CN115100481A - Textile qualitative classification method based on artificial intelligence - Google Patents
Textile qualitative classification method based on artificial intelligence Download PDFInfo
- Publication number
- CN115100481A CN115100481A CN202211022235.1A CN202211022235A CN115100481A CN 115100481 A CN115100481 A CN 115100481A CN 202211022235 A CN202211022235 A CN 202211022235A CN 115100481 A CN115100481 A CN 115100481A
- Authority
- CN
- China
- Prior art keywords
- classified
- image
- sampling
- textile
- textiles
- 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.)
- Granted
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/764—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Evolutionary Computation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Computing Systems (AREA)
- Databases & Information Systems (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
Abstract
The invention relates to the technical field of image recognition, in particular to a textile qualitative classification method based on artificial intelligence, which can be used for artificial intelligence systems and artificial intelligence optimization operation systems in the production field, can also be used for application software development such as computer vision software and the like, and comprises the following steps: acquiring each sampling image of the textiles to be classified, and determining the fiber vertical quantity of each sampling image; determining the sampling scale weight of each sampling image according to the fabric type of the textiles to be classified and the sampling scale of each sampling image; and determining the classification coefficient of the textiles to be classified according to the sampling scale weight and the fiber vertical quantity of each sampling image, and further determining the types of the textiles to be classified. The invention can be applied to corresponding electronic equipment, and the electronic equipment realizes the qualitative classification of textiles to be classified through the processes of calculation, data processing and the like by utilizing an image recognition technology.
Description
Technical Field
The invention relates to the technical field of image recognition, in particular to a textile qualitative classification method based on artificial intelligence.
Background
In the textile industry, the accurate classification of textile crafts is about important steps such as subsequent packaging, finishing and dyeing, for example, warp-knitted textile crafts and weft-knitted textile crafts with fabric types are completely different in dyeing and finishing steps and ironing steps, and some important production parameters of blended textile crafts and interwoven textile crafts with fabric types are also different in setting. Therefore, in order to ensure the quality of textile handicrafts, it is necessary to qualitatively classify the textile handicrafts when they are manufactured.
The traditional classification mode is to use the gray level co-occurrence matrix in the image recognition technology to perform qualitative classification, namely to perform classification through various statistical characteristics of the gray level co-occurrence matrix, which is essentially an empirical summary of statistical indexes of a large amount of image data, and the accuracy of the method is low. With the development of computer vision technology and artificial intelligence systems, in the prior art, besides a method of manual visual inspection, a method of qualitative classification of textile artworks by using computer vision, artificial intelligence and other technologies appears, for example, an artificial neural network algorithm is used for qualitative classification of textile artworks, but the method needs a large amount of image data with labels to train a network, the data cost is high, the labels need to be manually marked, manpower resources are wasted, the workload of workers is increased, and the artificial neural network algorithms corresponding to the textile artworks of different fabric types are different, so that the integration of the classification methods cannot be realized, and the applicability of the qualitative classification of the textile artworks is low.
Disclosure of Invention
In order to solve the problem of low applicability of the existing textile handicraft qualitative classification method, the invention aims to provide a textile qualitative classification method based on artificial intelligence.
The invention provides a textile qualitative classification method based on artificial intelligence, which comprises the following steps:
acquiring a surface image of a textile to be classified, and performing multi-scale down sampling on the surface image to obtain each sampling image and the sampling scale thereof;
determining a gray level run matrix of each angle corresponding to each sampling image according to each sampling image, and further determining a fiber vertical quantity corresponding to each sampling image;
acquiring the fabric type of the textile to be classified, and determining the sampling scale weight corresponding to each sampling image according to the fabric type of the textile to be classified and the sampling scale corresponding to each sampling image;
determining the classification coefficient of the textiles to be classified according to the sampling scale weight and the fiber vertical quantity corresponding to each sampling image;
and determining the type of the textiles to be classified according to the classification coefficient of the textiles to be classified.
Further, the step of determining the fiber verticality corresponding to each sampling image comprises:
determining a first gray level run difference matrix and a second gray level run difference matrix corresponding to each sampling image according to the gray level run matrix of each angle corresponding to each sampling image;
and determining the fiber vertical quantity corresponding to each sampling image according to the first gray level run-length difference matrix and the second gray level run-length difference matrix corresponding to each sampling image and each sampling image.
