |
| CN114419004A (en) * | 2022-01-21 | 2022-04-29 | 佛山技研智联科技有限公司 | Fabric flaw detection method and device, computer equipment and readable storage medium |
| WO2022185474A1 (en) * | 2021-03-04 | 2022-09-09 | 日本電気株式会社 | Training device, training method, inspection device, inspection method, and recording medium |
| CN116008289A (en) * | 2023-02-07 | 2023-04-25 | 浙江聚优非织造材料科技有限公司 | Nonwoven product surface defect detection method and system |
| JP7517586B2 (en) | 2021-03-04 | 2024-07-17 | 日本電気株式会社 | Learning device, learning method, inspection device, inspection method, and program |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106845556A (en) * | 2017-02-09 | 2017-06-13 | 东华大学 | A kind of fabric defect detection method based on convolutional neural networks |
| CN106996935A (en) * | 2017-02-27 | 2017-08-01 | 华中科技大学 | A kind of multi-level fuzzy judgment Fabric Defects Inspection detection method and system |
| CN107369155A (en) * | 2017-07-24 | 2017-11-21 | 广东工业大学 | A kind of cloth surface defect detection method and its system based on machine vision |
| CN108133473A (en) * | 2017-12-21 | 2018-06-08 | 江南大学 | Warp knitted jacquard fabric defect detection method based on Gabor filtering and deep neural network |
| CN108414525A (en) * | 2018-01-30 | 2018-08-17 | 广东溢达纺织有限公司 | Fabric defect detection method, device, computer equipment and storage medium |
Family Cites Families (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1948603A (en) * | 2006-11-10 | 2007-04-18 | 苏州大学 | Method of identifying woven fabric defect |
| US10467459B2 (en) * | 2016-09-09 | 2019-11-05 | Microsoft Technology Licensing, Llc | Object detection based on joint feature extraction |
| CN107169956B (en) * | 2017-04-28 | 2020-02-14 | 西安工程大学 | Color woven fabric defect detection method based on convolutional neural network |
| CN108288272A (en) * | 2018-02-08 | 2018-07-17 | 首都师范大学 | Yarn recognition methods and device |
| CN109215022A (en) * | 2018-09-05 | 2019-01-15 | 深圳灵图慧视科技有限公司 | Cloth inspection method, device, terminal device, server, storage medium and system |
| CN109300117A (en) * | 2018-09-05 | 2019-02-01 | 深圳灵图慧视科技有限公司 | Nerve network system, electronic equipment and machine readable media |
| CN109325940A (en) * | 2018-09-05 | 2019-02-12 | 深圳灵图慧视科技有限公司 | Textile detecting method and device, computer equipment and computer-readable medium |
| CN109187579A (en) * | 2018-09-05 | 2019-01-11 | 深圳灵图慧视科技有限公司 | Fabric defect detection method and device, computer equipment and computer-readable medium |
-
2018
- 2018-09-05 CN CN201811037595.2A patent/CN109187579A/en active Pending
-
2019
- 2019-07-22 WO PCT/CN2019/096973 patent/WO2020048248A1/en active Application Filing
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106845556A (en) * | 2017-02-09 | 2017-06-13 | 东华大学 | A kind of fabric defect detection method based on convolutional neural networks |
| CN106996935A (en) * | 2017-02-27 | 2017-08-01 | 华中科技大学 | A kind of multi-level fuzzy judgment Fabric Defects Inspection detection method and system |
| CN107369155A (en) * | 2017-07-24 | 2017-11-21 | 广东工业大学 | A kind of cloth surface defect detection method and its system based on machine vision |
| CN108133473A (en) * | 2017-12-21 | 2018-06-08 | 江南大学 | Warp knitted jacquard fabric defect detection method based on Gabor filtering and deep neural network |
| CN108414525A (en) * | 2018-01-30 | 2018-08-17 | 广东溢达纺织有限公司 | Fabric defect detection method, device, computer equipment and storage medium |
Non-Patent Citations (1)
| Title |
|---|
| NGAN HENRY Y.T.等: "Automated fabric defect detection-A review", 《IMAGE AND VISION COMPUTING》 * |
Cited By (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2020048248A1 (en) * | 2018-09-05 | 2020-03-12 | 深圳灵图慧视科技有限公司 | Textile defect detection method and apparatus, and computer device and computer-readable medium |
| CN110188833B (en) * | 2019-06-04 | 2021-06-18 | 北京字节跳动网络技术有限公司 | Method and apparatus for training a model |
| CN110188833A (en) * | 2019-06-04 | 2019-08-30 | 北京字节跳动网络技术有限公司 | Method and apparatus for training pattern |
| CN110570397A (en) * | 2019-08-13 | 2019-12-13 | 创新奇智(重庆)科技有限公司 | Method for detecting ready-made clothes printing defects based on deep learning template matching algorithm |
| CN110889838A (en) * | 2019-11-26 | 2020-03-17 | 武汉纺织大学 | Fabric defect detection method and device |
