CN112465810A - Method for detecting and classifying defects of textiles - Google Patents
Method for detecting and classifying defects of textiles Download PDFInfo
- Publication number
- CN112465810A CN112465810A CN202011477224.3A CN202011477224A CN112465810A CN 112465810 A CN112465810 A CN 112465810A CN 202011477224 A CN202011477224 A CN 202011477224A CN 112465810 A CN112465810 A CN 112465810A
- Authority
- CN
- China
- Prior art keywords
- layer
- neural network
- image
- convolutional neural
- enhanced
- 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
- 230000007547 defect Effects 0.000 title claims abstract description 73
- 239000004753 textile Substances 0.000 title claims abstract description 58
- 238000000034 method Methods 0.000 title claims abstract description 38
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 59
- 238000001514 detection method Methods 0.000 claims abstract description 37
- 238000012549 training Methods 0.000 claims abstract description 26
- 238000012360 testing method Methods 0.000 claims abstract description 16
- 238000012795 verification Methods 0.000 claims abstract description 13
- 238000007781 pre-processing Methods 0.000 claims abstract description 8
- 238000011176 pooling Methods 0.000 claims description 20
- 230000004913 activation Effects 0.000 claims description 14
- 230000006870 function Effects 0.000 claims description 14
- 238000013135 deep learning Methods 0.000 claims description 7
- 230000037303 wrinkles Effects 0.000 claims description 4
- 230000004075 alteration Effects 0.000 claims description 3
- 238000004140 cleaning Methods 0.000 claims description 3
- 230000000694 effects Effects 0.000 abstract description 5
- 239000004744 fabric Substances 0.000 description 4
- 238000004422 calculation algorithm Methods 0.000 description 3
- 238000013528 artificial neural network Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000001914 filtration Methods 0.000 description 2
- 238000004519 manufacturing process Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000001228 spectrum Methods 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 1
- 238000005284 basis set Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000002156 mixing Methods 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000004611 spectroscopical analysis Methods 0.000 description 1
- 238000007619 statistical method Methods 0.000 description 1
- 238000003860 storage Methods 0.000 description 1
- 238000010200 validation analysis Methods 0.000 description 1
Images
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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
-
- 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)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Computing Systems (AREA)
- Biomedical Technology (AREA)
- General Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Molecular Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Health & Medical Sciences (AREA)
- Evolutionary Biology (AREA)
- Quality & Reliability (AREA)
- Investigating Materials By The Use Of Optical Means Adapted For Particular Applications (AREA)
Abstract
The invention provides a method for detecting and classifying defects of textile in order to overcome the defects of low detection efficiency and unsatisfactory effect caused by directly applying a convolutional neural network to defect detection and classification of textiles, which comprises the following steps: collecting textile defect images to form a data set, and preprocessing the data set; dividing the preprocessed data set into a training set, a verification set and a test set; constructing an enhanced Alexnet convolutional neural network; inputting the training set into an enhanced Alexnet convolutional neural network for training, adjusting parameters of the training set and storing the parameters; inputting the verification set into an enhanced Alexnet convolutional neural network which completes training for verification, outputting a detection and classification result of the textile defects, and calculating the accuracy of the detection and classification result; if the accuracy is lower than the preset threshold value, the training is carried out again; and inputting the test set into the trained enhanced Alexnet convolutional neural network, and outputting to obtain the detection and classification results of the textile defects.
Description
Technical Field
The invention relates to the technical field of machine vision, in particular to a method for detecting and classifying defects of textiles.
Background
The currently commonly used defect detection algorithms mainly include 6 types: a statistical method, a frequency spectrum method, a self-learning method, a model method, a mixing algorithm and a pattern method based on airspace. Among them, frequency spectroscopy is most popular because it has a stable characteristic when feature extraction is performed. The frequency spectrum method has the following methods in practical application: fourier transform methods, Gabor filtering methods, wavelet transforms, and some other filtering methods.
