CN107870172A - A kind of Fabric Defects Inspection detection method based on image procossing - Google Patents
A kind of Fabric Defects Inspection detection method based on image procossing Download PDFInfo
- Publication number
- CN107870172A CN107870172A CN201710547024.2A CN201710547024A CN107870172A CN 107870172 A CN107870172 A CN 107870172A CN 201710547024 A CN201710547024 A CN 201710547024A CN 107870172 A CN107870172 A CN 107870172A
- Authority
- CN
- China
- Prior art keywords
- image
- fabric
- defect
- defects
- texture
- 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
- 239000004744 fabric Substances 0.000 title claims abstract description 177
- 230000007547 defect Effects 0.000 title claims abstract description 175
- 238000001514 detection method Methods 0.000 title claims abstract description 35
- 238000007689 inspection Methods 0.000 title abstract description 26
- 238000000034 method Methods 0.000 claims abstract description 41
- 238000012545 processing Methods 0.000 claims abstract description 37
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 35
- 238000001914 filtration Methods 0.000 claims abstract description 27
- 230000006870 function Effects 0.000 claims abstract description 17
- 238000000605 extraction Methods 0.000 claims abstract description 10
- 238000000354 decomposition reaction Methods 0.000 claims description 22
- 239000013598 vector Substances 0.000 claims description 18
- 238000005311 autocorrelation function Methods 0.000 claims description 16
- 210000002569 neuron Anatomy 0.000 claims description 15
- 230000002411 adverse Effects 0.000 claims description 12
- 238000009826 distribution Methods 0.000 claims description 11
- 230000008569 process Effects 0.000 claims description 11
- 230000011218 segmentation Effects 0.000 claims description 11
- 238000004364 calculation method Methods 0.000 claims description 9
- 239000011159 matrix material Substances 0.000 claims description 9
- 230000009466 transformation Effects 0.000 claims description 9
- 230000000694 effects Effects 0.000 claims description 8
- 238000012549 training Methods 0.000 claims description 8
- 238000013528 artificial neural network Methods 0.000 claims description 7
- 230000002159 abnormal effect Effects 0.000 claims description 6
- 230000008859 change Effects 0.000 claims description 6
- 238000004590 computer program Methods 0.000 claims description 6
- 230000000877 morphologic effect Effects 0.000 claims description 6
- 238000010586 diagram Methods 0.000 claims description 5
- 239000006002 Pepper Substances 0.000 claims description 3
- 238000004891 communication Methods 0.000 claims description 3
- 238000005530 etching Methods 0.000 claims description 3
- 238000002474 experimental method Methods 0.000 claims description 3
- 239000000835 fiber Substances 0.000 claims description 3
- 238000012886 linear function Methods 0.000 claims description 3
- 229920000728 polyester Polymers 0.000 claims description 3
- 238000005070 sampling Methods 0.000 claims description 3
- 238000007790 scraping Methods 0.000 claims description 3
- 238000003860 storage Methods 0.000 claims description 3
- 238000012546 transfer Methods 0.000 claims description 3
- 238000009941 weaving Methods 0.000 claims description 3
- 238000003708 edge detection Methods 0.000 claims 1
- 238000003672 processing method Methods 0.000 claims 1
- 238000007726 management method Methods 0.000 abstract description 16
- 238000004519 manufacturing process Methods 0.000 abstract description 9
- 239000004753 textile Substances 0.000 abstract description 6
- 238000004043 dyeing Methods 0.000 abstract description 5
- 230000010354 integration Effects 0.000 abstract description 5
- 238000012544 monitoring process Methods 0.000 abstract description 5
- 238000005516 engineering process Methods 0.000 abstract description 3
- 238000007619 statistical method Methods 0.000 abstract description 2
- 230000002708 enhancing effect Effects 0.000 abstract 2
- 208000028804 PERCHING syndrome Diseases 0.000 abstract 1
- 238000010606 normalization Methods 0.000 abstract 1
- 238000011160 research Methods 0.000 abstract 1
- 238000012360 testing method Methods 0.000 abstract 1
- 230000009286 beneficial effect Effects 0.000 description 2
- 230000004438 eyesight Effects 0.000 description 2
- 238000002372 labelling Methods 0.000 description 2
- 235000002566 Capsicum Nutrition 0.000 description 1
- 235000016761 Piper aduncum Nutrition 0.000 description 1
- 235000017804 Piper guineense Nutrition 0.000 description 1
- 244000203593 Piper nigrum Species 0.000 description 1
- 235000008184 Piper nigrum Nutrition 0.000 description 1
- 230000003044 adaptive effect Effects 0.000 description 1
- 230000004075 alteration Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 238000011049 filling Methods 0.000 description 1
- 238000010191 image analysis Methods 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 230000004630 mental health Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000005693 optoelectronics Effects 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 238000012827 research and development Methods 0.000 description 1
- 150000003839 salts Chemical class 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000011179 visual inspection Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/84—Systems specially adapted for particular applications
- G01N21/88—Investigating the presence of flaws or contamination
- G01N21/8851—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
- 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
- G01N2021/8887—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
Landscapes
- Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Theoretical Computer Science (AREA)
- Signal Processing (AREA)
- Health & Medical Sciences (AREA)
- Quality & Reliability (AREA)
- Chemical & Material Sciences (AREA)
- Analytical Chemistry (AREA)
- Biochemistry (AREA)
- General Health & Medical Sciences (AREA)
- Immunology (AREA)
- Pathology (AREA)
- Treatment Of Fiber Materials (AREA)
- Investigating Materials By The Use Of Optical Means Adapted For Particular Applications (AREA)
Abstract
The invention discloses a kind of Fabric Defects Inspection detection method based on image procossing, this method comprises the following steps：Self-adaptive solution algorithm is taken according to on-site actual situations；（2）Enhancing processing is sharpened to the image after denoising to strengthen the grain details of image fault and edge contour, makes the extraction of subsequent characteristics value relatively reliable；（3）The fault of image is done using morphology operations and circulating area notation and split, filtering is handled with circulation enhancing；（4）Fabric feature value is extracted and normalization；（5）The identification and classification of fabric defects.The Fabric Defects Inspection detection method based on image procossing, research and develop a set of fabric automatic test system based on image procossing and information integration, complete the functions such as Automatic Detection of Fabric Defects, finished product grading, statistical analysis of quality, information sharing, Centralized Monitoring management, during solving textile dyeing and finishing industry using Traditional Man perching and artificial statistics, the common technology problems such as labor intensity is big, production efficiency is low be present.
Description
Technical Field
The invention relates to the technical field of cloth defect detection methods based on image processing, in particular to a cloth defect detection method based on image processing.
Background
At present, the traditional fabric inspection method in the textile dyeing and finishing industry adopts a manual cloth inspection mode that one person is responsible for one cloth inspection machine, so that the method is suitable for various fabrics. Workers need to complete two tasks during operation: on one hand, various fabric defects are identified through manual visual inspection, and the defects are classified; and on the other hand, the fabric defect information is recorded in a paper report in time, and is handed to an administrator for manual statistics and recording into a computer for archiving after the operation is finished.
The working process of the manual cloth inspection is as follows: the fabric is driven by the fabric guide roller, passes through the cloth inspection table, workers visually detect the fabric slowly advancing on the cloth inspection table, and repeatedly inspect the cloth cover. In the process, firstly, the cloth cover is searched and checked by using human eyes, the defect that the human eyes are very easy to fatigue exists, the eyesight of cloth inspectors can be damaged, and the cloth inspecting machine is tedious and is heavy physical labor which is very easy to influence the physical and mental health of cloth inspectors; and secondly, the manual detection depends on the accuracy of the cloth inspection worker in mastering national standards, the proficiency of work and the experience. Therefore, the manual cloth inspection has the problems of low efficiency, easy omission and error detection, and difficult error detection, which causes the problems of product quality, untimely feedback, claim for delivery and the like. Meanwhile, how to realize effective integration of defect data, how to realize informatization management of the defect data to a certain extent, and how to standardize the sources of various report data, improve the correctness of the report data, reduce heavy manual statistical operation, improve the working efficiency of each workshop, fundamentally reduce manual misoperation, strengthen the monitoring of the production process, achieve defect data sharing, and are another problem in the current textile dyeing and finishing industry.
In recent years, some textile dyeing and finishing equipment companies in China have developed mature defect information management systems, and the defect information management systems are applied to the traditional cloth inspecting machines to obtain good effects. For example, a cloth inspection machine control management information system developed by a certain company in the Foshan City facilitates cloth inspection operators to quickly input cloth defect information and help users to issue related cloth inspection reports (defect distribution maps, loss statistics, cloth inspection detail reports, cloth inspection summary reports, four-component system reports and the like) according to four-component system international standards. Meanwhile, the bar code label is convenient for cloth information management, and reduces the workload of manual label filling and human errors. In addition, data exchange with the ERP of the user can be conveniently carried out, and seamless connection is achieved. For example, in a cloth inspection system developed by a Hangzhou company, the system is convenient for recording and processing inspection data such as defect positions, defect lengths and the like, storing the data and providing a finished product inspection analysis table, so that the complex manual statistics work is reduced, and meanwhile, a client can log in by adopting a web, so that the order progress can be conveniently inquired in real time, and the yield information can be shared.
