CN113554080A - Non-woven fabric defect detection and classification method and system based on machine vision - Google Patents

Non-woven fabric defect detection and classification method and system based on machine vision Download PDF

Info

Publication number
CN113554080A
CN113554080A CN202110801896.3A CN202110801896A CN113554080A CN 113554080 A CN113554080 A CN 113554080A CN 202110801896 A CN202110801896 A CN 202110801896A CN 113554080 A CN113554080 A CN 113554080A
Authority
CN
China
Prior art keywords
image
woven fabric
defect
calculating
area
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
Application number
CN202110801896.3A
Other languages
Chinese (zh)
Inventor
尹文芳
谭艾琳
郭东妮
向泽军
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Changsha Chaint Robotics Co Ltd
Original Assignee
Changsha Chaint Robotics Co Ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Changsha Chaint Robotics Co Ltd filed Critical Changsha Chaint Robotics Co Ltd
Priority to CN202110801896.3A priority Critical patent/CN113554080A/en
Publication of CN113554080A publication Critical patent/CN113554080A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • 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
    • 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
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T5/00—Image enhancement or restoration
    • G06T5/70—Denoising; Smoothing
    • 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
    • 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
    • G01N2021/8854—Grading and classifying of flaws
    • 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
    • G01N2021/8887—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 based on image processing techniques
    • 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
    • G06T2207/30—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)
  • General Physics & Mathematics (AREA)
  • Physics & Mathematics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Data Mining & Analysis (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Evolutionary Biology (AREA)
  • Evolutionary Computation (AREA)
  • Artificial Intelligence (AREA)
  • General Engineering & Computer Science (AREA)
  • Signal Processing (AREA)
  • Health & Medical Sciences (AREA)
  • Chemical & Material Sciences (AREA)
  • Analytical Chemistry (AREA)
  • Biochemistry (AREA)
  • General Health & Medical Sciences (AREA)
  • Immunology (AREA)
  • Pathology (AREA)
  • Quality & Reliability (AREA)
  • Image Analysis (AREA)
  • Investigating Materials By The Use Of Optical Means Adapted For Particular Applications (AREA)

Abstract

The invention relates to a non-woven fabric flaw detection and classification method and system based on machine vision, wherein the method comprises the following steps: (1) collecting an image; (2) preprocessing an image; (3) extracting an image detail signal: analyzing and extracting detail signals of the non-woven fabric defect image by adopting wavelet transform; (4) calculating a texture feature image and reconstructing the feature image: calculating texture information of the detail signal of the image to obtain a characteristic image with the capability of distinguishing a normal image and a defect image of the non-woven fabric, reconstructing the characteristic image and obtaining a reconstructed characteristic image; (5) calculating an anomaly score: calculating the abnormal score of the non-woven fabric characteristic image so as to judge whether the flaw appears; (6) and (3) calculating a defect area by difference: carrying out differential operation on an original image to be detected and a reconstructed characteristic image to obtain a flaw area, and carrying out post-processing on the flaw to obtain a complete flaw area; (7) classifying defective areas: and constructing a feature vector and training a classifier to classify the defective region.