Further, a calculation formula for determining the fiber vertical amount corresponding to each sampling image is determined:
wherein, the first and the second end of the pipe are connected with each other,
is as followsiThe amount of fiber verticality corresponding to each sampled image,
is as followsiThe values of the elements with the gray level k and the run length d in the first gray level run difference matrix corresponding to each sampled image,
first, theiThe gray level of the second gray level run-length difference matrix corresponding to each sampling image is K, and the value of the element with the run length of d is KiThe number of gray scale levels in the first gray scale run-length difference matrix and the second gray scale run-length difference matrix corresponding to each sampled image, D is the secondiAnd the maximum value of the run length in the first gray level run difference matrix and the second gray level run difference matrix corresponding to each sampling image.
Further, if the fabric type of the textile to be classified is a knitted fabric, determining a calculation formula of sampling scale weights corresponding to the sampling images:
wherein the content of the first and second substances,
when the fabric type of the textile to be classified is knitted fabriciThe sampling scale weights corresponding to the individual sampled images,
is a firstiThe corresponding sampling scale of each of the sampled images,
is the maximum value in the corresponding sampling scale of each sampling image.
Further, if the fabric type of the textiles to be classified is woven fabric, determining a calculation formula of the sampling scale weight corresponding to each sampling image:
wherein the content of the first and second substances,
when the fabric type of the textile to be classified is woven fabriciThe sampling scale weights corresponding to the individual sampled images,
is as followsiThe corresponding sampling scale of each of the sampled images,
is the maximum value in the sampling scale corresponding to each sampling image.
Further, if the fabric type of the textiles to be classified is knitted fabric, determining a calculation formula of the classification coefficient of the textiles to be classified:
wherein the content of the first and second substances,
the classification coefficient of the textile to be classified when the type of the textile to be classified is the knitted fabric,
when the fabric type of the textile to be classified is knitted fabriciThe sampling scale weights corresponding to the individual sampled images,
is as followsiThe amount of fiber verticality corresponding to each sampled image,Ithe number of each sampled image.
Further, if the fabric type of the textiles to be classified is woven fabric, determining a calculation formula of the classification coefficient of the textiles to be classified:
wherein, the first and the second end of the pipe are connected with each other,
the classification coefficient of the textiles to be classified is determined when the types of the textiles to be classified are woven fabrics,
when the fabric type of the textile to be classified is woven fabriciThe sampling scale weights corresponding to the individual sampled images,
is a firstiThe amount of fiber verticality corresponding to each sampled image,Ithe number of each sampled image.
Further, the step of determining the kind of textiles to be classified comprises:
when the fabric type of the textiles to be classified is knitted fabrics, if the classification coefficient of the textiles to be classified of the knitted fabrics is larger than or equal to a first preset classification coefficient and smaller than or equal to a second preset classification coefficient, the textiles to be classified are warp-knitted textiles, and if the classification coefficient of the textiles to be classified of the knitted fabrics is larger than the second preset classification coefficient and smaller than or equal to a third preset classification coefficient, the textiles to be classified are weft-knitted textiles;
when the fabric type of the textile to be classified is woven fabric, if the classification coefficient of the textile to be classified of the woven fabric is greater than or equal to a first preset classification coefficient and less than or equal to a second preset classification coefficient, the textile to be classified is a blended machine textile, and if the classification coefficient of the textile to be classified of the woven fabric is greater than the second preset classification coefficient and less than or equal to a third preset classification coefficient, the textile to be classified is an interweaving machine textile.
Further, the step of determining the gray level run matrix of each angle corresponding to each sampled image includes:
performing graying processing on each sampling image according to each sampling image to obtain each sampling image after graying processing;
determining the gray level corresponding to each gray value in each sampling image according to each sampling image after the graying processing;
and determining the gray run matrix of each angle corresponding to each sampling image according to the gray level corresponding to each gray value in each sampling image and each sampling image.
Further, the step of determining the gray level corresponding to each gray value in each sampled image includes:
obtaining a gray level histogram of each sampled image according to each sampled image after the graying processing;
obtaining a one-dimensional Gaussian mixture model of each sampled image according to the gray level histogram of each sampled image, and further obtaining each sub-Gaussian model corresponding to each sampled image;
and acquiring the weight of each sub-Gaussian model, and determining the gray level corresponding to each gray value in each sampled image according to the gray histogram of each sampled image, each sub-Gaussian model corresponding to each sampled image and the weight of each sub-Gaussian model.