The cloth inspection machine information management system in the domestic market reduces heavy manual statistics operation to a certain extent, so that the production management mode of an enterprise is gradually transited to a computer information management system from the original manual statistics mode, and then gradually transits to networked management of all information, thereby achieving the sharing of production information in a local area network, improving the integration level of information, realizing the real-time and accuracy of defect data, and being beneficial to each workshop to make timely and accurate production process monitoring. However, the operation mode still continues to use a manual cloth inspecting mode that one person is responsible for one cloth inspecting machine, and the automatic detection of the fabric defects by replacing the manual vision by the machine vision is not solved.
In contrast, foreign research and development in the aspect of fabric automatic detection technology and defect information management systems have achieved certain achievements, and are successfully marketed and applied to actual production. Such as an on-line fabric inspection system developed by Obdix optoelectronics, germany, an I-TEX proof system of israel micro silks (EVS), a proof system of birio, and a Fabriscan automatic proof system of Uster, switzerland, using neural network identification technology. These cloth inspecting systems are implemented by using very expensive hardware, for example, the image analysis tool of the cloth inspecting system of the company BARCO uses a CPU1000 with high price and a medium-sized computer with a memory of 128M to achieve the purpose of high-speed detection. The above systems are too expensive, high in maintenance cost and not timely in after-sale service, so that domestic enterprises are difficult to accept. Meanwhile, foreign fabric defect detection systems have limited adaptability to fabric varieties and capability of distinguishing defect types. The adaptive objects of the systems are basically monochromatic and simple-structure fabrics, the detected defects are few in types, and some systems require very obvious defect edges in images, so that the system has large limitation in practical application, and has no market in China.
Disclosure of Invention
The present invention is directed to a method for detecting a cloth defect based on image processing, so as to solve the problems mentioned in the background art.
In order to achieve the purpose, the invention provides the following technical scheme: a cloth defect detection method based on image processing comprises the following steps:
(2) adopting a self-adaptive denoising algorithm according to the actual situation on site;
(2) carrying out sharpening enhancement processing on the denoised image so as to enhance the texture detail and the edge contour of the image defect and ensure that the subsequent characteristic value is more reliably extracted;
(3) utilizing morphological operation and a circulation region marking method to segment the defects of the image, and carrying out filtering and circulation enhancement processing;
(4) extracting and normalizing the characteristic value of the fabric;
(5) and identifying and classifying the fabric defects.
Preferably, the method for detecting the cloth defects based on the image processing comprises the steps of (1) adopting a self-adaptive denoising algorithm according to the actual situation of a field, analyzing a noise source and noise characteristics in the fabric image acquisition process, performing an experiment, correspondingly adopting algorithms such as a space domain or a frequency domain for preprocessing, adopting median filtering if the salt-pepper noise occupies a dominant position, adopting wavelet threshold or Kalman filtering for denoising if the Gaussian noise occupies a dominant position, and adopting the combination of the three for denoising according to the actual effect.
Preferably, the method for detecting cloth defects based on image processing includes that step (2) includes obtaining defect edges by using an edge detector such as a prwitter operator, a sober operator, a laplace operator, or Canny.
Preferably, in the image processing-based cloth defect detection method, in the step (3), the defects of the image are segmented by using morphological operation and a flow region labeling method, and the filtering and the flow enhancement processing are performed by,
calculating the size of a repeating unit on a warp structure and a weft structure in an image of a standard fabric image after binaryzation by adopting an autocorrelation function, and taking the size as a detection window;
the method is characterized in that an opening operation is used for enabling the outline of a defect image to be smooth, burrs and isolated points can be removed, angles are sharpened, small grooves and grooves are filled up by using a closing operation, the two operations are used for different occasions respectively, the two operations are not limited in sequence and are carried out according to actual experimental effects, and the purpose is to hopefully etch a standard longitude and latitude structure repeated structural unit in a fabric image by using an etching operation, namely a so-called background, but the defect information in the fabric image, namely a so-called target, is recovered while the background is etched;
enhancement and calculation of flow-through region by using correlation algorithm
The geometric characteristics of the extracted defects are calculated according to the following formula, and the shape attribute category of the defects is judged according to the characteristic values
Squareness: r is A0/A; a0 is the area of the region, a is the area of the minimum bounding rectangle of the region, the probability of being a rectangle is high when R is 1, and the maximum of a circle when R is PI/4
The volume state ratio T is a/b; a b is the length and width of the minimum bounding rectangle of the region, T1 is square or circle, T >1 is the slender figure
The circularity is a quantity defining the degree of similarity to a circle, As is the area of the connected component, Ls is the perimeter of the connected component, and the circularity is calculated As follows
For circular defects, the circularity is at a maximum, and the more complex the defect shape, the smaller the value. Thus, circularity may be used as a measure of the complexity or roughness of the defect shape
The sphericity C is Ur/Pr, Ur is the average distance from the region barycenter to the contour point, Pr is the mean square error from the region barycenter to the contour point
Central moments, which can be solved for (p, q) th moment using OpenCV with a specialized function
The major axis and minor axis, the major axis and minor axis of the minimum circumscribed rectangle, are specifically a set of optimal units found by linear transformation using Hotelling transformation
Orthogonal basis vectors are used for representing original defect samples again by linear combination of the set of basis vectors. Setting the number of edge pixel points of the defect image as k, and taking each edge coordinate point of the defect as a two-dimensional vector Xi＝[ai，bi]T (i ═ 1, 2, …, k) according to all XiCalculating mean vector m of defect edge coordinatesxSum covariance matrix cx
Due to cxIs a 2 × 2 symmetric matrix, so the number of eigenvectors is only 2, and the matrix composed of these 2 eigenvectors can be expressed as (e) ═ a1，e2) T, wherein e1The eigenvector corresponding to the largest eigenvalue, e2A new set of vectors can be obtained by using A as an operation:
Yi＝A(Xi-mx)(i＝1，2，…，K).
thereby establishing a new coordinate system: using the leaf centroid, the mean vector m, x as the origin, and e1As direction of the horizontal axis of the new coordinates, e2And the length-width ratio of the minimum circumscribed rectangle of the leaves can be obtained after the coordinate rotation transformation:
wherein xmaxAnd xminRespectively the maximum value and the minimum value of the abscissa of the outline of the defect; y ismaxAnd yminRespectively the maximum value and the minimum value of the ordinate of the defect profile
Area, area is defined as the total number of pixels in the connected domain. The size of the connected domain after the binarization processing is measured, and the formula is as follows:
the perimeter or boundary length refers to the length of the boundary contour line surrounding a connected domain. The calculation formula of the perimeter is defined as follows
Wherein, the number of pixels with even direction code on the Ne boundary line, No is the number of pixels with odd direction code on the boundary line,
according to the parameters, the defects are classified according to the specific quality specifications in the factory, the defects are generally used as threshold values, and the effective defect texture feature extraction and classification identification are judged only when the area of a certain region is judged to be between two threshold values
Mallat wavelet fast decomposition is carried out on fabric defect image
The wavelet decomposition adopts dbN wavelet system, coifN wavelet system or symN wavelet system with regularity, orthogonality, symmetry and tight support as wavelet base,
the decomposition formula is as follows:
the specific decomposition block diagram is as shown in the attached figure (5):
whereinAndthe method comprises the following steps of respectively representing detail subgraphs of an image in the vertical direction, the horizontal direction and the diagonal direction, wherein the fabric image is a two-dimensional signal, the two-dimensional wavelet can be represented by the product of two one-dimensional wavelet functions, h and g filters are respectively applied in the horizontal direction and the vertical direction to convert the two-dimensional discrete wavelet into two one-dimensional discrete wavelets, each layer of decomposition is carried out, the two-dimensional discrete wavelet transform can generate four sub-images, the next-level wavelet transform is carried out on the basis of the previous-level low-frequency subgraph, and then, a multi-level wavelet decomposition subgraph can be obtained, and only one layer of wavelet decomposition is carried out to obtain 4 sub:
low-pass filtering and down-sampling are carried out on the rows and the columns of the image to obtain a low-frequency subgraph (LL) corresponding to the general picture of the original image;
respectively carrying out high-pass and low-pass filtering on the rows and the columns to obtain a horizontal subgraph (HL) which reflects the horizontal details of the image;
low-pass and high-pass filtering are respectively carried out on the rows and the columns to obtain a vertical subgraph (LH) which reflects the vertical details of the image;
high-pass filtering is carried out on the rows and the columns to obtain a diagonal subgraph (HH) which reflects the details of universal joint along the diagonal,
fabric texture feature value extraction
After the wavelet decomposition of the fabric image is completed, characteristic values can be extracted from the decomposed sub-images, whether the fabric image contains defects or not is further analyzed, and the positions of the defects are determined. The common fabric is formed by interweaving warp and weft yarns according to a rule according to a weaving process, so that the common fabric has obvious texture characteristics. Fabric defects are detected by identifying defects based on the fact that the texture characteristics of the defects are different from those of a defect-free fabric, and the texture characteristics of the fabric mainly comprise structural characteristics, roughness, periodicity, directionality, continuity and the like. The invention adopts the gray level statistical characteristic of the fabric texture image and judges whether the defects exist or not by comparing the fabric texture characteristics. The scheme adopts the following 5 characteristics, namely energy, variance, extreme error, entropy and adverse moment, Hi and j are set as gray values at pixel points (i and j) of the image, M is the number of lines of the image, N is the number of columns of the image, H is the gray average value of the image, and the definition and calculation formula of five characteristic values are as follows,
energy of
The intensity of the overall gray level of the fabric texture image, namely the uniformity of the gray level distribution of the image, is reflected. The scheme adopts energy, namely second moment, as a first characteristic value, and an expression of the first characteristic value is defined as
To reduce the number of stages, use may also be made of
Variance (variance)
In order to reflect the discrete condition of gray value distribution in the fabric image, the scheme adopts variance as a second characteristic value, and an expression is defined as
Extreme difference
In order to reflect the difference degree of the gray value of the fabric image, the scheme adopts the range difference as a third characteristic value, and the expression is defined as
Where Mh represents the meridional range and Mv represents the meridional range.