Description

Non-woven fabric defect detection and classification method and system based on machine vision
Technical Field
The invention relates to the field of machine vision detection, in particular to a non-woven fabric flaw detection and classification method and system based on machine vision.
Background
With the continuous expansion of the market of non-woven fabrics in the global field, the research and development of non-woven fabrics has never stopped, and new varieties emerge endlessly, so the application in each field is more and more extensive: disposable surgical masks, protective clothing and the like used in the medical industry, waterproof coiled materials used in the building industry, industrial filtering materials and the like. However, the production of quality nonwovens requires strict control of the procedure and quality of each process to ensure that good nonwoven articles are ultimately produced.
At present, most factories still adopt a mode of manually seeing the non-woven fabric to check whether flaws exist on the surface of the non-woven fabric, and one worker needs to continuously check the surface of the non-woven fabric with a width of more than 4m for 8 hours under the condition of insufficient light. Firstly, the attention of workers is concentrated for 30 minutes, and most of non-woven fabric defects are consistent with the color of the cloth cover, so that the defects cannot be distinguished by human eyes. In addition, the speed of cloth viewing with naked eyes is low. Therefore, the manual cloth inspection has low accuracy and low speed, and defects and missing inspection are easy to occur to cause defective products to flow out. Therefore, adopt machine vision system to detect the flaw on non-woven fabrics surface automatically, can realize higher rate of accuracy, guarantee on-line measuring, can practice thrift the cost of labor for the enterprise simultaneously to strict controlling the non-woven fabrics quality.
The prior non-woven fabric defect detection methods are as follows:
1. based on the gray level binarization method, because the non-woven fabric imaging has the condition of uneven brightness, the processing adaptability of simply carrying out binarization on the image is very poor, and the situation that the flaws can not be generated by complete segmentation can occur;
2. based on a template matching method, a template is established for the defects of the existing non-woven fabric, so that the defects can be identified by a system, but the established template is invalid usually due to the large variation range of various defects of the non-woven fabric, and the applicability in the industry is poor;
3. the non-woven fabric defect positioning method based on Gabor filtering extracts the cloth surface characteristics by using a plurality of Gabor filtering, and has the advantages of large calculated amount, good accuracy and poor real-time performance.
In view of the foregoing, it is desirable to provide a method and a system for detecting and classifying defects of a non-woven fabric based on machine vision, which can effectively distinguish normal images from defect images of the non-woven fabric, and have high accuracy and small computation workload.
Disclosure of Invention
The invention aims to provide a non-woven fabric flaw detection and classification method and system based on machine vision, which can effectively distinguish a non-woven fabric normal image and a flaw image, and has high accuracy and small calculation amount.
The above purpose is realized by the following technical scheme: a non-woven fabric flaw detection and classification method based on machine vision comprises the following steps:
(1) image acquisition: collecting a non-woven fabric image;
(2) preprocessing an image;
(3) extracting an image detail signal: analyzing and extracting detail signals of the non-woven fabric defect image by adopting wavelet transform;
(4) calculating a texture feature image and reconstructing the feature image: calculating texture information of the detail signals of the image processed in the step (3) to obtain a characteristic image capable of distinguishing a normal image and a defect image of the non-woven fabric, and reconstructing the characteristic image by learning the data distribution of the characteristic image of the non-woven fabric to obtain a reconstructed characteristic image;
(5) calculating an anomaly score: calculating the abnormal score of the non-woven fabric characteristic image in the step (4) so as to judge whether the flaw occurs;
(6) and (3) calculating a defect area by difference: carrying out differential operation on an original image to be detected and a reconstructed characteristic image to obtain a flaw area, and carrying out post-processing on the flaw to obtain a complete flaw area;
(7) classifying defective areas: and (5) constructing a feature vector and training a classifier to classify the flaw area obtained in the step (6).
The invention provides a method for unsupervised defect detection and classification of various non-woven fabrics, which mainly solves the detection and classification problems of seven defects of non-woven fabric yarn breakage, net turning, molten drop, glue spot, wrinkle, hole breaking and foreign matter, thereby realizing unmanned fabric inspection and ensuring the quality of non-woven fabric products. When the method is applied specifically, collected non-woven fabric images are preprocessed, detail signals of non-woven fabric defect images are extracted, abnormal image scores are calculated, whether defects exist or not is judged, a defect area is obtained by difference of an original image and a reconstructed feature image, a classifier is trained by using feature vectors of the defect images, an optimal parameter combination is searched through experiments, and the trained classifier is used for classifying the defects of the non-woven fabrics.
Preferably, a further technical solution is that, in the step (3), wavelet progressive decomposition is performed on the preprocessed nonwoven fabric image by using wavelets, when each layer of decomposition is performed, the smooth signal is decomposed into a high-frequency signal and a low-frequency signal, and each layer of decomposition generates details in horizontal, vertical and diagonal directions, and decomposes corresponding components from the horizontal, vertical and diagonal directions, and selects the component with the most obvious flaw for superposition.
A very important problem in wavelet analysis is the selection of the optimal wavelet basis. The wavelet is an orthogonal wavelet function with tight support, has a simpler structure, is a single rectangular wave with a support domain in a range, has the characteristic of simple calculation, is not only orthogonal, but also orthogonal with the integral displacement of the wavelet. When the method is applied, the image is subjected to wavelet progressive decomposition from multiple directions, and when the image is decomposed, a source signal is marked as O and H is usednRepresenting the high-frequency signal decomposed by the n-th layer by LnRepresenting the low-frequency signal obtained by the n-th layer decomposition, and when the first layer is decomposed, O ═ H1+L1When the second layer is decomposed L1＝H2+L2By analogy, each layer of decomposition will produce details in the horizontal, vertical and diagonal directions.
Preferably, a further technical scheme is that texture filtering is adopted in the step (4) to calculate texture information, wherein convolution operation is performed on the image to be detected by using different specific masks to extract required texture information, the extracted high-frequency detail information is calculated on key edge information of the flaw by using multi-dimension texture filtering, and the optimal texture filtering is selected according to the extraction effect to highlight the flaw characteristic.
Thus, the features of the normal image texture and the flaw image texture of the non-woven fabric are extracted from multiple dimensions, and the obtained wavelet decomposition reconstruction image is filtered by a texture method of multiple directions and dimensions, such as: extracting corresponding texture information of the low frequency, the edge, the point, the ripple and the like, and selecting the optimal texture filtering to highlight the flaw characteristics according to the extraction effect.
Preferably, in the step (4), the non-woven fabric image is processed by using a mask having a size of 5 × 5, and the texture analysis mask for each dimension of 5 × 5 is as follows:
l＝[a 4a 6a 4a a]
e＝[-a -2a 0 2a a]
s＝[-a 0 2a 0 -a]
w＝[-a 2a 0 -2a a]
r＝[a -4a 6a -4a a]
wherein, the mask l is level and is used for calculating the relative low-frequency details in the image by convolution; the mask e means edge and is used for calculating the edge details of the image by convolution; the mask s is spot and is used for calculating the detail of the spot class in the image by convolution; the mask w means wave and is used for calculating the waveform details of the image by convolution; the mask r is ripple and is used for calculating image ripple details through convolution, a is a factor base number and is used for adjusting the mask strength, a is a positive integer, the mask is used in a combined mode, a plurality of masks and an image to be detected are synthesized after convolution operation is carried out on the masks, a multi-channel image is formed, reconstruction of a characteristic image is completed, and the convolution operation is as follows:
Figure BDA0003164914550000041
wherein, IfilteredIs a feature image filtered by a texture method, I is an image detail signal, Filter is a texture filtering mask in the horizontal direction, the vertical direction and the diagonal direction,
Figure BDA0003164914550000042
the representative image is convolved with a filter mask. Masks are used in combination, such as 'el', 'ee', 'se', etc., with vertical weighting within brackets to the factor on each pixel. Texture filtering is an important algorithm for texture analysis, and the principle is to utilize different specific masks to perform convolution operation on an image to be detected so as to extract required information such as low frequency and edges. And calculating the extracted high-frequency detail information by using a multi-dimensional texture method for filtering the key edge information of the flaw, wherein a multi-channel image formed after convolution operation is the reconstructed characteristic image.
Preferably, in the step (5), an anomaly score is constructed by an auto-encoder, and an anomaly score of the nonwoven fabric feature image is calculated from the original feature image and the reconstructed feature image, wherein the anomaly score calculation formula is as follows:
Figure BDA0003164914550000043
in the above formula, x is the feature image, z is the hidden variable mapped from the feature image by the self-encoder, and ziThe computation of the anomaly score consists of KL divergence and reconstruction error, subject to a distribution q (z | x), where KL (q (z | x) | p (x)) is KL divergence, used to measure the quantitative difference computation between the original distribution p (x) of the feature image and the distribution q (z | x) of the variables reconstructed by the encoder,
Figure BDA0003164914550000044
setting an abnormal threshold value alpha for a reconstruction error after the original characteristic image is coded by a self-coder, judging that the image to be detected has a defect when an abnormal score S (x) is larger than alpha, obtaining a defect area by performing differential operation on the original image to be detected and the reconstructed characteristic image in the step (6), and obtaining a complete defect area by performing post-processing on the defect. The abnormal score of the image is calculated, and the basic idea here is to compare the distribution of the normal image with the distribution of the sample to be detected.
Preferably, a further technical solution is that, in the step (6), it is checked whether there is a trained gaussian mixture classifier model, and if there is a trained gaussian mixture classifier model, the obtained image of the defective region is input into a classifier for classification; if not, starting to train the classifier: calculating a plurality of characteristics of the obtained image of the defect area and forming a characteristic vector group, inputting each defect type and the corresponding characteristic vector group into a Gaussian mixture classifier for training to obtain the corresponding defect type, then performing iterative optimization on parameters to obtain a trained Gaussian mixture classifier model, and then returning to input the obtained image of the defect area into the classifier for classification.
The method comprises the steps of extracting features of normal image textures and flaw image textures of the non-woven fabric from multiple dimensions by adopting texture filtering, carrying out abnormal detection on a feature map, reconstructing data to be detected by learning the distribution of normal data, and carrying out differential operation on the image to be detected and a reconstructed image to obtain a flaw area; and (3) using a parameter matrix formed by the contrast, entropy, autocorrelation coefficient and inverse difference moment of seven defect images as a defect characteristic vector, training a Gaussian mixture classifier, searching an optimal parameter combination through experiments, and classifying seven non-woven fabric defects by the trained Gaussian mixture classifier.
Preferably, a further technical solution is that the feature vector group includes region basic features, region shape features, and gray texture features, the region basic features at least include an area, a roundness, and a maximum diameter, the region shape features at least include a small circumscribed circle radius, a minimum circumscribed rectangle width and a connectivity, and the gray texture features at least include an average value, a standard deviation, an entropy, and an anisotropy.
Preferably, a further technical solution is that the step (2) includes image cropping, gray level correction, and image denoising, and the specific steps are as follows:
(2.1) image cropping: selecting the cloth surface part of the whole obtained image by a binarization method for cutting, detecting the edge of the cloth surface by an edge detection method, setting the corresponding edge cutting width of the cloth surface edge cutting position according to the actual production condition, and selecting an interested area by taking the width as a standard;
(2.2) gradation correction: separating bright and dark areas of the cut interesting area, carrying out histogram equalization on the bright area and the dark area to respectively obtain a histogram L and a histogram D, carrying out histogram equalization operation on the whole interesting area to obtain a histogram A, applying histogram conversion from the histogram L to the histogram A to the bright area, and applying histogram conversion from the histogram D to the histogram A to the darker area;
and (2.3) transforming the image after gray correction to a frequency domain through Fourier transformation, zeroing the gray of the position of the interference striation in the frequency domain, and transforming the interference striation back to a time domain through inverse Fourier transformation to obtain the image after de-noising.
And due to the fact that the digital light source controller and the trigger frequency of the camera cause interference horizontal stripes accidentally caused by collision, the non-woven fabric backlight image is transformed to a frequency domain through Fourier transformation, the gray level of the position of the interference horizontal stripes in the frequency domain is set to be zero, and the interference horizontal stripes are transformed back to a time domain through inverse Fourier transformation to obtain a denoised image. The method collects the non-woven fabric backlight image, cuts the region of interest, corrects the brightness unevenness of the non-woven fabric backlight image and eliminates interference horizontal stripes from a frequency domain. When the non-woven fabric defect image is preprocessed, smoothing and denoising can be performed by using average filtering, bilateral filtering, Gaussian filtering and the like.
In order to achieve the purpose, the invention also provides a non-woven fabric flaw detection and classification system based on machine vision, which realizes non-woven fabric flaw detection and classification through any non-woven fabric flaw detection and classification method based on machine vision.
Preferably, the industrial personal computer comprises a memory and a processor, wherein a program is stored in the memory, and when the processor calls the program in the memory, the provided flaw detection and classification method can be realized.
Preferably, the non-woven fabric defect detecting and classifying system based on machine vision further comprises a mechanical support, a light source controller and a linear array lens, wherein the linear array lens is arranged on a camera to form an integral structure, the camera is connected with an industrial personal computer through a network cable, the light source controller is connected with a light source through a control line and used for controlling the brightness of the light source, and the camera and the light source are arranged on the mechanical support.
When the system is used, the mechanical support is arranged on a non-woven fabric production machine, the camera is preferably an industrial linear array camera, the light source is arranged at a position 10mm below the non-woven fabric, the camera is integrally arranged at a position 2.5m above the non-woven fabric, and the light source controller is connected with the light source through a light source control line to control the brightness of the light source.