The invention has the following beneficial effects:
the invention provides a textile qualitative classification method based on artificial intelligence, which can be used for an artificial intelligence system, an artificial intelligence optimization operation system and an artificial intelligence middleware in the production field, and particularly relates to a method for identifying each acquired sampling image of a textile to be classified by electronic equipment by utilizing an image identification technology so as to obtain the fiber vertical quantity and the sampling scale weight of each sampling image, determining the classification coefficient of the textile to be classified according to the fiber vertical quantity and the sampling scale weight of each sampling image, and further determining the type of the textile to be classified.
The invention combines the operation results of the textile qualitative classification into one type, namely, the type of the textile to be classified can be determined through the classification coefficient of the textile to be classified, and the applicability of the textile qualitative classification is improved. In addition, the invention obtains each sampling image by carrying out multi-scale down sampling on the surface images of the textiles, and determines the types of the textiles to be classified by carrying out data processing on the image characteristics in each sampling image, thereby effectively improving the accuracy of qualitative classification of the textiles to be classified. Therefore, the textile qualitative classification method based on artificial intelligence provided by the invention can be used for developing application software such as computer vision software.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions and advantages of the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments of the present invention, and other drawings can be obtained by those skilled in the art without creative efforts.
FIG. 1 is a flow chart of a method for qualitative classification of textiles according to the invention based on artificial intelligence;
fig. 2 is a flowchart of determining a gray scale level corresponding to each gray scale value in each sample image according to an embodiment of the present invention.
Detailed Description
To further explain the technical means and effects of the present invention adopted to achieve the predetermined objects, the following detailed description of the embodiments, structures, features and effects of the technical solutions according to the present invention will be given with reference to the accompanying drawings and preferred embodiments. In the following description, different references to "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, the particular features, structures, or characteristics may be combined in any suitable manner in one or more embodiments.
Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.
In the prior art, a corresponding classification method is designed according to different classification characteristics of textiles, so that the purpose of qualitatively classifying the textiles is achieved by acquiring various types of image characteristics of textile images, namely, the prior art is adapted to the qualitative classification of textiles of different fabric types by acquiring various types of image characteristics of the textile images. For example, the classification according to the fiber type can be classified into blending and interweaving, that is, the type of the textile whose fabric type is woven fabric includes blending and interweaving, and the classification according to the weaving manner can be classified into weft knitting and warp knitting, that is, the type of the textile whose fabric type is knitted fabric includes weft knitting and warp knitting. The qualitative classification of the textiles of these different fabric types requires a separate design of the method of extracting image features and the method of classifying image features, which is time-consuming and labor-consuming. Based on the analysis, the implementation utilizes the same type of image characteristics to qualitatively classify the textiles with the fabric types of woven fabrics and knitted fabrics.
The embodiment provides a textile qualitative classification method based on artificial intelligence, and a corresponding flow chart is shown in fig. 1, and the method comprises the following steps:
(1) and acquiring a surface image of the textile to be classified, and performing multi-scale down sampling on the surface image to obtain each sampling image and the sampling scale thereof.
In the process of qualitatively classifying textiles, firstly, the textiles to be classified are shot by an industrial camera to obtain a surface image of the textiles to be classified, the image specification of the surface image is that the length and the width are equal, namely the shape of the surface image is a square. And then, carrying out multi-scale pyramid down-sampling on the surface images of the textiles to be classified to obtain each sampling image, wherein each sampling image has a corresponding sampling scale. Marking individual sample images as
，
，IIn order to perform the sampling times of the multiscale pyramid downsampling on the surface image of the textile to be classified,
the surface image before the down-sampling of the multi-scale pyramid is performed, that is, the sampling image with the sampling scale of 0. The process of performing multiscale pyramid downsampling on a surface image is prior art and is not within the scope of the present invention, and is not described in detail herein.
(2) According to each sampling image, determining a gray level run matrix of each angle corresponding to each sampling image, and further determining a fiber vertical quantity corresponding to each sampling image, wherein the method comprises the following steps of:
and (2-1) determining a gray run matrix of each angle corresponding to each sampling image according to each sampling image.