Entropy of the entropy
In order to reflect the roughness and the non-uniformity of the texture of the fabric image, entropy is adopted as a fourth characteristic value, and if the texture is dense, the entropy value is large; and if the texture is sparse, the entropy value is small. An expression is defined as
Wherein,eh represents the entropy value of the latitudinal texture, and Ev represents the entropy value of the longitudinal texture.
Moment of adverse difference
In order to express the homogeneity of the fabric image texture, namely reflect the degree of local change of the image texture, the scheme adopts the adverse difference moment as a fifth characteristic value, the adverse difference moment is small, and the change between different texture areas of the image is large. The inverse difference moment is large, the variation among different texture areas of the image is small, and the expression is defined as
Determination of a segmentation window for a fabric image
To judge whether a defect exists and the type of the defect through the difference of the characteristic values. In the step, window segmentation is carried out on a wavelet decomposed subgraph so as to accurately position defects, a characteristic value of each segmented window is calculated, if the defects exist, the characteristic value of the window at the defect position is abnormal, the step analyzes the texture of the image by using an autocorrelation function method to determine the size of the window, and the autocorrelation function definition formula of the image in the longitude and latitude directions is shown as
Wherein Hi, j is the gray value at the pixel point (i, j) of the image, M is the number of rows of the image, N is the number of columns of the image, C (h, 0) and C (v, 0) respectively represent the weft autocorrelation function and warp autocorrelation function of the fabric image,
the period of the function curve calculated by the above formula is repeated as a period reflecting the texture of the fabric. And respectively determining the number of the repeated pixels of the weft yarns and the warp yarns in the fabric image, thereby determining the size of a segmentation detection window as shown in the figure: the calculated results, etc. 4 are given as examples, and M128, N128,
dividing the latitudinal texture subimage, namely the horizontal texture subimage, into 32 long rectangular blocks with the same size along the longitudinal direction, wherein each block has 4 x 128 pixels, and calculating a characteristic value of each rectangular block; for a warp, i.e. vertical, texture sub-image, it is divided into 32 blocks in the weft direction, each block of pixels 128 x 4.
Preferably, in the image processing-based cloth defect detection method, the step (4) of extracting and normalizing the characteristic values of the fabric comprises the first step of respectively calculating 5 characteristic values in each segmentation sub-window, normalizing the 5 characteristic values according to the following formula,
wherein x is the absolute characteristic value, x is the mean value, xmax，xminRespectively the maximum value and the minimum value of the absolute characteristic value of the normal fabric without defects, y is the relative characteristic value of the fabric to be detected, thus, the relative characteristic value normalized by the linear function is completely mapped in the interval (1, 1),
respectively obtaining the following series of characteristic value curve distribution graphs by taking the defect-free defects and the flying defects as examples, wherein the abscissa is the serial number of 1-32 sub-windows;
and observing and analyzing whether the characteristic value in each window is abnormal or not, if the characteristic value curve of the fabric fluctuates in a proper range, determining that the fabric is normal and has no defects, and if the characteristic value curve has fluctuation obviously beyond the range.
Preferably, the image processing-based cloth defect detecting method, step (5) textile defect identifying and classifying method, step 1, determining the number of neurons in input layer and output layer according to the number of identified types and the number of identified characteristic values
Determining ten fabric defect characteristic values of latitudinal energy, latitudinal variance, latitudinal extreme error, latitudinal entropy value, latitudinal inverse error moment, longitudinal energy, longitudinal variance, longitudinal extreme error, longitudinal entropy value and longitudinal inverse error moment as the number of neurons in the input layer; the number of nodes of the output layer is taken as the category number in the classification network, the number of neurons of the output layer is taken as 10 in the scheme, and the 9 fabric defects and the defect-free defects respectively represent white bars, miscellaneous fibers, broken yarns, long residues, broken buttons, broken polyester yarns, thread scraping, dirt and flying
Step 2, determining the number of hidden layer neurons
The number of hidden layer neurons is determined with reference to the following formula,
wherein I is the number of hidden layer nodes, n is the number of input layer nodes, m is the number of output layer nodes, a is a tuning constant between 1 and 10,
step 3, selecting transfer functions and learning rates of the hidden layer and the output layer according to specific conditions
And 4, training data by adopting various BP training algorithms to determine weight values and bias values of the neural network, wherein the weight values and the bias values comprise momentum BP algorithm, learning rate variable BP algorithm, elastic BP algorithm, gradient variable BP algorithm, quasi-Newton algorithm, LM algorithm and the like, and the specific process comprises the following steps:
after a neural network weight and a bias parameter are trained in the step 5, a high-level computer program with the functions of collection, filtering and denoising, feature extraction and classification recognition is realized by using high-level programming languages such as vc + +, vb and the like, and the high-level computer program is used as the front end of the whole defect information management system to complete a B/S or C/S system software architecture;
and 6, transmitting the identification result and field information such as time, equipment number and the like to a certain WEB server substation through network communication protocols such as a network (tcpip, can bus, modbus485/232) and the like, entering a background database for storage, and fusing with an enterprise center ERP system.
Compared with the prior art, the invention has the beneficial effects that: according to the cloth defect detection method based on image processing, a set of automatic fabric inspection system based on image processing and information integration is developed to complete functions of automatic fabric defect detection, finished product grading, quality statistical analysis, information sharing, centralized monitoring management and the like, so that common technical problems of high labor intensity, high labor cost, easiness in missed inspection and false inspection, large fluctuation of cloth inspection effect, fussy information statistics, low production efficiency and the like in the conventional manual cloth inspection and manual statistics in the textile dyeing and finishing industry are solved, and the purposes of reducing labor cost, realizing automatic monitoring and information integration of enterprise production lines, meeting quality management requirements and improving automatic production and informatization management levels are achieved.
Drawings
FIG. 1 is a distribution diagram of a defect-free characteristic value curve of the present invention.
FIG. 2 is a graph of a characteristic value profile of flying defects in accordance with the present invention.
FIG. 3 is a flowchart of the bp training algorithm for training data according to the present invention.
FIG. 4 is a graph of the longitudinal and radial autocorrelation functions of the present invention.
FIG. 5 is a schematic diagram showing the detailed exploded view in step (3) of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention. Referring to fig. 1, the present invention provides a technical solution:
example 1:
a cloth defect detection method based on image processing comprises the following steps:
(3) adopting a self-adaptive denoising algorithm according to the actual situation on site;
(2) carrying out sharpening enhancement processing on the denoised image so as to enhance the texture detail and the edge contour of the image defect and ensure that the subsequent characteristic value is more reliably extracted;
(3) utilizing morphological operation and a circulation region marking method to segment the defects of the image, and carrying out filtering and circulation enhancement processing;
(4) extracting and normalizing the characteristic value of the fabric;
(5) and identifying and classifying the fabric defects.
Example 2:
according to the cloth defect detection method based on image processing in embodiment 1, the method of adopting the self-adaptive denoising algorithm according to the actual situation on site in step (1) is to analyze the noise source and the noise characteristics in the fabric image acquisition process, carry out the pretreatment by adopting the algorithms of airspace or frequency domain and the like correspondingly in the experiment, adopt median filtering if the salt and pepper noise occupies the dominant position, adopt wavelet threshold or Kalman filtering denoising if the Gaussian noise occupies the dominant position, and can also adopt the combination denoising of the three according to the actual effect.