Alexnet is a convolutional neural network with great innovation proposed by Alex in 2012, after which deeper and larger convolutional neural networks are proposed. However, these networks are proposed for a large data set (imagenet) with 1000 types of images, each type of image has 1000 images, and direct application to images of textiles is not an optimal method, because the types of defects of textiles are few (only 5 types), and the backgrounds of these images are similar and not as diverse as those of the imagenet, and direct application of convolutional neural networks such as Alexnet to defect detection classification of textiles has problems of low detection efficiency and unsatisfactory effect.
Disclosure of Invention
The invention provides a method for detecting and classifying defects of textile defects, aiming at overcoming the defects of low detection efficiency and unsatisfactory effect caused by directly applying convolutional neural networks such as Alexnet and the like to defect detection and classification of textiles in the prior art.
In order to solve the technical problems, the technical scheme of the invention is as follows:
a method for detecting and classifying textile defects comprises the following steps:
s1: collecting textile defect images to form a data set, and preprocessing the data set;
s2: dividing the preprocessed data set into a training set, a verification set and a test set;
s3: constructing an enhanced Alexnet convolutional neural network by utilizing a caffe deep learning framework; the enhanced Alexnet convolutional neural network comprises a convolutional layer, a ReLU activation function layer, a pooling layer, a filter layer, a full connection layer and a classification layer;
s4: inputting the training set into an enhanced Alexnet convolutional neural network for training, adjusting and storing parameters of the enhanced Alexnet convolutional neural network;
s5: inputting the verification set into an enhanced Alexnet convolutional neural network which completes training for verification, outputting a detection and classification result of the textile defects, and calculating the accuracy of the detection and classification result; if the accuracy is lower than the preset threshold, skipping to execute the step S4;
s6: and inputting the test set into the trained enhanced Alexnet convolutional neural network for testing, and outputting to obtain the detection and classification results of the textile defects.
Preferably, the defect types of the textile defect image include hole, stain, thread end, wrinkle, normal.
Preferably, in the step of S1, the step of preprocessing the data set includes:
s1.1: performing data cleaning on the image;
s1.2: converting the image format into a jpg format;
s1.3: copying the value of a single channel in the image to the other two channels aiming at the image of the single channel to form a three-channel image;
s1.4: adjusting pixels of the image to 256 × 256;
s1.5: and carrying out image expansion on the image.
Preferably, in step S1.4, the step of image expanding the image includes: the method comprises the steps of rotating an image by 90 degrees, rotating the image by 180 degrees, inputting Gaussian noise and changing image chromatic aberration.
Preferably, the enhanced Alexnet convolutional neural network comprises an 8-layer structure connected in sequence, wherein:
the first layer comprises a convolution layer, a ReLU activation function layer and a pooling layer which are connected in sequence;
the second layer comprises a convolution layer with a filling edge set to 2;
the third layer comprises a convolution layer and a pooling layer which are connected in sequence;
the fourth layer comprises a convolution layer and a ReLU activation function layer which are connected in sequence;
the fifth layer comprises a convolution layer and a ReLU activation function layer which are connected in sequence;
the sixth layer comprises a pooling layer;
the seventh and eighth layers comprise fully connected layers with a core number of 2048.
Preferably, in the enhanced Alexnet convolutional neural network, the convolution kernel size of the first convolutional layer is 9 × 9; the convolution layers of the second layer and the third layer are convolution layers with convolution kernel size of 3 x 3 using the same strategy, and the number of convolution kernels is 192.
Preferably, the output of the eighth fully connected layer is connected to the input of the softmax classifier.
Preferably, the output of the eighth layer full link layer is connected to the input of the SVM classifier.
Preferably, the fully connected layer employs a Dropout strategy.
Preferably, in the step S4, in the training process of the enhanced Alexnet convolutional neural network, the initial learning rate is set to 0.001, the initial weight is randomly set, the bias is set to 0.1, the number of iterations is set to 10000, and the number of pictures in each iteration is set to 150.
Compared with the prior art, the technical scheme of the invention has the beneficial effects that: according to the method, the enhanced Alexnet convolutional neural network is constructed by utilizing a caffe deep learning framework, the traditional Alexnet convolutional neural network is improved, the calculated amount of the convolutional neural network is reduced, and meanwhile, the textile defect detection and classification efficiency is effectively improved.