Firstly, it should be noted that if the fabric type of the textile is knitted fabric, the gray level run in the gray level run matrix corresponding to the surface image of the textile represents the knitting direction of the textile, the knitted woolen loops of the textile are different in the degree of protruding from the surface of the textile in different knitting directions, and the gray level of the surface of the textile is different when the illumination conditions of the whole textile surface are the same; if the fabric type of the textile is woven fabric, different fibers in the yarns of the textile have different light reflection capabilities, that is, the gray scales corresponding to different fibers are different, the gray scale run in the gray scale run matrix corresponding to the surface image of the textile represents the continuous occurrence length of elements with the same gray scale value in a certain direction in the textile, and the larger the length of the gray scale run is, the larger the probability that a certain type of fibers or loops of wool appear in the direction is. In this embodiment, in order to subsequently facilitate determining the types of textiles to be classified of different fabric types, it is necessary to determine the gray level run matrix of each angle corresponding to each sampled image, that is, determine the gray level run information in the gray level run matrix of each angle, and the steps include:
and (2-1-1) carrying out graying processing on each sampling image according to each sampling image to obtain each sampling image after graying processing.
In order to determine the gray value of each pixel point in each sampled image, the present embodiment performs graying processing on each sampled image obtained in step (1), so as to obtain each grayed sampled image. The graying process is prior art and is not within the scope of the present invention, and will not be described in detail herein.
(2-1-2) determining the gray level corresponding to each gray value in each sampled image according to each sampled image after the graying processing, wherein a flow chart is shown as fig. 2, and the steps comprise:
and (2-1-2-1) obtaining a gray level histogram of each sampling image according to each sampling image after the graying processing.
And (3) performing gray histogram statistics on each grayed sampling image according to each grayed sampling image obtained in the step (2-1-1), so as to obtain a gray histogram of each sampling image, wherein the gray histogram represents the frequency of each gray value in the sampling image in the whole sampling image. The process of gray histogram statistics is prior art and is not within the scope of the present invention, and will not be described in detail herein.
(2-1-2-2) obtaining a one-dimensional Gaussian mixture model of each sampled image according to the gray level histogram of each sampled image, and further obtaining each sub-Gaussian model corresponding to each sampled image.
In this embodiment, first, the frequency of each gray value in each sample image appearing in the entire sample image is obtained from the gray histogram of each sample image. Then, each gray value in each sampling image and the frequency of the gray value appearing in the whole sampling image are taken as sample data, and an EM (Expectation-maximization) algorithm is utilized to fit the gray histogram of each sampling image, so that a one-dimensional Gaussian mixture model corresponding to each sampling image is obtained, wherein the one-dimensional Gaussian mixture model can describe the probability of each gray value in each sampling image appearing in the whole sampling image. And finally, obtaining each sub-Gaussian model corresponding to each sampling image according to the one-dimensional Gaussian mixture model corresponding to each sampling image.
It should be noted that the number of sub-gaussian models in the one-dimensional gaussian mixture model isKThe number of sub-Gaussian models is a self-defined parameter, and the embodiment will be describedKSetting the number as 10, and arranging the sub-Gaussian models from large to small according to the average value, wherein the serial numbers of the sub-Gaussian models are marked as 1,2, …,K. The process of obtaining a one-dimensional gaussian mixture model by fitting using an EM algorithm is prior art and is not within the scope of the present invention, and is not described in detail herein.
(2-1-2-3) obtaining the weight of each sub-Gaussian model, and determining the gray level corresponding to each gray value in each sampled image according to the gray histogram of each sampled image, each sub-Gaussian model corresponding to each sampled image and the weight of each sub-Gaussian model.
In this embodiment, each sub-gaussian model corresponding to each sampled image has a weight corresponding to each sub-gaussian model, and according to the gray level histogram of each sampled image, each sub-gaussian model corresponding to each sampled image and its weight, the probability of each gray level value in each sampled image appearing in the whole sampled image can be calculated, and the calculation formula is as follows:
wherein the content of the first and second substances,
for the first in each sampled imageHThe probability of an individual gray value appearing in the entire sampled image,
for each sampled imagekThe weight of the sub-gaussian model is,
for the first in each sampled imageHThe gray value of the image is at the second of the corresponding sampling imagekThe probability value of the sub-gaussian model,Kthe number of sub-gaussian models for each sampled image.
According to the calculation formula of the probability of each gray value in each sampling image appearing in the whole sampling image, each gray value in the sampling image corresponds to each gray valueKAn
ObtainingKAn
The sequence number of the sub-gaussian model corresponding to the maximum value is recorded as the gray level of the corresponding gray value, so as to obtain the gray level corresponding to each gray value in each sampled image.
To this end, for each gray value in each sample image, there is one gray level, and there are multiple gray levels in each sample image.
And (2-1-3) determining a gray run matrix of each angle corresponding to each sampling image according to the gray level corresponding to each gray value in each sampling image and each sampling image.