Example 3:
the method for detecting cloth defects based on image processing according to embodiment 1 or 2, wherein the step (2) comprises finding the defect edge by using an edge detector such as a prwitter operator, a sober operator, a laplace operator, or Canny.
Example 4:
according to the image processing-based cloth defect detection method of embodiment 1, 2 or 3, the step (3) uses morphological operation and flow-through region labeling method to divide the defects of the image, filter and perform the flow-through enhancement processing,
calculating the size of a repeating unit on a warp structure and a weft structure in an image of a standard fabric image after binaryzation by adopting an autocorrelation function, and taking the size as a detection window;
the method is characterized in that an opening operation is used for enabling the outline of a defect image to be smooth, burrs and isolated points can be removed, angles are sharpened, small grooves and grooves are filled up by using a closing operation, the two operations are used for different occasions respectively, the two operations are not limited in sequence and are carried out according to actual experimental effects, and the purpose is to hopefully etch a standard longitude and latitude structure repeated structural unit in a fabric image by using an etching operation, namely a so-called background, but the defect information in the fabric image, namely a so-called target, is recovered while the background is etched;
enhancement and calculation of flow-through region by using correlation algorithm
The geometric characteristics of the extracted defects are calculated according to the following formula, and the shape attribute category of the defects is judged according to the characteristic values
Squareness: r is A0/A; a0 is the area of the region, a is the area of the minimum bounding rectangle of the region, the probability of being a rectangle is high when R is 1, and the maximum of a circle when R is PI/4
The volume state ratio T is a/b; a b is the length and width of the minimum bounding rectangle of the region, T1 is square or circle, T >1 is the slender figure
The circularity is a quantity defining the degree of similarity to a circle, As is the area of the connected component, Ls is the perimeter of the connected component, and the circularity is calculated As follows
For circular defects, the circularity is at a maximum, and the more complex the defect shape, the smaller the value. Thus, circularity may be used as a measure of the complexity or roughness of the defect shape
The sphericity C is Ur/Pr, Ur is the average distance from the region barycenter to the contour point, Pr is the mean square error from the region barycenter to the contour point
Central moments, which can be solved for (p, q) th moment using OpenCV with a specialized function
Major and minor axes, major and minor axes of the smallest circumscribed rectangle, specifically by linear transformation using Hotelling's transformationThe transformation finds a group of optimal unit orthogonal basis vectors, and the linear combination of the group of basis vectors is used for re-representing the original defect samples. Setting the number of edge pixel points of the defect image as k, and taking each edge coordinate point of the defect as a two-dimensional vector Xi＝[ai，bi]T (i ═ 1, 2, …, k) according to all XiCalculating mean vector m of defect edge coordinatesxSum covariance matrix cx
Due to cxIs a 2 × 2 symmetric matrix, so the number of eigenvectors is only 2, and the matrix composed of these 2 eigenvectors can be expressed as (e) ═ a1，e2) T, wherein e1The eigenvector corresponding to the largest eigenvalue, e2A new set of vectors can be obtained by using A as an operation:
Yi＝A(Xi-mx)(i＝1，2，…，K).
thereby establishing a new coordinate system: using the leaf centroid, the mean vector m, x as the origin, and e1As direction of the horizontal axis of the new coordinates, e2And the length-width ratio of the minimum circumscribed rectangle of the leaves can be obtained after the coordinate rotation transformation:
wherein xmaxAnd xminRespectively the maximum value and the minimum value of the abscissa of the outline of the defect; y ismaxAnd yminRespectively the maximum value and the minimum value of the ordinate of the defect profile
Area, area is defined as the total number of pixels in the connected domain. The size of the connected domain after the binarization processing is measured, and the formula is as follows:
the perimeter or boundary length refers to the length of the boundary contour line surrounding a connected domain. The calculation formula of the perimeter is defined as follows
Wherein, the number of pixels with even direction code on the Ne boundary line, No is the number of pixels with odd direction code on the boundary line,
according to the parameters, the defects are classified according to the specific quality specifications in the factory, the defects are generally used as threshold values, and the effective defect texture feature extraction and classification identification are judged only when the area of a certain region is judged to be between two threshold values
Mallat wavelet fast decomposition is carried out on fabric defect image
The wavelet decomposition adopts dbN wavelet system, coifN wavelet system or symN wavelet system with regularity, orthogonality, symmetry and tight support as wavelet base,
the decomposition formula is as follows:
the specific exploded block diagram is as shown in figure 5:
whereinAndthe method comprises the following steps of respectively representing detail subgraphs of an image in the vertical direction, the horizontal direction and the diagonal direction, wherein the fabric image is a two-dimensional signal, the two-dimensional wavelet can be represented by the product of two one-dimensional wavelet functions, h and g filters are respectively applied in the horizontal direction and the vertical direction to convert the two-dimensional discrete wavelet into two one-dimensional discrete wavelets, each layer of decomposition is carried out, the two-dimensional discrete wavelet transform can generate four sub-images, the next-level wavelet transform is carried out on the basis of the previous-level low-frequency subgraph, and then, a multi-level wavelet decomposition subgraph can be obtained, and only one layer of wavelet decomposition is carried out to obtain 4 sub:
low-pass filtering and down-sampling are carried out on the rows and the columns of the image to obtain a low-frequency subgraph (LL) corresponding to the general picture of the original image;
respectively carrying out high-pass and low-pass filtering on the rows and the columns to obtain a horizontal subgraph (HL) which reflects the horizontal details of the image;
low-pass and high-pass filtering are respectively carried out on the rows and the columns to obtain a vertical subgraph (LH) which reflects the vertical details of the image;
and carrying out high-pass filtering on the rows and the columns to obtain a diagonal subgraph (HH) reflecting details universal along the diagonal.
Fabric texture feature value extraction
After the wavelet decomposition of the fabric image is completed, characteristic values can be extracted from the decomposed sub-images, whether the fabric image contains defects or not is further analyzed, and the positions of the defects are determined. The common fabric is formed by interweaving warp and weft yarns according to a rule according to a weaving process, so that the common fabric has obvious texture characteristics. Fabric defects are detected by identifying defects based on the fact that the texture characteristics of the defects are different from those of a defect-free fabric, and the texture characteristics of the fabric mainly comprise structural characteristics, roughness, periodicity, directionality, continuity and the like. The invention adopts the gray level statistical characteristic of the fabric texture image and judges whether the defects exist or not by comparing the fabric texture characteristics. The scheme adopts the following 5 characteristics, namely energy, variance, extreme error, entropy and adverse moment, Hi and j are set as gray values at pixel points (i and j) of the image, M is the number of lines of the image, N is the number of columns of the image, H is the gray average value of the image, and the definition and calculation formula of five characteristic values are as follows,
energy of
The intensity of the overall gray level of the fabric texture image, namely the uniformity of the gray level distribution of the image, is reflected. The scheme adopts energy, namely second moment, as a first characteristic value, and an expression of the first characteristic value is defined as
To reduce the number of stages, use may also be made of
Variance (variance)
In order to reflect the discrete condition of gray value distribution in the fabric image, the scheme adopts variance as a second characteristic value, and an expression is defined as
Extreme difference
In order to reflect the difference degree of the gray value of the fabric image, the scheme adopts the range difference as a third characteristic value, and the expression is defined as
Where Mh represents the meridional range and Mv represents the meridional range.
Entropy of the entropy
In order to reflect the roughness and the non-uniformity of the texture of the fabric image, entropy is adopted as a fourth characteristic value, and if the texture is dense, the entropy value is large; and if the texture is sparse, the entropy value is small. An expression is defined as
Wherein,eh represents the entropy value of the latitudinal texture, and Ev represents the entropy value of the longitudinal texture.
Moment of adverse difference
In order to express the homogeneity of the fabric image texture, namely reflect the degree of local change of the image texture, the scheme adopts the adverse difference moment as a fifth characteristic value, the adverse difference moment is small, and the change between different texture areas of the image is large. The inverse difference moment is large, the variation among different texture areas of the image is small, and the expression is defined as
Determination of a segmentation window for a fabric image
To judge whether a defect exists and the type of the defect through the difference of the characteristic values. In the step, window segmentation is carried out on a wavelet decomposed subgraph so as to accurately position defects, a characteristic value of each segmented window is calculated, if the defects exist, the characteristic value of the window at the defect position is abnormal, the step analyzes the texture of the image by using an autocorrelation function method to determine the size of the window, and the autocorrelation function definition formula of the image in the longitude and latitude directions is shown as
Wherein Hi, j is the gray value at the pixel point (i, j) of the image, M is the number of rows of the image, N is the number of columns of the image, C (h, 0) and C (v, 0) respectively represent the weft autocorrelation function and warp autocorrelation function of the fabric image,
the period of the function curve calculated by the above formula is repeated as a period reflecting the texture of the fabric. And respectively determining the number of the repeated pixels of the weft yarns and the warp yarns in the fabric image, thereby determining the size of a segmentation detection window as shown in the figure: the calculated results, etc. 4 are given as examples, and M128, N128,
dividing the latitudinal texture subimage, namely the horizontal texture subimage, into 32 long rectangular blocks with the same size along the longitudinal direction, wherein each block has 4 x 128 pixels, and calculating a characteristic value of each rectangular block; for a warp, i.e. vertical, texture sub-image, it is divided into 32 blocks in the weft direction, each block of pixels 128 x 4.