The method comprises the steps of preprocessing a non-woven fabric image, decomposing high-frequency and low-frequency components of the image from a frequency domain, carrying out texture filtering on the high-frequency component, calculating an image abnormal score by using a filtering result, judging whether a flaw occurs, differentiating an original image and a reconstructed image to obtain a flaw area, using a parameter matrix formed by the contrast, entropy, autocorrelation coefficient and inverse difference moment of seven kinds of flaw images as a flaw characteristic vector, training a GMM classifier, searching an optimal parameter combination through experiments, and classifying seven kinds of non-woven fabric flaws by the trained GMM classifier. Compared with the prior art, the method extracts the non-woven fabric high-frequency component image by wavelet decomposition, greatly reduces redundant information of the whole image, highlights the flaw characteristic by texture filtering, can effectively distinguish the non-woven fabric normal image from the flaw image, and has small computation amount; seven defects of the non-woven fabric are classified by using a Gaussian mixture classifier, and a Gaussian mixture model is commonly used for unsupervised clustering tasks, so that a good classification effect can be achieved. Under the condition that positive and negative samples of the non-woven fabric image data are quite unbalanced, the distribution of normal data is learned by adopting an abnormal detection method, the data to be detected is reconstructed, whether flaws exist is judged by calculating abnormal scores, the defects are segmented by using the image to be detected and the reconstructed image through differential operation, the non-supervised detection of the non-woven fabric defects is realized, and the accuracy is high.
Drawings
The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate an embodiment of the invention and, together with the description, serve to explain the invention and not to limit the invention.
FIG. 1 is a schematic flow chart of a method for detecting and classifying defects of a non-woven fabric based on machine vision according to an embodiment of the present invention;
fig. 2 is a result diagram of defect detection classification according to an embodiment of the present invention, in which diagram (a) is a net-turning detection result, diagram (b) is a broken filament detection result, diagram (c) is a wrinkle detection result, diagram (d) is a droplet detection result, diagram (e) is a spot detection result, diagram (f) is an undrawn filament (hole) detection result, diagram (g) is a foreign object detection result, and the detection result is that a defect position is marked with red to display a detected defect area.
Detailed Description
The present invention will now be described in detail with reference to the drawings, which are given by way of illustration and explanation only and should not be construed to limit the scope of the present invention in any way. Furthermore, features from embodiments in this document and from different embodiments may be combined accordingly by a person skilled in the art from the description in this document.
The embodiment of the invention provides a non-woven fabric flaw detection and classification method based on machine vision, which comprises the following steps of:
(1) image acquisition: collecting a non-woven fabric image;
(2) preprocessing an image;
(3) extracting an image detail signal: analyzing and extracting detail signals of the non-woven fabric defect image by adopting wavelet transform;
(4) calculating a texture feature image and reconstructing the feature image: calculating texture information of the detail signals of the image processed in the step (3) to obtain a characteristic image capable of distinguishing a normal image and a defect image of the non-woven fabric, and reconstructing the characteristic image by learning the data distribution of the characteristic image of the non-woven fabric to obtain a reconstructed characteristic image;
(5) calculating an anomaly score: calculating the abnormal score of the non-woven fabric characteristic image in the step (4) so as to judge whether the flaw occurs;
(6) and (3) calculating a defect area by difference: carrying out differential operation on an original image to be detected and a reconstructed characteristic image to obtain a flaw area, and carrying out post-processing on the flaw to obtain a complete flaw area;
(7) classifying defective areas: and (5) constructing a feature vector and training a classifier to classify the flaw area obtained in the step (6).
The invention provides a method for unsupervised defect detection and classification of various non-woven fabrics, as shown in figure 2, which mainly solves the detection and classification problems of seven defects of non-woven fabric yarn breakage, net turning, molten drop, glue spot, wrinkle, hole breaking and foreign matter, thereby realizing unmanned fabric inspection and ensuring the quality of non-woven fabric products. When the method is applied specifically, collected non-woven fabric images are preprocessed, detail signals of non-woven fabric defect images are extracted, abnormal image scores are calculated, whether defects exist or not is judged, a defect area is obtained by difference of an original image and a reconstructed feature image, a classifier is trained by using feature vectors of the defect images, an optimal parameter combination is searched through experiments, and the trained classifier is used for classifying the defects of the non-woven fabrics.
On the basis of the above embodiment, in another embodiment of the present invention, as shown in fig. 1, in the step (3), wavelet decomposition is performed on the preprocessed nonwoven fabric image by using wavelet, when each layer of decomposition is performed, the smooth signal is decomposed into a high frequency signal and a low frequency signal, and each layer of decomposition generates details in horizontal, vertical and diagonal directions, and decomposes out corresponding components from the horizontal, vertical and diagonal directions, and selects the component with the most obvious flaw for superposition.
A very important problem in wavelet analysis is the selection of the optimal wavelet basis. The wavelet is an orthogonal wavelet function with tight support, the structure is simple, the wavelet is a single rectangular wave with a support domain in a range, the calculation is simple, and the orthogonal wavelet is orthogonalAnd orthogonal to its integer displacement. When the method is applied, the image is subjected to wavelet progressive decomposition from multiple directions, and when the image is decomposed, a source signal is marked as O and H is usednRepresenting the high-frequency signal decomposed by the n-th layer by LnRepresenting the low-frequency signal obtained by the n-th layer decomposition, and when the first layer is decomposed, O ═ H1+L1When the second layer is decomposed L1＝H2+L2By analogy, each layer of decomposition will produce details in the horizontal, vertical and diagonal directions.
On the basis of the above embodiment, in another embodiment of the present invention, as shown in fig. 1, texture filtering is adopted in the step (4) to calculate texture information, wherein convolution operation is performed on the image to be detected by using different specific masks, so as to extract the required texture information, the extracted high-frequency detail information is calculated by using multi-dimensional texture filtering on key edge information of the defect, and an optimal texture filtering is selected to highlight the defect feature according to the extraction effect.
Thus, the features of the normal image texture and the flaw image texture of the non-woven fabric are extracted from multiple dimensions, and the obtained wavelet decomposition reconstruction image is filtered by a texture method of multiple directions and dimensions, such as: extracting corresponding texture information of the low frequency, the edge, the point, the ripple and the like, and selecting the optimal texture filtering to highlight the flaw characteristics according to the extraction effect.
In another embodiment of the present invention based on the above embodiment, in the step (4), the non-woven fabric image is processed by using a mask with a size of 5 × 5, and the texture analysis mask for each dimension of 5 × 5 is as follows:
l＝[a 4a 6a 4a a]
e＝[-a -2a 0 2a a]
s＝[-a 0 2a 0 -a]
w＝[-a 2a 0 -2a a]
r＝[a -4a 6a -4a a]
wherein, the mask l is level and is used for calculating the relative low-frequency details in the image by convolution; the mask e means edge and is used for calculating the edge details of the image by convolution; the mask s is spot and is used for calculating the detail of the spot class in the image by convolution; the mask w means wave and is used for calculating the waveform details of the image by convolution; the mask r is ripple and is used for calculating image ripple details through convolution, a is a factor base number and is used for adjusting the mask strength, a is a positive integer, the mask is used in a combined mode, a plurality of masks and an image to be detected are synthesized after convolution operation is carried out on the masks, a multi-channel image is formed, reconstruction of a characteristic image is completed, and the convolution operation is as follows:
Figure BDA0003164914550000101
wherein, IfilteredIs a feature image filtered by a texture method, I is an image detail signal, Filter is a texture filtering mask in the horizontal direction, the vertical direction and the diagonal direction,
Figure BDA0003164914550000102