| CN111402226A (en) * | 2020-03-13 | 2020-07-10 | 浙江工业大学 | Surface defect detection method based on cascade convolution neural network |
| CN112907521A (en) * | 2021-02-05 | 2021-06-04 | 上海展湾信息科技有限公司 | SSD and GLCM based welding air hole target detection method |
| WO2022185474A1 (en) * | 2021-03-04 | 2022-09-09 | 日本電気株式会社 | Training device, training method, inspection device, inspection method, and recording medium |
| JP7517586B2 (en) | 2021-03-04 | 2024-07-17 | 日本電気株式会社 | Learning device, learning method, inspection device, inspection method, and program |
| CN113723527A (en) * | 2021-09-01 | 2021-11-30 | 重庆赛迪奇智人工智能科技有限公司 | Classification model training method, flaw detection device and electronic equipment |
| CN114419004A (en) * | 2022-01-21 | 2022-04-29 | 佛山技研智联科技有限公司 | Fabric flaw detection method and device, computer equipment and readable storage medium |
| CN116008289A (en) * | 2023-02-07 | 2023-04-25 | 浙江聚优非织造材料科技有限公司 | Nonwoven product surface defect detection method and system |
| CN116008289B (en) * | 2023-02-07 | 2024-06-07 | 浙江聚优非织造材料科技有限公司 | Nonwoven product surface defect detection method and system |
Also Published As
| Publication number | Publication date |
|---|---|
| WO2020048248A1 (en) | 2020-03-12 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109187579A (en) | Fabric defect detection method and device, computer equipment and computer-readable medium | |
| CN111738315B (en) | Image classification method based on countermeasure fusion multi-source transfer learning | |
| CN109711474A (en) | A kind of aluminium material surface defects detection algorithm based on deep learning | |
| CN109509187B (en) | Efficient inspection algorithm for small defects in large-resolution cloth images | |
| CN108537136A (en) | The pedestrian's recognition methods again generated based on posture normalized image | |
| CN107330453A (en) | The Pornographic image recognizing method of key position detection is recognized and merged based on substep | |
| CN109598287A (en) | The apparent flaws detection method that confrontation network sample generates is generated based on depth convolution | |
| CN105574550A (en) | Vehicle identification method and device | |
| CN108520114A (en) | A kind of textile cloth defect detection model and its training method and application | |
| CN109543688A (en) | A kind of novel meter reading detection and knowledge method for distinguishing based on multilayer convolutional neural networks | |
| CN105654066A (en) | Vehicle identification method and device | |
| CN108681746A (en) | A kind of image-recognizing method, device, electronic equipment and computer-readable medium | |
| CN109598234A (en) | Critical point detection method and apparatus | |
| CN104200478B (en) | Low-resolution touch screen image defect detection method based on sparse representation | |
| CN109886153A (en) | A kind of real-time face detection method based on depth convolutional neural networks | |
| CN110119688A (en) | A kind of Image emotional semantic classification method using visual attention contract network | |
| CN109509171A (en) | A kind of Fabric Defects Inspection detection method based on GMM and image pyramid | |
| CN107220971A (en) | A kind of Lung neoplasm feature extracting method based on convolutional neural networks and PCA | |
| CN112001219B (en) | Multi-angle multi-face recognition attendance checking method and system | |
| CN109919045A (en) | Small scale pedestrian detection recognition methods based on concatenated convolutional network | |
| CN109670517A (en) | Object detection method, device, electronic equipment and target detection model | |
| CN106874929A (en) | A kind of pearl sorting technique based on deep learning | |
| CN109325940A (en) | Textile detecting method and device, computer equipment and computer-readable medium | |
| CN110245683A (en) | The residual error relational network construction method that sample object identifies a kind of less and application | |
| CN109977832A (en) | A kind of image processing method, device and storage 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 | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20190111 |
|
| WD01 | Invention patent application deemed withdrawn after publication |