Drawings
Fig. 1 is a flowchart of a method for detecting and classifying textile defects according to example 1.
Fig. 2 is a schematic structural diagram of the enhanced Alexnet convolutional neural network of embodiment 2.
Detailed Description
The drawings are for illustrative purposes only and are not to be construed as limiting the patent;
for the purpose of better illustrating the embodiments, certain features of the drawings may be omitted, enlarged or reduced, and do not represent the size of an actual product;
it will be understood by those skilled in the art that certain well-known structures in the drawings and descriptions thereof may be omitted.
The technical solution of the present invention is further described below with reference to the accompanying drawings and examples.
Example 1
This embodiment proposes a method for detecting and classifying textile defects, which is a flowchart of the method for detecting and classifying textile defects of this embodiment, as shown in fig. 1.
In this embodiment, the defect types of the collected textile defect images include 5 types, such as hole, stain, thread end, wrinkle, normal, and the like.
The method for detecting and classifying textile defects provided by the embodiment comprises the following steps:
s1: collecting textile defect images to form a data set, and preprocessing the data set.
Wherein the step of preprocessing the data set comprises:
s1.1: performing data cleaning on the image, and detecting whether polluted data exist in the data set or not;
s1.2: converting the image format into a jpg format to enable the image to conform to the image format accepted by a caffe deep learning framework;
s1.3: copying a single-channel value in the image to the other two channels to form a three-channel image aiming at the single-channel image;
s1.4: adjusting pixels of the image to 256 × 256;
s1.5: image expansion is performed on the images to increase the number of images of textile defects.
Further, when the image is subjected to image expansion, the image is subjected to 90 ° rotation, the image is subjected to 180 ° rotation, gaussian noise is input, image color difference is changed, and the like.
S2: the preprocessed data set is divided into a training set, a validation set, and a test set.
The training set is used for inputting into an enhanced Alexnet convolutional neural network for training; the verification set is used for verifying the training result of the enhanced Alexnet convolutional neural network and verifying the training result according to the accuracy rate of the output detection classification result; and the test set is used for testing the final detection classification result of the enhanced Alexnet convolutional neural network.
S3: and constructing an enhanced Alexnet convolutional neural network by using a caffe deep learning framework.
The enhanced Alexnet convolutional neural network in this embodiment includes a convolutional layer, a ReLU activation function layer, a pooling layer, a filter layer, a full-link layer, and a classification layer, and specifically, the enhanced Alexnet convolutional neural network includes 8 layers of structures connected in sequence:
the first layer comprises a convolution layer, a ReLU activation function layer and a pooling layer which are connected in sequence, wherein the convolution kernel of the convolution layer is 9 x 9;
the second layer comprises a convolution layer with a filling edge set to 2; the convolution layer with convolution kernel size of 3 x 3 and convolution kernel number of 192;
the third layer comprises a convolution layer and a pooling layer which are connected in sequence; the convolution layer with convolution kernel size of 3 x 3 and convolution kernel number of 192;
the fourth layer comprises a convolution layer and a ReLU activation function layer which are connected in sequence;
the fifth layer comprises a convolution layer and a ReLU activation function layer which are connected in sequence;
the sixth layer comprises a pooling layer;
the seventh and eighth layers comprise fully connected layers with a core number of 2048.
The pooling layers in this embodiment have the same structure and all adopt the maximum pooling strategy.
The full connection layer in this embodiment adopts Dropout strategy.
And the output end of the last full connection layer in the Alexnet convolutional neural network is connected with the input end of a softmax classifier, or the softmax classifier is replaced by an SVM classifier and is connected with the output end of the last full connection layer in the Alexnet convolutional neural network. In this embodiment, a multi-classification SVM classifier implemented based on hindeloss is connected to the output end of the last full connection layer in the Alexnet convolutional neural network.