the representative image is convolved with a filter mask. Masks are used in combination, e.g. 'el', 'ee', 'se', etc., and the values in brackets are factors that are weighted to each pixel. Texture filtering is an important algorithm for texture analysis, and the principle is to utilize different specific masks to perform convolution operation on an image to be detected so as to extract required information such as low frequency and edges. And calculating the extracted high-frequency detail information by using a multi-dimensional texture method for filtering the key edge information of the flaw, wherein a multi-channel image formed after convolution operation is the reconstructed characteristic image.
In another embodiment of the present invention, in addition to the above embodiment, in the step (5), an anomaly score is constructed by a self-encoder, and the anomaly score of the nonwoven fabric feature image is calculated from the original feature image and the reconstructed feature image, where the anomaly score calculation formula is:
Figure BDA0003164914550000103
in the above formula, x is the feature image, z is the hidden variable mapped from the feature image by the self-encoder, and ziObedience distribution q (z | x), xThe calculation of the constant fraction is composed of KL divergence and reconstruction error, wherein KL (q (z | x) | p (x)) is KL divergence and is used for measuring the quantitative difference calculation between the original distribution p (x) of the characteristic image and the variable distribution q (z | x) reconstructed by the encoder,
Figure BDA0003164914550000104
setting an abnormal threshold value alpha for a reconstruction error after the original characteristic image is coded by a self-coder, judging that the image to be detected has a defect when an abnormal score S (x) is larger than alpha, obtaining a defect area by performing differential operation on the original image to be detected and the reconstructed characteristic image in the step (6), and obtaining a complete defect area by performing post-processing on the defect. The abnormal score of the image is calculated, and the basic idea here is to compare the distribution of the normal image with the distribution of the sample to be detected.
On the basis of the above embodiment, in another embodiment of the present invention, as shown in fig. 1, in the step (6), it is checked whether there is a trained gaussian mixture classifier model, and if there is, the obtained image of the defective region is input into a classifier for classification; if not, starting to train the classifier: calculating a plurality of characteristics of the obtained image of the defect area and forming a characteristic vector group, inputting each defect type and the corresponding characteristic vector group into a Gaussian mixture classifier for training to obtain the corresponding defect type, then performing iterative optimization on parameters to obtain a trained Gaussian mixture classifier model, and then inputting the obtained image of the defect area into the trained classifier for classification.
The method comprises the steps of extracting features of normal image textures and flaw image textures of the non-woven fabric from multiple dimensions by adopting texture filtering, carrying out abnormal detection on a feature map, reconstructing data to be detected by learning the distribution of normal data, and carrying out differential operation on the image to be detected and a reconstructed image to obtain a flaw area; and (3) using a parameter matrix formed by the contrast, entropy, autocorrelation coefficient and inverse difference moment of seven defect images as a defect characteristic vector, training a Gaussian mixture classifier, searching an optimal parameter combination through experiments, and classifying seven non-woven fabric defects by the trained Gaussian mixture classifier.
On the basis of the above embodiment, in another embodiment of the present invention, as shown in fig. 1, the feature vector group includes region basic features, region shape features and gray texture features, the region basic features at least include an area, a roundness and a maximum diameter, the region shape features at least include a small circumscribed circle radius, a minimum circumscribed rectangle width and a connectivity, and the gray texture features at least include an average value, a standard deviation, an entropy and an anisotropy.
On the basis of the above embodiment, in another embodiment of the present invention, as shown in fig. 1, the step (2) includes image cropping, gray level correction, and image denoising, and the specific steps are as follows:
(2.1) image cropping: selecting the cloth surface part of the whole obtained image by a binarization method for cutting, detecting the edge of the cloth surface by an edge detection method, setting the corresponding edge cutting width of the cloth surface edge cutting position according to the actual production condition, and selecting an interested area by taking the width as a standard;
(2.2) gradation correction: separating bright and dark areas of the cut interesting area, carrying out histogram equalization on the bright area and the dark area to respectively obtain a histogram L and a histogram D, carrying out histogram equalization operation on the whole interesting area to obtain a histogram A, applying histogram conversion from the histogram L to the histogram A to the bright area, and applying histogram conversion from the histogram D to the histogram A to the darker area;
and (2.3) transforming the image after gray correction to a frequency domain through Fourier transformation, zeroing the gray of the position of the interference striation in the frequency domain, and transforming the interference striation back to a time domain through inverse Fourier transformation to obtain the image after de-noising.
And due to the fact that the digital light source controller and the trigger frequency of the camera cause interference horizontal stripes accidentally caused by collision, the non-woven fabric backlight image is transformed to a frequency domain through Fourier transformation, the gray level of the position of the interference horizontal stripes in the frequency domain is set to be zero, and the interference horizontal stripes are transformed back to a time domain through inverse Fourier transformation to obtain a denoised image. The method collects the non-woven fabric backlight image, cuts the region of interest, corrects the brightness unevenness of the non-woven fabric backlight image and eliminates interference horizontal stripes from a frequency domain. When the non-woven fabric defect image is preprocessed, smoothing and denoising can be performed by using average filtering, bilateral filtering, Gaussian filtering and the like.
The invention also provides a non-woven fabric flaw detection and classification system based on machine vision, and the embodiment is that the non-woven fabric flaw detection and classification is realized through any one of the non-woven fabric flaw detection and classification methods based on machine vision.
Preferably, the industrial personal computer comprises a memory and a processor, wherein a program is stored in the memory, and when the processor calls the program in the memory, the provided flaw detection and classification method can be realized.
On the basis of the above embodiment, in another embodiment of the present invention, the non-woven fabric defect detecting and classifying system based on machine vision further includes a mechanical support, a light source controller and a linear array lens, the linear array lens is arranged on a camera to form an integral structure, the camera is connected with an industrial personal computer through a network cable, the light source controller is connected with a light source through a control line and is used for controlling the brightness of the light source, and the camera and the light source are arranged on the mechanical support.
When the system is used, the mechanical support is arranged on a non-woven fabric production machine, the camera is preferably an industrial linear array camera, the light source is arranged at a position 10mm below the non-woven fabric, the camera is integrally arranged at a position 2.5m above the non-woven fabric, and the light source controller is connected with the light source through a light source control line to control the brightness of the light source.
The method comprises the steps of preprocessing a non-woven fabric image, decomposing high-frequency and low-frequency components of the image from a frequency domain, carrying out texture filtering on the high-frequency component, calculating an image abnormal score by using a filtering result, judging whether a flaw occurs, differentiating an original image and a reconstructed image to obtain a flaw area, using a parameter matrix formed by the contrast, entropy, autocorrelation coefficient and inverse difference moment of seven kinds of flaw images as a flaw characteristic vector, training a GMM classifier, searching an optimal parameter combination through experiments, and classifying seven kinds of non-woven fabric flaws by the trained GMM classifier.
It will be apparent to those skilled in the art that various modifications and adaptations can be made without departing from the principles of the invention and these are intended to be within the scope of the invention.