Example 5:
the image processing-based cloth defect detecting method according to embodiment 1 or 2 or 3 or 4, step (4) a fabric feature value extracting and normalizing method, wherein in the first step, 5 feature values in each of the divided sub-windows are respectively calculated, the 5 feature values are normalized according to the following formula,
wherein x is the absolute characteristic value, x is the mean value, xmax，xminRespectively the maximum value and the minimum value of the absolute characteristic value of the normal fabric without defects, y is the relative characteristic value of the fabric to be detected, thus, the relative characteristic value normalized by the linear function is completely mapped in the interval (1, 1),
respectively obtaining the following series of characteristic value curve distribution graphs by taking the defect-free defects and the flying defects as examples, wherein the abscissa is the serial number of 1-32 sub-windows;
and observing and analyzing whether the characteristic value in each window is abnormal or not, if the characteristic value curve of the fabric fluctuates in a proper range, determining that the fabric is normal and has no defects, and if the characteristic value curve has fluctuation obviously beyond the range.
Example 6:
the image processing-based cloth defect detection method of embodiments 1 or 2 or 3 or 4 or 5, step (5) a fabric defect identification and classification method, step 1, determining the number of neurons in input and output layers based on the number of identified classes and the number of identified eigenvalues
Determining ten fabric defect characteristic values of latitudinal energy, latitudinal variance, latitudinal extreme error, latitudinal entropy value, latitudinal inverse error moment, longitudinal energy, longitudinal variance, longitudinal extreme error, longitudinal entropy value and longitudinal inverse error moment as the number of neurons in the input layer; the number of nodes of the output layer is taken as the category number in the classification network, the number of neurons of the output layer is taken as 10 in the scheme, and the 9 fabric defects and the defect-free defects respectively represent white bars, miscellaneous fibers, broken yarns, long residues, broken buttons, broken polyester yarns, thread scraping, dirt and flying
Step 2, determining the number of hidden layer neurons
The number of hidden layer neurons is determined with reference to the following formula,
wherein I is the number of hidden layer nodes, n is the number of input layer nodes, m is the number of output layer nodes, a is a tuning constant between 1 and 10,
step 3, selecting transfer functions and learning rates of the hidden layer and the output layer according to specific conditions
And 4, training data by adopting various BP training algorithms to determine weight values and bias values of the neural network, wherein the weight values and the bias values comprise momentum BP algorithm, learning rate variable BP algorithm, elastic BP algorithm, gradient variable BP algorithm, quasi-Newton algorithm, LM algorithm and the like, and the specific process comprises the following steps:
after a neural network weight and a bias parameter are trained in the step 5, a high-level computer program with the functions of collection, filtering and denoising, feature extraction and classification recognition is realized by using high-level programming languages such as vc + +, vb and the like, and the high-level computer program is used as the front end of the whole defect information management system to complete a B/S or C/S system software architecture;
and 6, transmitting the identification result and field information such as time, equipment number and the like to a certain WEB server substation through network communication protocols such as a network (tcpip, can bus, modbus485/232) and the like, entering a background database for storage, and fusing with an enterprise center ERP system.
Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that changes, modifications, substitutions and alterations can be made in these embodiments without departing from the principles and spirit of the invention, the scope of which is defined in the appended claims and their equivalents.
Claims (6)
1. A cloth defect detection method based on image processing is characterized in that: the method comprises the following steps:
(1) adopting a self-adaptive denoising algorithm according to the actual situation on site;
(2) carrying out sharpening enhancement processing on the denoised image so as to enhance the texture detail and the edge contour of the image defect and ensure that the subsequent characteristic value is more reliably extracted;
(3) utilizing morphological operation and a circulation region marking method to segment the defects of the image, and carrying out filtering and circulation enhancement processing;
(4) extracting and normalizing the characteristic value of the fabric;
(5) and identifying and classifying the fabric defects.
2. The image processing-based cloth defect detection method of claim 1, wherein: the method for adopting the self-adaptive denoising algorithm according to the actual situation on site in the step (1) is to analyze the noise source and the noise characteristics in the fabric image acquisition process, carry out the pretreatment by adopting algorithms such as airspace or frequency domain and the like correspondingly in the experiment, adopt median filtering if the salt-pepper noise occupies the dominant position, adopt wavelet threshold or Kalman filtering denoising if the Gaussian noise occupies the dominant position, and adopt the combination of the three for denoising according to the actual effect.
3. The image processing-based cloth defect detection method of claim 1, wherein: and (2) solving the flaw edge by utilizing an edge detection operator such as a prwitter operator, a sober operator, a laplace operator or Canny.
4. The image processing-based cloth defect detection method of claim 1, wherein: step (3) using morphological operation and circulation region marking method to divide the image defect, filtering and circulation enhancement processing method,
calculating the size of a repeating unit on a warp structure and a weft structure in an image of a standard fabric image after binaryzation by adopting an autocorrelation function, and taking the size as a detection window;
the method is characterized in that an opening operation is used for enabling the outline of a defect image to be smooth, burrs and isolated points can be removed, angles are sharpened, small grooves and grooves are filled up by using a closing operation, the two operations are used for different occasions respectively, the two operations are not limited in sequence and are carried out according to actual experimental effects, and the purpose is to hopefully etch a standard longitude and latitude structure repeated structural unit in a fabric image by using an etching operation, namely a so-called background, but the defect information in the fabric image, namely a so-called target, is recovered while the background is etched;
enhancement and calculation of flow-through region by using correlation algorithm
The geometric characteristics of the extracted defects are calculated according to the following formula, and the shape attribute category of the defects is judged according to the characteristic values
Squareness: r is A0/A; a0 is the area of the region, a is the area of the minimum bounding rectangle of the region, the probability of being a rectangle is high when R is 1, and the maximum of a circle when R is PI/4
The volume state ratio T is a/b; a b is the length and width of the minimum bounding rectangle of the region, T1 is square or circle, T >1 is the slender figure
The circularity is a quantity defining the degree of similarity to a circle, As is the area of the connected component, Ls is the perimeter of the connected component, and the circularity is calculated As follows
For circular defects, the circularity is at a maximum, and the more complex the defect shape, the smaller the value. Thus, circularity may be used as a measure of the complexity or roughness of the defect shape
The sphericity C is Ur/Pr, Ur is the average distance from the region barycenter to the contour point, Pr is the mean square error from the region barycenter to the contour point
Central moments, which can be solved for (p, q) th moment using OpenCV with a specialized function
The major axis minor axis, the major axis and the minor axis of the minimum circumscribed rectangle, specifically, a group of optimal unit orthogonal basis vectors are found through linear transformation by utilizing Hotelling transformation, and the original defect samples are re-represented by linear combination of the group of basis vectors. Setting the number of edge pixel points of the defect image as k, and taking each edge coordinate point of the defect as a two-dimensional vector Xi＝[ai，bi]T (i ═ 1, 2, …, k) according to all XiCalculating mean vector m of defect edge coordinatesxSum covariance matrix cx
Due to cxIs a 2 × 2 symmetric matrix, so the number of eigenvectors is only 2, and the matrix composed of these 2 eigenvectors can be expressed as (e) ═ a1，e2) T, wherein e1The eigenvector corresponding to the largest eigenvalue, e2A new set of vectors can be obtained by using A as an operation:
Yi＝A(Xi-mx) (i＝1，2，…，K).