S4: inputting the training set into an enhanced Alexnet convolutional neural network for training, adjusting and storing parameters of the enhanced Alexnet convolutional neural network;
s5: inputting the verification set into an enhanced Alexnet convolutional neural network which completes training for verification, outputting a detection and classification result of the textile defects, and calculating the accuracy of the detection and classification result; if the accuracy is lower than the preset threshold, skipping to execute the step S4;
s6: and inputting the test set into the trained enhanced Alexnet convolutional neural network for testing, and outputting to obtain the detection and classification results of the textile defects.
In this embodiment, an Alexnet convolutional neural network is improved, a smaller convolutional kernel is adopted, and a structure which has little influence on detection and classification effects, such as a normalization layer in the Alexnet convolutional neural network, is deleted, so that the depth of the convolutional neural network is increased while parameters are reduced, and an enhanced Alexnet convolutional neural network with less calculation amount and higher processing efficiency is obtained.
Example 2
The present embodiment applies the method for detecting and classifying textile defects, which is proposed in embodiment 1, to the detection and classification of textile defects of a data set acquired by a Tilda database. The image types in the data set acquired by the Tilda database comprise 5 types of broken holes, dirt, line ends, folds, normality and the like, and 300 images exist in each image type.
Firstly, a data set consisting of images of defects of the textile is collected through a Tilda database, and the data set is preprocessed. Specifically, a data cleansing operation is performed on the data set to meet the requirements of training and testing.
All pictures in the dataset are then converted to an image format that can be input into the enhanced Alexnet convolutional neural network constructed by the caffe deep learning framework. The picture of the original Tilda database is in a black and white tif image format with 768 × 512 pixels, and considering that the tif format image is not suitable for the enhanced Alexnet convolutional neural network in the embodiment, and the tif format image occupies a larger storage space and requires a larger GPU memory during training, the image format is converted into a jpg format in the embodiment.
Furthermore, a single-channel image in the data set is converted into a three-channel image, and for the single-channel image, a single-channel value in the image is copied to the other two channels to form the three-channel image. The pixels of the Tilda database picture are then adjusted from 768 × 512 to 256 × 256.
In order to increase the diversity of the images, the highlighting after the preprocessing is further subjected to image expansion, which comprises the steps of rotating the images by 90 degrees, rotating the images by 180 degrees, inputting Gaussian noise, changing the chromatic aberration of the images and the like, and expanding the number of the images to be 4 times and the total number to be 8105, wherein 1653 are the images of the broken hole defects, 1390 are the images of the stain defects, 1456 are the images of the normal defects, 1646 are the images of the thread head defects, and 1960 are the images of the wrinkle defects. Wherein training set (6484 sheets) and verification set (1621 sheets) are randomly divided at a ratio of 8:2, and a test set is additionally arranged, wherein the test set comprises 20 sheets of each type of defect images.
And constructing an enhanced Alexnet convolutional neural network by using a caffe deep learning framework. The structure diagram of the enhanced Alexnet convolutional neural network is shown in FIG. 2. In this embodiment, the original input of the enhanced Alexnet convolutional neural network is 256 × 3, and the input image is randomly cropped in the data layer to obtain 227 × 3 image.
In the first layer, the convolution kernel size of the convolution layer is 9 x 9; after passing through the convolution layer, the image size was 56 × 96, and then the image data size was 28 × 96 after passing through the ReLU activation function layer and the pooling layer in this order.
In the second layer, the fill edge of the convolution layer is 2, the convolution kernel size is 3 × 3, and no pooling layer is provided, so the output image data size is 28 × 96.
The third layer is a convolution layer which adopts the same strategy as the second layer, the convolution kernel size is 3 x 3, the pooling layer is arranged, and the size of output image data is 14 x 192.
The fourth layer has the same strategy as the fifth layer, and comprises a convolution layer and a ReLU activation function layer which are connected in sequence, and the sizes of output image data are 14 × 256 and 14 × 192 respectively.
The sixth layer is a pooling layer outputting data of size 7 x 192, which converts the data to a long vector output of length 9408.