Claims (10)

1. A non-woven fabric flaw detection and classification method based on machine vision is characterized by comprising the following steps:
(1) image acquisition: collecting a non-woven fabric image;
(2) preprocessing an image;
(3) extracting an image detail signal: analyzing and extracting detail signals of the non-woven fabric defect image by adopting wavelet transform;
(4) calculating a texture feature image and reconstructing the feature image: calculating texture information of the detail signals of the image processed in the step (3) to obtain a characteristic image capable of distinguishing a normal image and a defect image of the non-woven fabric, and reconstructing the characteristic image by learning the data distribution of the characteristic image of the non-woven fabric to obtain a reconstructed characteristic image;
(5) calculating an anomaly score: calculating the abnormal score of the non-woven fabric characteristic image in the step (4) so as to judge whether the flaw occurs;
(6) and (3) calculating a defect area by difference: carrying out differential operation on an original image to be detected and a reconstructed characteristic image to obtain a flaw area, and carrying out post-processing on the flaw to obtain a complete flaw area;
(7) classifying defective areas: and (5) constructing a feature vector and training a classifier to classify the flaw area obtained in the step (6).
2. The method according to claim 1, wherein the step (3) is implemented by performing wavelet decomposition on the preprocessed nonwoven fabric image, wherein the smooth signal is decomposed into a high frequency signal and a low frequency signal in each layer of decomposition, and each layer of decomposition generates horizontal, vertical and diagonal details, decomposes corresponding components from the horizontal, vertical and diagonal directions, and selects the most obvious component of the defects for superposition.
3. The method according to claim 2, wherein texture filtering is used to calculate texture information in step (4), wherein different specific masks are used to perform convolution operation with the image to be detected to extract the required texture information, the extracted high-frequency detail information is calculated by using multi-dimensional texture filtering to calculate key edge information of the defect, and the optimal texture filtering is selected to highlight the defect feature according to the extraction effect.
4. The method for detecting and classifying defects in non-woven fabrics based on machine vision according to claim 3, wherein in the step (4), a 5x5 mask is used for processing the non-woven fabric image, and the texture analysis mask of each dimension of 5x5 is as follows:
l＝[a 4a 6a 4a a]
e＝[-a -2a 0 2a a]
s＝[-a 0 2a 0 -a]
w＝[-a 2a 0 -2a a]
r＝[a -4a 6a -4a a]
the mask l is level and is used for calculating the relative low-frequency details in the image through convolution; the mask e is edge and is used for calculating the edge details of the image by convolution; the mask s is spot and is used for calculating the detail of the spot class in the image by convolution; the mask w is wave and is used for calculating the waveform details of the image by convolution; the mask r is ripple and is used for convolution calculation of image ripple details, a is a factor base number and is used for adjusting the mask strength, a is a positive integer, the masks are combined for use when in use, a plurality of masks and an image to be detected are synthesized after convolution operation is carried out on the masks, a multi-channel image is formed, reconstruction of a characteristic image is completed, and the convolution operation is as follows:
Figure FDA0003164914540000021
Ifilteredis a feature image filtered by a texture method, I is an image detail signal, Filter is a texture filtering mask in the horizontal direction, the vertical direction and the diagonal direction,
Figure FDA0003164914540000022
the representative image is convolved with a filter mask.
5. The method for detecting and classifying defects of non-woven fabric based on machine vision according to claim 3 or 4, wherein in the step (5), an anomaly score is constructed through a self-encoder, and the anomaly score of the non-woven fabric characteristic image is calculated according to the original characteristic image and the reconstructed characteristic image, and the anomaly score calculation formula is as follows:
Figure FDA0003164914540000023
where x is the feature image, z is the hidden variable mapped from the feature image with the self-encoder, ziThe computation of the anomaly score consists of KL divergence and reconstruction error, subject to a distribution q (z | x), where KL (q (z | x) | p (x)) is KL divergence, used to measure the quantitative difference computation between the original distribution p (x) of the feature image and the distribution q (z | x) of the variables reconstructed by the encoder,
Figure FDA0003164914540000031
setting an abnormal threshold value alpha for a reconstruction error after the original characteristic image is coded by a self-coder, judging that the image to be detected has a defect when an abnormal score S (x) is larger than alpha, obtaining a defect area by performing differential operation on the original image to be detected and the reconstructed characteristic image in the step (6), and obtaining a complete defect area by performing post-processing on the defect.
6. The method according to claim 5, wherein in the step (6), whether a trained Gaussian mixture classifier model exists is checked, and if so, the obtained image of the defect region is input into a classifier for classification; if not, starting to train the classifier: calculating a plurality of characteristics of the obtained image of the defect area and forming a characteristic vector group, inputting each defect type and the corresponding characteristic vector group into a Gaussian mixture classifier for training to obtain the corresponding defect type, then performing iterative optimization on parameters to obtain a trained Gaussian mixture classifier model, and then returning to input the obtained image of the defect area into the classifier for classification.
7. The method of claim 6, wherein the set of feature vectors comprises region basis features including at least an area, a roundness and a maximum diameter, region shape features including at least a small circumscribed circle radius, a minimum circumscribed rectangle width height and a connectivity, and gray texture features including at least an average value, a standard deviation, an entropy and an anisotropy.
8. The non-woven fabric defect detection and classification method based on machine vision according to any one of claims 1 to 4, characterized in that the step (2) comprises image cutting, gray level correction and image denoising, and the specific steps are as follows:
(2.1) image cropping: selecting the cloth surface part of the whole obtained image by a binarization method for cutting, detecting the edge of the cloth surface by an edge detection method, setting the corresponding edge cutting width of the cloth surface edge cutting position according to the actual production condition, and selecting an interested area by taking the width as a standard;
(2.2) gradation correction: separating bright and dark areas of the cut interesting area, carrying out histogram equalization on the bright area and the dark area to respectively obtain a histogram L and a histogram D, carrying out histogram equalization operation on the whole interesting area to obtain a histogram A, applying histogram conversion from the histogram L to the histogram A to the bright area, and applying histogram conversion from the histogram D to the histogram A to the darker area;
and (2.3) transforming the image after gray correction to a frequency domain through Fourier transformation, zeroing the gray of the position of the interference striation in the frequency domain, and transforming the interference striation back to a time domain through inverse Fourier transformation to obtain the image after de-noising.
9. The non-woven fabric flaw detection and classification system based on the machine vision is characterized in that the non-woven fabric flaw detection and classification is realized through the non-woven fabric flaw detection and classification method based on the machine vision, and the system comprises an industrial personal computer, a light source and a camera, wherein the industrial personal computer is in communication connection with the camera, the camera is used for collecting non-woven fabric images and transmitting the collected images to the industrial personal computer, and the industrial personal computer is used for image processing and calculation.
10. The non-woven fabric defect detecting and classifying system based on the machine vision as claimed in claim 9, further comprising a mechanical support, a light source controller and a linear array lens, wherein the linear array lens is arranged on a camera to form an integral structure, the camera is connected with an industrial personal computer through a network cable, the light source controller is connected with a light source through a control line and is used for controlling the brightness of the light source, and the camera and the light source are arranged on the mechanical support.
CN202110801896.3A 2021-07-15 2021-07-15 Non-woven fabric defect detection and classification method and system based on machine vision Pending CN113554080A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110801896.3A CN113554080A (en) 2021-07-15 2021-07-15 Non-woven fabric defect detection and classification method and system based on machine vision