thereby establishing a new coordinate system: using the leaf centroid, the mean vector m, x as the origin, and e1As direction of the horizontal axis of the new coordinates, e2And the length-width ratio of the minimum circumscribed rectangle of the leaves can be obtained after the coordinate rotation transformation:
wherein xmaxAnd xminRespectively the maximum value and the minimum value of the abscissa of the outline of the defect; y ismaxAnd yminRespectively the maximum value and the minimum value of the ordinate of the defect profile
Area, area is defined as the total number of pixels in the connected domain. The size of the connected domain after the binarization processing is measured, and the formula is as follows:
the perimeter or boundary length refers to the length of the boundary contour line surrounding a connected domain. The calculation formula of the perimeter is defined as follows
Wherein, the number of pixels with even direction code on the Ne boundary line, No is the number of pixels with odd direction code on the boundary line,
according to the parameters, the defects are classified according to the specific quality specifications in the factory, the defects are generally used as threshold values, and the effective defect texture feature extraction and classification identification are judged only when the area of a certain region is judged to be between two threshold values
Mallat wavelet fast decomposition is carried out on fabric defect image
The wavelet decomposition adopts dbN wavelet system, coifN wavelet system or symN wavelet system with regularity, orthogonality, symmetry and tight support as wavelet base,
the decomposition formula is as follows:
the specific exploded block diagram is as shown in figure 5:
whereinAndthe method comprises respectively representing detail subgraphs of an image in the vertical direction, the horizontal direction and the diagonal direction, wherein a fabric image is a two-dimensional signal, a two-dimensional wavelet can be represented by the product of two one-dimensional wavelet functions, and the two-dimensional discrete wavelet can be converted into two one-dimensional discrete small wavelets by respectively applying h and g filters in the horizontal direction and the vertical directionWave, every time through one layer of decomposition, the two-dimensional discrete wavelet transform can produce four sub-images, the next level of wavelet transform is carried out on the basis of the previous level of low-frequency subgraph, and then the multilevel wavelet decomposition subgraph can be obtained, and the scheme only carries out one layer of wavelet decomposition to obtain 4 sub-images with the size of one fourth:
low-pass filtering and down-sampling are carried out on the rows and the columns of the image to obtain a low-frequency subgraph (LL) corresponding to the general picture of the original image;
respectively carrying out high-pass and low-pass filtering on the rows and the columns to obtain a horizontal subgraph (HL) which reflects the horizontal details of the image;
low-pass and high-pass filtering are respectively carried out on the rows and the columns to obtain a vertical subgraph (LH) which reflects the vertical details of the image;
and carrying out high-pass filtering on the rows and the columns to obtain a diagonal subgraph (HH) reflecting details universal along the diagonal.
Fabric texture feature value extraction
After the wavelet decomposition of the fabric image is completed, characteristic values can be extracted from the decomposed sub-images, whether the fabric image contains defects or not is further analyzed, and the positions of the defects are determined. The common fabric is formed by interweaving warp and weft yarns according to a rule according to a weaving process, so that the common fabric has obvious texture characteristics. Fabric defects are detected by identifying defects based on the fact that the texture characteristics of the defects are different from those of a defect-free fabric, and the texture characteristics of the fabric mainly comprise structural characteristics, roughness, periodicity, directionality, continuity and the like. The invention adopts the gray level statistical characteristic of the fabric texture image and judges whether the defects exist or not by comparing the fabric texture characteristics. The scheme adopts the following 5 characteristics, namely energy, variance, extreme error, entropy and adverse moment, Hi and j are set as gray values at pixel points (i and j) of the image, M is the number of lines of the image, N is the number of columns of the image, H is the gray average value of the image, and the definition and calculation formula of five characteristic values are as follows,
energy of
The intensity of the overall gray level of the fabric texture image, namely the uniformity of the gray level distribution of the image, is reflected. The scheme adopts energy, namely second moment, as a first characteristic value, and an expression of the first characteristic value is defined as
To reduce the number of stages, use may also be made of
Variance (variance)
In order to reflect the discrete condition of gray value distribution in the fabric image, the scheme adopts variance as a second characteristic value, and an expression is defined as
Extreme difference
In order to reflect the difference degree of the gray value of the fabric image, the scheme adopts the range difference as a third characteristic value, and the expression is defined as
Where Mh represents the meridional range and Mv represents the meridional range.
Entropy of the entropy
In order to reflect the roughness and the non-uniformity of the texture of the fabric image, entropy is adopted as a fourth characteristic value, and if the texture is dense, the entropy value is large; and if the texture is sparse, the entropy value is small. An expression is defined as
Wherein,eh represents the entropy value of the latitudinal texture, and Ev represents the entropy value of the longitudinal texture.
Moment of adverse difference
In order to express the homogeneity of the fabric image texture, namely reflect the degree of local change of the image texture, the scheme adopts the adverse difference moment as a fifth characteristic value, the adverse difference moment is small, and the change between different texture areas of the image is large. The inverse difference moment is large, the variation among different texture areas of the image is small, and the expression is defined as
Determination of a segmentation window for a fabric image
To judge whether a defect exists and the type of the defect through the difference of the characteristic values. In the step, window segmentation is carried out on a wavelet decomposed subgraph so as to accurately position defects, a characteristic value of each segmented window is calculated, if the defects exist, the characteristic value of the window at the defect position is abnormal, the step analyzes the texture of the image by using an autocorrelation function method to determine the size of the window, and the autocorrelation function definition formula of the image in the longitude and latitude directions is shown as
Wherein Hi, j is the gray value at the pixel point (i, j) of the image, M is the number of rows of the image, N is the number of columns of the image, C (h, 0) and C (v, 0) respectively represent the weft autocorrelation function and warp autocorrelation function of the fabric image,
the period of the function curve calculated by the above formula is repeated as a period reflecting the texture of the fabric. And respectively determining the number of the repeated pixels of the weft yarns and the warp yarns in the fabric image, thereby determining the size of a segmentation detection window as shown in the figure: the calculated results, etc. 4 are given as examples, and M128, N128,
dividing the latitudinal texture subimage, namely the horizontal texture subimage, into 32 long rectangular blocks with the same size along the longitudinal direction, wherein each block has 4 x 128 pixels, and calculating a characteristic value of each rectangular block; for a warp, i.e. vertical, texture sub-image, it is divided into 32 blocks in the weft direction, each block of pixels 128 x 4.
5. The image processing-based cloth defect detection method of claim 1, wherein: step (4) the method for extracting and normalizing the characteristic values of the fabric comprises the first step of respectively calculating 5 characteristic values in each segmentation sub-window and normalizing the 5 characteristic values according to the following formula,
wherein x is the absolute characteristic value, x is the mean value, xmax，xminRespectively the maximum value and the minimum value of the absolute characteristic value of the normal fabric without defects, y is the relative characteristic value of the fabric to be detected, thus, the relative characteristic value normalized by the linear function is completely mapped in the interval (1, 1),
respectively obtaining the following series of characteristic value curve distribution graphs by taking the defect-free defects and the flying defects as examples, wherein the abscissa is the serial number of 1-32 sub-windows;
and observing and analyzing whether the characteristic value in each window is abnormal or not, if the characteristic value curve of the fabric fluctuates in a proper range, determining that the fabric is normal and has no defects, and if the characteristic value curve has fluctuation obviously beyond the range.