The seventh layer and the eighth layer are fully connected layers adopting a Dropout strategy, and the Dropout value of the fully connected layers is set to be 0.5. And the core numbers of the sixth layer and the seventh layer are 2048, the core number of the eighth layer is 5, the output end of the full-connected layer in the eighth layer is connected into an SVM classifier, and 5 types of textile defect detection and classification results are output. The SVM classifier is adopted to replace the traditional softmax classifier, so that a better classification result can be obtained.
The pooling layers in this embodiment all adopt the maximum pooling strategy, and their kernel _ size is 3 and stride is 2.
Further, in this embodiment, other conventional textile defect detection methods are adopted to perform the detection and classification of textile defects on the Tilda database, including:
(1)PCA+FFN(Rebhi,A.；Benmhammed,I.；Abid,S.；Fnaiech,F.Fabric Defect Detection Using Local Homogeneity Analysis and Neural Network.J.Photonics 2015,2015,1–9.)；
(2)MLP neural networks(Tabassian,M.；Ghaderi,R.；Ebrahimpour,R.Knitted Fabric Defect Classification for Uncertain Labels Based on Dempster–Shafer Theory of Evidence.Expert Syst.Appl.2011,38(5),5259–5267.)；
(3)Sparse coding(Zhu,Qiuping；Wu,Minyuan；Li,Jie；Deng,Dexiang.Fabric defect detection via small scale over-complete basis set.Textile Research Journal 2014,vol.84,no.15,pp.1634–1649.)；
(4)GoogLeNet；
(5)LeNet-5(Jing,J.；Ma,H.；Zhang,H.Automatic Fabric Defect Detection Using a Deep Convolutional Neural Network.Color.technol.2019,135(3),213–223.)；
(6)VGG16(Jing,J.；Ma,H.；Zhang,H.Automatic Fabric Defect Detection Using a Deep Convolutional Neural Network.Color.technol.2019,135(3),213–223.)；
(7)Alexnet。
as shown in table 1 below, the classification accuracy of detecting and classifying the Tilda data set by using the above-mentioned textile defect detecting method and the textile defect detecting and classifying method provided by the present invention is shown.
TABLE 1 Classification accuracy for detection classification of Tilda datasets
Wherein, Improved Alexnet represents that the enhanced Alexnet convolutional neural network is matched with a softmax classifier; improved Alexnet-SVM represents the enhanced Alexnet convolutional neural network of the present invention in cooperation with an SVM classifier.
As can be seen from the above table, compared with other conventional textile defect detection methods, the enhanced Alexnet convolutional neural network provided in this embodiment has a higher detection and classification accuracy, and particularly, when the enhanced Alexnet convolutional neural network provided by the present invention is used in cooperation with an SVM classifier to detect and classify textile defects, the classification accuracy is as high as 99%, compared with the conventional textile defect detection method, the algorithm complexity can be effectively simplified, and the detection and classification efficiency can be effectively improved. In specific application, the method for detecting and classifying textile defects, which is provided by the embodiment, is applied to an upper computer for automatically detecting the defects, and is matched with image acquisition equipment applied to textile production, so that the quality of the textile can be detected in real time, the defect detection and classification results can be output, and the automatic effect of the textile production can be greatly improved.
The same or similar reference numerals correspond to the same or similar parts;
the terms describing positional relationships in the drawings are for illustrative purposes only and are not to be construed as limiting the patent;
it should be understood that the above-described embodiments of the present invention are merely examples for clearly illustrating the present invention, and are not intended to limit the embodiments of the present invention. Other variations and modifications will be apparent to persons skilled in the art in light of the above description. And are neither required nor exhaustive of all embodiments. Any modification, equivalent replacement, and improvement made within the spirit and principle of the present invention should be included in the protection scope of the claims of the present invention.