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110801896.3A CN113554080A (en) 2021-07-15 2021-07-15 Non-woven fabric defect detection and classification method and system based on machine vision

Publications (1)

Publication Number Publication Date
CN113554080A true CN113554080A (en) 2021-10-26

Family

ID=78103243

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110801896.3A Pending CN113554080A (en) 2021-07-15 2021-07-15 Non-woven fabric defect detection and classification method and system based on machine vision

Country Status (1)

Country Link
CN (1) CN113554080A (en)

Cited By (11)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114345753A (en) * 2021-12-13 2022-04-15 贝塔科技(苏州)有限公司 Production method for connecting printing garment production raw materials of thermal sublimation process into Internet of things
CN114419039A (en) * 2022-03-28 2022-04-29 武汉市融科优品装饰材料有限公司 Decorative wallpaper defect detection method based on template matching
CN114411337A (en) * 2022-01-21 2022-04-29 山东晶创新材料科技有限公司 Preparation method of ultra-wide polypropylene filament non-woven fabric
CN114486903A (en) * 2021-12-06 2022-05-13 浙江大学台州研究院 Gray-scale self-adaptive coiled material detection system, device and algorithm
CN114757900A (en) * 2022-03-31 2022-07-15 启东新朋莱纺织科技有限公司 Artificial intelligence-based textile defect type identification method
CN115016394A (en) * 2022-07-18 2022-09-06 西安交通大学 Flaw cutter point identification method based on flaw type
CN115049651A (en) * 2022-08-15 2022-09-13 邹城市天晖软件科技有限公司 Metal plate stamping abnormity detection method
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
CN116485795A (en) * 2023-06-19 2023-07-25 湖南隆深氢能科技有限公司 Coil coating production line flaw detection method and system
CN117723564A (en) * 2024-02-18 2024-03-19 青岛华康塑料包装有限公司 Packaging bag printing quality detection method and system based on image transmission
CN118067741A (en) * 2024-04-18 2024-05-24 成都考拉悠然科技有限公司 Mini-LED glue surface defect detection method based on photometric stereo method

Citations (15)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103604809A (en) * 2013-10-22 2014-02-26 江南大学 Pattern cloth flaw online visual inspection method
CN104198498A (en) * 2014-09-12 2014-12-10 河海大学常州校区 Method and device for detecting cloth flaws based on adaptive orthogonal wavelet transform
CN104751472A (en) * 2015-04-10 2015-07-01 浙江工业大学 Fabric defect detection method based on B-spline wavelets and deep neural network
CN105548212A (en) * 2016-02-03 2016-05-04 杭州晶耐科光电技术有限公司 System and method for online automatically detecting defects on glass surface of touch screen
CN105701477A (en) * 2016-02-19 2016-06-22 中原工学院 Fabric defect detection method based on steady wavelet transform visual saliency
CN105718842A (en) * 2014-12-02 2016-06-29 中国科学院沈阳自动化研究所 Machine vision-based detection method for transmission line strand breakage fault
CN106373123A (en) * 2016-09-21 2017-02-01 哈尔滨工业大学 K_tSL central clustering algorithm-based industrial component surface defect detection method
CN106934801A (en) * 2017-03-01 2017-07-07 西南科技大学 A kind of fluorescentmagnetic particle(powder) defect inspection method based on Laws texture filterings
CN109308462A (en) * 2018-09-10 2019-02-05 中国民航大学 A kind of finger vein and phalangeal configurations area-of-interest localization method
CN109509171A (en) * 2018-09-20 2019-03-22 江苏理工学院 A kind of Fabric Defects Inspection detection method based on GMM and image pyramid
CN109934802A (en) * 2019-02-02 2019-06-25 浙江工业大学 A kind of Fabric Defects Inspection detection method based on Fourier transformation and morphological image
CN111257348A (en) * 2020-03-27 2020-06-09 河海大学常州校区 LED light guide plate defect detection method based on machine vision
CN111724372A (en) * 2020-06-19 2020-09-29 深圳新视智科技术有限公司 Method, terminal and storage medium for detecting cloth defects based on antagonistic neural network
CN111784644A (en) * 2020-06-11 2020-10-16 上海布眼人工智能科技有限公司 Printing defect detection method and system based on deep learning
CN112037166A (en) * 2020-07-10 2020-12-04 武汉迈格驷友科技有限公司 Surface defect detection method and detection device