6. The image processing-based cloth defect detection method of claim 1, wherein: step (5) identification and classification method of fabric defects, step 1, identifying the number of characteristic values to determine the number of neurons in an input layer and an output layer according to the number of identification types
Determining ten fabric defect characteristic values of latitudinal energy, latitudinal variance, latitudinal extreme error, latitudinal entropy value, latitudinal inverse error moment, longitudinal energy, longitudinal variance, longitudinal extreme error, longitudinal entropy value and longitudinal inverse error moment as the number of neurons in the input layer; the number of nodes of the output layer is taken as the category number in the classification network, the number of neurons of the output layer is taken as 10 in the scheme, and the 9 fabric defects and the defect-free defects respectively represent white bars, miscellaneous fibers, broken yarns, long residues, broken buttons, broken polyester yarns, thread scraping, dirt and flying
Step 2, determining the number of hidden layer neurons
The number of hidden layer neurons is determined with reference to the following formula,
wherein I is the number of hidden layer nodes, n is the number of input layer nodes, m is the number of output layer nodes, a is a tuning constant between 1 and 10,
step 3, selecting transfer functions and learning rates of the hidden layer and the output layer according to specific conditions
And 4, training data by adopting various BP training algorithms to determine weight values and bias values of the neural network, wherein the weight values and the bias values comprise momentum BP algorithm, learning rate variable BP algorithm, elastic BP algorithm, gradient variable BP algorithm, quasi-Newton algorithm, LM algorithm and the like, and the specific process comprises the following steps:
after a neural network weight and a bias parameter are trained in the step 5, a high-level computer program with the functions of collection, filtering and denoising, feature extraction and classification recognition is realized by using high-level programming languages such as vc + +, vb and the like, and the high-level computer program is used as the front end of the whole defect information management system to complete a B/S or C/S system software architecture;
and 6, transmitting the identification result and field information such as time, equipment number and the like to a certain WEB server substation through network communication protocols such as a network (tcpip, can bus, modbus485/232) and the like, entering a background database for storage, and fusing with an enterprise center ERP system.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710547024.2A CN107870172A (en) | 2017-07-06 | 2017-07-06 | A kind of Fabric Defects Inspection detection method based on image procossing |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710547024.2A CN107870172A (en) | 2017-07-06 | 2017-07-06 | A kind of Fabric Defects Inspection detection method based on image procossing |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN107870172A true CN107870172A (en) | 2018-04-03 |
Family
ID=61761811
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710547024.2A Pending CN107870172A (en) | 2017-07-06 | 2017-07-06 | A kind of Fabric Defects Inspection detection method based on image procossing |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107870172A (en) |
Cited By (35)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108921819A (en) * | 2018-05-29 | 2018-11-30 | 黎明职业大学 | A kind of cloth examination device and method based on machine vision |
| CN109035135A (en) * | 2018-07-13 | 2018-12-18 | 常州宏大智能装备产业发展研究院有限公司 | A kind of on-line automatic pattern adjustment method of fabric based on machine vision |
| CN109035195A (en) * | 2018-05-08 | 2018-12-18 | 武汉纺织大学 | A kind of fabric defect detection method |
| CN109100360A (en) * | 2018-07-05 | 2018-12-28 | 福建省国业科技发展有限公司 | The method that ring signal and initial fixed character carry out flaw positioning based on |
| CN109272494A (en) * | 2018-08-31 | 2019-01-25 | 龙山县惹巴妹手工织品有限公司 | A kind of toy watch leather fabric detection method |
| 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 |
| CN109344868A (en) * | 2018-08-28 | 2019-02-15 | 广东奥普特科技股份有限公司 | A kind of universal method for distinguishing axisymmetric inhomogeneity object each other |
| CN109507192A (en) * | 2018-11-02 | 2019-03-22 | 江苏理工学院 | A kind of magnetic core detection method of surface flaw based on machine vision |
| CN109613006A (en) * | 2018-12-22 | 2019-04-12 | 中原工学院 | A kind of fabric defect detection method based on end-to-end neural network |
| CN109884066A (en) * | 2019-01-22 | 2019-06-14 | 江苏恒力化纤股份有限公司 | A kind of method of offline inspection long filament silk volume greasy dirt fault |
| CN110084246A (en) * | 2019-04-17 | 2019-08-02 | 江南大学 | A kind of yarn dyed fabric fault automatic identifying method |
| CN110097538A (en) * | 2019-04-17 | 2019-08-06 | 江南大学 | A kind of online cloth examination device of loom and defects identification method |
| CN110376211A (en) * | 2019-08-07 | 2019-10-25 | 浙江大学台州研究院 | A kind of the synthetic leather crimping on-line measuring device and method of wet process gluing |
| CN111105398A (en) * | 2019-12-19 | 2020-05-05 | 昆明能讯科技有限责任公司 | Transmission line component crack detection method based on visible light image data |
| CN111402225A (en) * | 2020-03-13 | 2020-07-10 | 浙江工业大学 | Cloth folding false detection defect discrimination method |
| CN111537519A (en) * | 2020-04-11 | 2020-08-14 | 嘉兴市昊鸣纺织有限公司 | Automatic detection and supervision system for cloth |
| CN111862125A (en) * | 2019-04-28 | 2020-10-30 | 株式会社理光 | Contour segmentation method and device and computer readable storage medium |
| CN111882545A (en) * | 2020-07-30 | 2020-11-03 | 中原工学院 | Fabric defect detection method based on bidirectional information transmission and feature fusion |
| CN111929327A (en) * | 2020-09-09 | 2020-11-13 | 深兰人工智能芯片研究院（江苏）有限公司 | Cloth defect detection method and device |
| CN112330673A (en) * | 2020-12-11 | 2021-02-05 | 武汉纺织大学 | Woven fabric density detection method based on image processing |
| CN112365442A (en) * | 2020-10-13 | 2021-02-12 | 江苏恒力化纤股份有限公司 | Image analysis-based rectangular spinneret hole residual impurity detection method |
| CN112365448A (en) * | 2020-10-20 | 2021-02-12 | 天津大学 | Fabric defect detection method in warp knitting process |
| CN113139948A (en) * | 2021-04-28 | 2021-07-20 | 福建自贸试验区厦门片区Manteia数据科技有限公司 | Organ contour line quality evaluation method, device and system |
| CN113610852A (en) * | 2021-10-10 | 2021-11-05 | 江苏祥顺布业有限公司 | Yarn drafting quality monitoring method based on image processing |
| CN113643276A (en) * | 2021-08-29 | 2021-11-12 | 浙江工业大学 | Textile texture defect automatic detection method based on statistical analysis |
| CN113638104A (en) * | 2021-10-14 | 2021-11-12 | 海门市恒昌织带有限公司 | Intelligent yarn cleaning control method and system for bobbin winder |
| CN113888459A (en) * | 2021-05-17 | 2022-01-04 | 浙江大学绍兴微电子研究中心 | Texture image defect detection system and method based on self-adaptive texture features |
| CN114494259A (en) * | 2022-04-18 | 2022-05-13 | 南通东德纺织科技有限公司 | Cloth defect detection method based on artificial intelligence |
| CN114723705A (en) * | 2022-03-31 | 2022-07-08 | 海门市恒创织带有限公司 | Cloth flaw detection method based on image processing |
| CN115047070A (en) * | 2022-08-11 | 2022-09-13 | 江苏恒力化纤股份有限公司 | Fabric surface defect detection method based on friction vibration signal |
| CN115082460A (en) * | 2022-08-18 | 2022-09-20 | 聊城市恒丰电子有限公司 | Weaving production line quality monitoring method and system |
| CN116721105A (en) * | 2023-08-11 | 2023-09-08 | 山东淼珠生物科技有限公司 | Explosive bead production abnormality online detection method based on artificial intelligence |
| CN117132593A (en) * | 2023-10-25 | 2023-11-28 | 济宁华晟服装股份有限公司 | Cloth roughness detection method for resisting periodic texture influence |
| CN117475000A (en) * | 2023-12-28 | 2024-01-30 | 江苏恒力化纤股份有限公司 | Fabric selvedge positioning method |
| CN117911417A (en) * | 2024-03-20 | 2024-04-19 | 天津市凯瑞新材料科技有限公司 | Textile cloth cover defect detection method based on photoelectric detector |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104751443A (en) * | 2014-12-12 | 2015-07-01 | 郑州轻工业学院 | Cotton fault detecting and identifying method based on multi-spectrum technology |
-
2017
- 2017-07-06 CN CN201710547024.2A patent/CN107870172A/en active Pending
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104751443A (en) * | 2014-12-12 | 2015-07-01 | 郑州轻工业学院 | Cotton fault detecting and identifying method based on multi-spectrum technology |
Non-Patent Citations (4)
| Title |
|---|
| 刘晶璟等，: "基于链码的织物疵点图像形态特征提取", 《现代电子技术》 * |
| 周帅: "基于图像处理技术的织物疵点检测与识别技术研究", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
| 王锋: "基于数学形态学的织物疵点检测方法", 《测试技术学报》 * |
| 王静等: "树叶长宽比及边缘曲率的提取", 《华南师范大学学报（自然科学版）》 * |
Cited By (57)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109035195B (en) * | 2018-05-08 | 2021-11-30 | 武汉纺织大学 | Fabric defect detection method |
| CN109035195A (en) * | 2018-05-08 | 2018-12-18 | 武汉纺织大学 | A kind of fabric defect detection method |
| CN108921819B (en) * | 2018-05-29 | 2021-02-02 | 黎明职业大学 | Cloth inspecting device and method based on machine vision |
| CN108921819A (en) * | 2018-05-29 | 2018-11-30 | 黎明职业大学 | A kind of cloth examination device and method based on machine vision |
| CN109100360A (en) * | 2018-07-05 | 2018-12-28 | 福建省国业科技发展有限公司 | The method that ring signal and initial fixed character carry out flaw positioning based on |