Claims (10)
1. A method for detecting and classifying defects of textiles, which is characterized by comprising the following steps:
s1: collecting textile defect images to form a data set, and preprocessing the data set;
s2: dividing the preprocessed data set into a training set, a verification set and a test set;
s3: constructing an enhanced Alexnet convolutional neural network by utilizing a caffe deep learning framework; the enhanced Alexnet convolutional neural network comprises a convolutional layer, a ReLU activation function layer, a pooling layer, a filter layer, a full connection layer and a classification layer;
s4: inputting a training set into the enhanced Alexnet convolutional neural network for training, adjusting and storing parameters of the enhanced Alexnet convolutional neural network;
s5: inputting a verification set into the trained enhanced Alexnet convolutional neural network for verification, outputting a detection and classification result of the textile defects, and calculating the accuracy of the detection and classification result; if the accuracy is lower than the preset threshold, skipping to execute the step S4;
s6: and inputting the test set into the trained enhanced Alexnet convolutional neural network for testing, and outputting to obtain the detection and classification results of the textile defects.
2. A method of detecting and classifying textile defects according to claim 1 and wherein said defect types of said textile defect images include holes, stains, lint, wrinkles, normal.
3. A method for detecting and classifying textile defects according to claim 1, wherein said step of preprocessing said data set in step S1 comprises:
s1.1: performing data cleaning on the image;
s1.2: converting the image format into a jpg format;
s1.3: copying the value of a single channel in the image to the other two channels aiming at the image of the single channel to form a three-channel image;
s1.4: adjusting pixels of the image to 256 × 256;
s1.5: and carrying out image expansion on the image.
4. A method for detecting and classifying textile defects according to claim 3, wherein in said step S1.4, the step of image-expanding the image comprises: the method comprises the steps of rotating an image by 90 degrees, rotating the image by 180 degrees, inputting Gaussian noise and changing image chromatic aberration.
5. A method for detecting and classifying textile defects according to claim 1 and wherein said enhanced Alexnet convolutional neural network comprises a sequentially connected 8-layer structure in which:
the first layer comprises a convolution layer, a ReLU activation function layer and a pooling layer which are connected in sequence;
the second layer comprises a convolution layer with a filling edge set to 2;
the third layer comprises a convolution layer and a pooling layer which are connected in sequence;
the fourth layer comprises a convolution layer and a ReLU activation function layer which are connected in sequence;
the fifth layer comprises a convolution layer and a ReLU activation function layer which are connected in sequence;
the sixth layer comprises a pooling layer;
the seventh and eighth layers comprise fully connected layers with a core number of 2048.
6. A method of detecting and classifying textile defects according to claim 5 and wherein said enhanced Alexnet convolutional neural network has a convolutional kernel size of 9 x 9 for the first convolutional layer; the convolution layers of the second layer and the third layer are convolution layers with convolution kernel size of 3 x 3 using the same strategy, and the number of convolution kernels is 192.
7. A method of detecting and classifying textile defects according to claim 6 and wherein said eighth fully bonded ply has an output connected to an input of a softmax classifier.
8. A method of detecting and classifying textile defects according to claim 6 and wherein the output of said eighth fully-connected layer is connected to the input of an SVM classifier.
9. A method of detecting and classifying textile defects according to claim 7 or 8, characterized in that said fully bonded layer employs the Dropout strategy.