Patent Citations (15)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103604809A (en) * 2013-10-22 2014-02-26 江南大学 Pattern cloth flaw online visual inspection method
CN104198498A (en) * 2014-09-12 2014-12-10 河海大学常州校区 Method and device for detecting cloth flaws based on adaptive orthogonal wavelet transform
CN105718842A (en) * 2014-12-02 2016-06-29 中国科学院沈阳自动化研究所 Machine vision-based detection method for transmission line strand breakage fault
CN104751472A (en) * 2015-04-10 2015-07-01 浙江工业大学 Fabric defect detection method based on B-spline wavelets and deep neural network
CN105548212A (en) * 2016-02-03 2016-05-04 杭州晶耐科光电技术有限公司 System and method for online automatically detecting defects on glass surface of touch screen
CN105701477A (en) * 2016-02-19 2016-06-22 中原工学院 Fabric defect detection method based on steady wavelet transform visual saliency
CN106373123A (en) * 2016-09-21 2017-02-01 哈尔滨工业大学 K_tSL central clustering algorithm-based industrial component surface defect detection method
CN106934801A (en) * 2017-03-01 2017-07-07 西南科技大学 A kind of fluorescentmagnetic particle(powder) defect inspection method based on Laws texture filterings
CN109308462A (en) * 2018-09-10 2019-02-05 中国民航大学 A kind of finger vein and phalangeal configurations area-of-interest localization method
CN109509171A (en) * 2018-09-20 2019-03-22 江苏理工学院 A kind of Fabric Defects Inspection detection method based on GMM and image pyramid
CN109934802A (en) * 2019-02-02 2019-06-25 浙江工业大学 A kind of Fabric Defects Inspection detection method based on Fourier transformation and morphological image
CN111257348A (en) * 2020-03-27 2020-06-09 河海大学常州校区 LED light guide plate defect detection method based on machine vision
CN111784644A (en) * 2020-06-11 2020-10-16 上海布眼人工智能科技有限公司 Printing defect detection method and system based on deep learning
CN111724372A (en) * 2020-06-19 2020-09-29 深圳新视智科技术有限公司 Method, terminal and storage medium for detecting cloth defects based on antagonistic neural network
CN112037166A (en) * 2020-07-10 2020-12-04 武汉迈格驷友科技有限公司 Surface defect detection method and detection device

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
崔玲玲 等: "基于非下采样Contourlet域高斯混合模型的布匹瑕疵识别算法", 《吉林大学学报（工学版）》, 31 May 2013 (2013-05-31), pages 734 - 739 *
程光耀 等: "《时空域视频运动片段分割理论及应用》", 北京理工大学出版社, pages: 134 - 33 *

Cited By (18)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114486903A (en) * 2021-12-06 2022-05-13 浙江大学台州研究院 Gray-scale self-adaptive coiled material detection system, device and algorithm
CN114486903B (en) * 2021-12-06 2024-05-14 浙江大学台州研究院 Gray-scale self-adaptive coiled material detection system, device and algorithm
CN114345753A (en) * 2021-12-13 2022-04-15 贝塔科技(苏州)有限公司 Production method for connecting printing garment production raw materials of thermal sublimation process into Internet of things
CN114345753B (en) * 2021-12-13 2024-05-10 贝塔科技(苏州)有限公司 Production method for connecting printing garment production raw materials into Internet of things through thermal sublimation technology
CN114411337B (en) * 2022-01-21 2023-03-14 山东晶创新材料科技有限公司 Preparation method of ultra-wide polypropylene filament non-woven fabric
CN114411337A (en) * 2022-01-21 2022-04-29 山东晶创新材料科技有限公司 Preparation method of ultra-wide polypropylene filament non-woven fabric
CN114419039A (en) * 2022-03-28 2022-04-29 武汉市融科优品装饰材料有限公司 Decorative wallpaper defect detection method based on template matching
CN114757900A (en) * 2022-03-31 2022-07-15 启东新朋莱纺织科技有限公司 Artificial intelligence-based textile defect type identification method
CN115016394A (en) * 2022-07-18 2022-09-06 西安交通大学 Flaw cutter point identification method based on flaw type
CN115049651B (en) * 2022-08-15 2022-10-28 邹城市天晖软件科技有限公司 Metal plate stamping abnormity detection method
CN115049651A (en) * 2022-08-15 2022-09-13 邹城市天晖软件科技有限公司 Metal plate stamping abnormity detection method
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
CN116485795A (en) * 2023-06-19 2023-07-25 湖南隆深氢能科技有限公司 Coil coating production line flaw detection method and system
CN116485795B (en) * 2023-06-19 2023-09-01 湖南隆深氢能科技有限公司 Coil coating production line flaw detection method and system
CN117723564A (en) * 2024-02-18 2024-03-19 青岛华康塑料包装有限公司 Packaging bag printing quality detection method and system based on image transmission
CN117723564B (en) * 2024-02-18 2024-04-26 青岛华康塑料包装有限公司 Packaging bag printing quality detection method and system based on image transmission
CN118067741A (en) * 2024-04-18 2024-05-24 成都考拉悠然科技有限公司 Mini-LED glue surface defect detection method based on photometric stereo method

Similar Documents

Publication Publication Date Title
CN113554080A (en) Non-woven fabric defect detection and classification method and system based on machine vision
CN114549522B (en) Textile quality detection method based on target detection
CN111179225B (en) Test paper surface texture defect detection method based on gray gradient clustering
Clausi et al. Design-based texture feature fusion using Gabor filters and co-occurrence probabilities
CN100401043C (en) Image inspecting method and device for detecting faults
CN108765402B (en) Non-woven fabric defect detection and classification method
CN110445921A (en) A kind of Mobile phone screen backlight foreign matter defect diagnostic method and device based on machine vision
CN113505865B (en) Sheet surface defect image recognition processing method based on convolutional neural network
CN107220649A (en) A kind of plain color cloth defects detection and sorting technique
CN106846313A (en) Surface Flaw Detection method and apparatus
CN102879401A (en) Method for automatically detecting and classifying textile flaws based on pattern recognition and image processing
CN109472788B (en) Method for detecting flaw on surface of airplane rivet
CN113837974B (en) NSST domain power equipment infrared image enhancement method based on improved BEEPS filtering algorithm
CN110889837A (en) Cloth flaw detection method with flaw classification function
CN109785301A (en) A kind of rail corrugation automatic identifying method and appraisal procedure based on image procossing
Chen et al. Evaluating fabric pilling with light-projected image analysis
CN110084787B (en) Synthetic leather gluing online monitoring method based on machine vision
WO2006113979A1 (en) Method for identifying guignardia citricarpa
Daniel et al. Automatic road distress detection and analysis
CN104392441A (en) Method for detecting and evaluating spray rating of high anti-noise fabric based on image processing
CN112669265B (en) Method for realizing surface defect detection based on Fourier transform and image gradient characteristics
CN116228651A (en) Cloth defect detection method, system, equipment and medium
Oni et al. Patterned fabric defect detection and classification (FDDC) techniques: a review
Laggoune et al. Tree ring analysis
CN113673396A (en) Spore germination rate calculation method and 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
RJ01 Rejection of invention patent application after publication

Application publication date: 20211026

RJ01 Rejection of invention patent application after publication