| CN109100360B (en) * | 2018-07-05 | 2021-06-22 | 福建省国业科技发展有限公司 | Method for positioning flaw based on lap counting signal and initial fixed characteristic |
| CN109035135A (en) * | 2018-07-13 | 2018-12-18 | 常州宏大智能装备产业发展研究院有限公司 | A kind of on-line automatic pattern adjustment method of fabric based on machine vision |
| CN109344868A (en) * | 2018-08-28 | 2019-02-15 | 广东奥普特科技股份有限公司 | A kind of universal method for distinguishing axisymmetric inhomogeneity object each other |
| CN109344868B (en) * | 2018-08-28 | 2021-11-16 | 广东奥普特科技股份有限公司 | General method for distinguishing different types of objects which are mutually axisymmetric |
| CN109272494A (en) * | 2018-08-31 | 2019-01-25 | 龙山县惹巴妹手工织品有限公司 | A kind of toy watch leather fabric detection method |
| 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 |
| CN109507192A (en) * | 2018-11-02 | 2019-03-22 | 江苏理工学院 | A kind of magnetic core detection method of surface flaw based on machine vision |
| CN109507192B (en) * | 2018-11-02 | 2021-07-02 | 江苏理工学院 | Magnetic core surface defect detection method based on machine vision |
| CN109613006A (en) * | 2018-12-22 | 2019-04-12 | 中原工学院 | A kind of fabric defect detection method based on end-to-end neural network |
| CN109884066A (en) * | 2019-01-22 | 2019-06-14 | 江苏恒力化纤股份有限公司 | A kind of method of offline inspection long filament silk volume greasy dirt fault |
| CN110084246A (en) * | 2019-04-17 | 2019-08-02 | 江南大学 | A kind of yarn dyed fabric fault automatic identifying method |
| CN110097538A (en) * | 2019-04-17 | 2019-08-06 | 江南大学 | A kind of online cloth examination device of loom and defects identification method |
| CN111862125A (en) * | 2019-04-28 | 2020-10-30 | 株式会社理光 | Contour segmentation method and device and computer readable storage medium |
| CN113567447A (en) * | 2019-08-07 | 2021-10-29 | 浙江大学台州研究院 | Synthetic leather hemming online detection method |
| CN110376211A (en) * | 2019-08-07 | 2019-10-25 | 浙江大学台州研究院 | A kind of the synthetic leather crimping on-line measuring device and method of wet process gluing |
| CN111105398A (en) * | 2019-12-19 | 2020-05-05 | 昆明能讯科技有限责任公司 | Transmission line component crack detection method based on visible light image data |
| CN111402225B (en) * | 2020-03-13 | 2023-05-26 | 浙江工业大学 | Cloth folding false-detection defect discriminating method |
| CN111402225A (en) * | 2020-03-13 | 2020-07-10 | 浙江工业大学 | Cloth folding false detection defect discrimination method |
| CN111537519A (en) * | 2020-04-11 | 2020-08-14 | 嘉兴市昊鸣纺织有限公司 | Automatic detection and supervision system for cloth |
| CN111882545A (en) * | 2020-07-30 | 2020-11-03 | 中原工学院 | Fabric defect detection method based on bidirectional information transmission and feature fusion |
| CN111882545B (en) * | 2020-07-30 | 2023-07-25 | 中原工学院 | Fabric defect detection method based on bidirectional information transmission and feature fusion |
| CN111929327A (en) * | 2020-09-09 | 2020-11-13 | 深兰人工智能芯片研究院（江苏）有限公司 | Cloth defect detection method and device |
| CN112365442A (en) * | 2020-10-13 | 2021-02-12 | 江苏恒力化纤股份有限公司 | Image analysis-based rectangular spinneret hole residual impurity detection method |
| CN112365442B (en) * | 2020-10-13 | 2023-06-02 | 江苏恒力化纤股份有限公司 | Rectangular spinneret hole residual impurity detection method based on image analysis |
| CN112365448B (en) * | 2020-10-20 | 2022-08-30 | 天津大学 | Method for detecting fabric defects in warp knitting process |
| CN112365448A (en) * | 2020-10-20 | 2021-02-12 | 天津大学 | Fabric defect detection method in warp knitting process |
| CN112330673B (en) * | 2020-12-11 | 2021-07-06 | 武汉纺织大学 | Woven fabric density detection method based on image processing |
| CN112330673A (en) * | 2020-12-11 | 2021-02-05 | 武汉纺织大学 | Woven fabric density detection method based on image processing |
| CN113139948A (en) * | 2021-04-28 | 2021-07-20 | 福建自贸试验区厦门片区Manteia数据科技有限公司 | Organ contour line quality evaluation method, device and system |
| CN113139948B (en) * | 2021-04-28 | 2023-06-27 | 福建自贸试验区厦门片区Manteia数据科技有限公司 | Organ contour line quality assessment method, device and system |
| CN113888459A (en) * | 2021-05-17 | 2022-01-04 | 浙江大学绍兴微电子研究中心 | Texture image defect detection system and method based on self-adaptive texture features |
| CN113643276A (en) * | 2021-08-29 | 2021-11-12 | 浙江工业大学 | Textile texture defect automatic detection method based on statistical analysis |
| CN113643276B (en) * | 2021-08-29 | 2024-02-02 | 浙江工业大学 | Textile texture defect automatic detection method based on statistical analysis |
| CN113610852A (en) * | 2021-10-10 | 2021-11-05 | 江苏祥顺布业有限公司 | Yarn drafting quality monitoring method based on image processing |
| CN113610852B (en) * | 2021-10-10 | 2021-12-10 | 江苏祥顺布业有限公司 | Yarn drafting quality monitoring method based on image processing |
| CN113638104B (en) * | 2021-10-14 | 2022-02-08 | 海门市恒昌织带有限公司 | Intelligent yarn cleaning control method and system for bobbin winder |
| CN113638104A (en) * | 2021-10-14 | 2021-11-12 | 海门市恒昌织带有限公司 | Intelligent yarn cleaning control method and system for bobbin winder |
| CN114723705B (en) * | 2022-03-31 | 2023-08-22 | 深圳市启灵图像科技有限公司 | Cloth flaw detection method based on image processing |
| CN114723705A (en) * | 2022-03-31 | 2022-07-08 | 海门市恒创织带有限公司 | Cloth flaw detection method based on image processing |
| CN114494259A (en) * | 2022-04-18 | 2022-05-13 | 南通东德纺织科技有限公司 | Cloth defect detection method based on artificial intelligence |
| CN114494259B (en) * | 2022-04-18 | 2022-06-17 | 南通东德纺织科技有限公司 | Cloth defect detection method based on artificial intelligence |
| CN115047070A (en) * | 2022-08-11 | 2022-09-13 | 江苏恒力化纤股份有限公司 | Fabric surface defect detection method based on friction vibration signal |
| CN115047070B (en) * | 2022-08-11 | 2022-12-20 | 江苏恒力化纤股份有限公司 | Fabric surface defect detection method based on friction vibration signal |
| CN115082460A (en) * | 2022-08-18 | 2022-09-20 | 聊城市恒丰电子有限公司 | Weaving production line quality monitoring method and system |
| CN116721105A (en) * | 2023-08-11 | 2023-09-08 | 山东淼珠生物科技有限公司 | Explosive bead production abnormality online detection method based on artificial intelligence |
| CN116721105B (en) * | 2023-08-11 | 2023-10-20 | 山东淼珠生物科技有限公司 | Explosive bead production abnormality online detection method based on artificial intelligence |
| CN117132593A (en) * | 2023-10-25 | 2023-11-28 | 济宁华晟服装股份有限公司 | Cloth roughness detection method for resisting periodic texture influence |
| CN117132593B (en) * | 2023-10-25 | 2024-03-26 | 济宁华晟服装股份有限公司 | Cloth roughness detection method for resisting periodic texture influence |
| CN117475000A (en) * | 2023-12-28 | 2024-01-30 | 江苏恒力化纤股份有限公司 | Fabric selvedge positioning method |
| CN117475000B (en) * | 2023-12-28 | 2024-03-19 | 江苏恒力化纤股份有限公司 | Fabric selvedge positioning method |
| CN117911417A (en) * | 2024-03-20 | 2024-04-19 | 天津市凯瑞新材料科技有限公司 | Textile cloth cover defect detection method based on photoelectric detector |
| CN117911417B (en) * | 2024-03-20 | 2024-05-31 | 天津市凯瑞新材料科技有限公司 | Textile cloth cover defect detection method based on photoelectric detector |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107870172A (en) | A kind of Fabric Defects Inspection detection method based on image procossing | |
| Hanbay et al. | Fabric defect detection systems and methods—A systematic literature review | |
| CN103234976B (en) | Based on the online visible detection method of tricot machine Fabric Defect of Gabor transformation | |
| CN108090494B (en) | Textile defect identification method based on Gabor filter and support vector machine | |
| CN110473201A (en) | A kind of automatic testing method and device of disc surface defect | |
| CN109509171A (en) | A kind of Fabric Defects Inspection detection method based on GMM and image pyramid | |
| CN113435460A (en) | Method for identifying brilliant particle limestone image | |
| Shi et al. | A faster-rcnn based chemical fiber paper tube defect detection method | |
| Xie et al. | Fabric defect detection method combing image pyramid and direction template | |
| CN115601355A (en) | Method and device for detecting and classifying product surface defects and storage medium | |
| Daniel et al. | Automatic road distress detection and analysis | |
| Liang et al. | Automatic defect detection of texture surface with an efficient texture removal network | |
| Meng et al. | Woven fabric density measurement by using multi-scale convolutional neural networks | |
| Abdellah et al. | Defect detection and identification in textile fabric by SVM method | |
| CN114581805A (en) | Coating roller surface defect detection method adopting 3D line laser profile technology | |
| Oni et al. | Patterned fabric defect detection and classification (FDDC) techniques: a review | |
| Gyimah et al. | A discriminative deeplab model (ddlm) for surface anomaly detection and localization | |
| CN113936001A (en) | Textile surface flaw detection method based on image processing technology | |
| Tolba et al. | A self-organizing feature map for automated visual inspection of textile products | |
| Jeffrey Kuo et al. | Self-organizing map network for automatically recognizing color texture fabric nature | |
| Strokina et al. | Detection of curvilinear structures by tensor voting applied to fiber characterization | |
| Wang et al. | Automated inspection of carpets | |
| CN113870328A (en) | Liquid foreign matter visual detection method and system | |
| Miljanovic et al. | Detection of windows in facades using image processing algorithms | |
| Zhang et al. | Draw textured yarn packages hairiness defect detection based on the multi-directional anisotropic gaussian directional derivative |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20180403 |