10. A method for detecting and classifying textile defects according to claim 1, wherein in said step S4, during the training of the enhanced Alexnet convolutional neural network, the initial learning rate is set to 0.001, the initial weight is randomly set, the bias is set to 0.1, the number of iterations is set to 10000, and the number of pictures per iteration is set to 150.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011477224.3A CN112465810A (en) | 2020-12-15 | 2020-12-15 | Method for detecting and classifying defects of textiles |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011477224.3A CN112465810A (en) | 2020-12-15 | 2020-12-15 | Method for detecting and classifying defects of textiles |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112465810A true CN112465810A (en) | 2021-03-09 |
Family
ID=74802873
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011477224.3A Pending CN112465810A (en) | 2020-12-15 | 2020-12-15 | Method for detecting and classifying defects of textiles |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112465810A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113139577A (en) * | 2021-03-22 | 2021-07-20 | 广东省科学院智能制造研究所 | Deep learning image classification method and system based on deformable convolution network |
| CN115240144A (en) * | 2022-09-21 | 2022-10-25 | 青岛宏大纺织机械有限责任公司 | Method and system for intelligently identifying flaws in spinning twisting |
Citations (3)
| 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 |
| CN107169956A (en) * | 2017-04-28 | 2017-09-15 | 西安工程大学 | Yarn dyed fabric defect detection method based on convolutional neural networks |
| 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 |
-
2020
- 2020-12-15 CN CN202011477224.3A patent/CN112465810A/en active Pending
Patent Citations (3)
| 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 |
| CN107169956A (en) * | 2017-04-28 | 2017-09-15 | 西安工程大学 | Yarn dyed fabric defect detection method based on convolutional neural networks |
| 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 |
Non-Patent Citations (2)
| Title |
|---|
| WENXIAN ZENG ET AL: "Design of Intelligent Classroom Attendance System Based on Face Recognition", 《2019 IEEE 3RD INFORMATION TECHNOLOGY,NETWORKING,ELECTRONIC AND AUTOMATION CONTROL CONFERENCE》 * |
| 景军锋 等: "基于卷积神经网络的织物表面缺陷分类方法", 《测控技术》 * |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113139577A (en) * | 2021-03-22 | 2021-07-20 | 广东省科学院智能制造研究所 | Deep learning image classification method and system based on deformable convolution network |
| CN113139577B (en) * | 2021-03-22 | 2024-02-23 | 广东省科学院智能制造研究所 | Deep learning image classification method and system based on deformable convolution network |
| CN115240144A (en) * | 2022-09-21 | 2022-10-25 | 青岛宏大纺织机械有限责任公司 | Method and system for intelligently identifying flaws in spinning twisting |
| CN115240144B (en) * | 2022-09-21 | 2022-12-27 | 青岛宏大纺织机械有限责任公司 | Method and system for intelligently identifying flaws in spinning twisting |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| WO2021253632A1 (en) | Cloth defect detection method based on adversarial neural network, and terminal and storage medium | |
| Rasheed et al. | Fabric defect detection using computer vision techniques: a comprehensive review | |
| CN107341499B (en) | Fabric defect detection and classification method based on unsupervised segmentation and ELM | |
| CN110619618B (en) | Surface defect detection method and device and electronic equipment | |
| Li et al. | Deformable patterned fabric defect detection with fisher criterion-based deep learning | |
| CN108345818B (en) | Face living body detection method and device | |
| CN105678788B (en) | A kind of fabric defect detection method based on HOG and low-rank decomposition | |
| CN109766858A (en) | Three-dimensional convolution neural network hyperspectral image classification method combined with bilateral filtering | |
| CN104778692B (en) | A kind of fabric defect detection method optimized based on rarefaction representation coefficient | |
| CN110889838A (en) | Fabric defect detection method and device | |
| CN112950561B (en) | Optical fiber end face defect detection method, device and storage medium | |
| CN109543760A (en) | Confrontation sample testing method based on image filters algorithm | |
| CN112465810A (en) | Method for detecting and classifying defects of textiles | |
| CN109712113A (en) | A kind of fabric defect detection method based on cascade low-rank decomposition | |
| Drozdowski et al. | Detection of glasses in near-infrared ocular images | |
| CN107240086A (en) | A kind of fabric defects detection method based on integration nomography | |
| An et al. | Fabric defect detection using deep learning: An Improved Faster R-approach | |
| KR20230147130A (en) | Methods and apparatus for ranking images in a collection using image segmentation and image analysis | |
| Huang et al. | A method for identifying origin of digital images using a convolutional neural network | |
| Ando et al. | Improvements of classification accuracy of film defects by using gpu-accelerated image processing and machine learning frameworks | |
| CN113888459A (en) | Texture image defect detection system and method based on self-adaptive texture features | |
| KR102178238B1 (en) | Apparatus and method of defect classification using rotating kernel based on machine-learning | |
| CN201041547Y (en) | Device for analyzing and recognizing different types of green teas based on multi-spectrum image texture | |
| CN107180419A (en) | A kind of medium filtering detection method based on PCA networks | |
| Zhou et al. | JCS: An explainable surface defects detection method for steel sheet by joint classification and segmentation |
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: 20210309 |
|
| RJ01 | Rejection of invention patent application after publication |