Through each sampling image obtained in the step (1), the size of each sampling image can be determined, wherein the size refers to the length and the width of each sampling image, and each sampling image is formedSince the shape of the image is square, the length and width of each sample image are equal, and the size of each sample image is described asD. According to the size of each sampled imageDAnd (3) constructing a gray run matrix corresponding to each sampling image according to the gray level corresponding to each gray value in each sampling image obtained in the step (2-1-2). The gray level in the gray level run matrix corresponding to each sampling image is from 1 toKRun length from 1 toD，KThe number of gray level levels in the gray scale run matrix, i.e. the maximum value among the sequence numbers of the sub-gaussian models of each sampled image,Dis the maximum value of the run length in the gray scale run matrix, i.e. the size of each sampled image. The process of constructing the gray scale run matrix is prior art and is not within the scope of the present invention, and will not be described in detail herein.
In this embodiment, the gray level run matrix can be divided into 4 types according to the statistical angles of the gray level runs, and the statistical angles are 0 °,45 °,90 °, and 135 °, respectively. So far, each sampled image corresponds to a 4-angle gray scale run matrix.
(2-2) determining the vertical fiber quantity corresponding to each sampling image according to the gray level run matrix of each angle corresponding to each sampling image, wherein the method comprises the following steps:
and (2-2-1) determining a first gray level run-length difference matrix and a second gray level run-length difference matrix corresponding to each sampling image according to the gray level run-length matrix of each angle corresponding to each sampling image.
Firstly, it should be noted that determining the first gray level run-length difference matrix and the second gray level run-length difference matrix corresponding to each sampled image is helpful for subsequently determining the type of textiles to be classified, because:
when the textiles to be classified are of the knitwear fabric types, the weaving modes of the textiles to be classified need to be distinguished, and the weaving modes comprise weft knitting and warp knitting. The loops of the warp-knitted textile extend in a zigzag manner, namely the runs of the same gray level in the gray level run matrix of each angle corresponding to the surface image of the warp-knitted textile are perpendicular to each other, and the loops of the weft-knitted textile are arranged in a reciprocating parallel manner, namely the runs of the same gray level in the gray level run matrix of each angle corresponding to the surface image of the weft-knitted textile are not perpendicular to each other.
When the textiles to be classified are of the types of the woven fabrics, the fiber types of the textiles to be classified need to be distinguished, and the fiber types comprise blending and interweaving. The warps and the wefts of the same type of fibers of the blended textile are perpendicular to each other, namely the runs with the same gray level in the gray level run matrix of each angle corresponding to the surface image of the blended textile are perpendicular to each other, while the warps and the wefts of the same type of fibers of the interwoven textile are not perpendicular to each other, namely the runs with the same gray level in the gray level run matrix of each angle corresponding to the surface image of the interwoven textile are not perpendicular to each other, wherein the warps and the wefts refer to the yarns of the textile, and the yarns of the textile are composed of a plurality of fibers.
In the present embodiment, to determineiTaking the first gray level run difference matrix and the second gray level run difference matrix corresponding to each sampled image as an example, the first gray level run difference matrix and the second gray level run difference matrix are obtained according to the step (2-1)iThe gray scale run matrix of 4 angles corresponding to each sampling image is obtainediThe gray level run matrixes of 4 angles corresponding to the sampling images are respectively marked as
Calculating the firstiThe calculation formula of the first gray level run difference matrix and the second gray level run difference matrix corresponding to each sampling image is as follows:
wherein the content of the first and second substances,
is as followsiA first gray scale run-length difference matrix for each sampled image,
is as followsiThe gray scale run matrix with the angle of 0 degree is corresponding to each sampling image,
is a firstiThe sampled images correspond to a gray scale run matrix with an angle of 90 degrees,
first, theiA second gray scale travel difference matrix for each sampled image,
is a firstiThe sampled images correspond to a gray scale run matrix with an angle of 45 degrees,
first, theiThe gray scale run matrix with the angle of 135 degrees corresponds to each sampling image.
Reference toiAnd determining a first gray level run difference matrix and a second gray level run difference matrix corresponding to each sampling image to obtain the first gray level run difference matrix and the second gray level run difference matrix corresponding to each sampling image. So far, each sampling image corresponds to two gray level run-length difference matrixes.
In addition, the first stepiFirst gray level run-length difference matrix corresponding to each sampling image
And a second gray scale travel difference matrix
The inner element is the absolute value of the difference value of the corresponding element in the gray level run matrix of two different angles, the first one isiThe more the numerical value of the element in the gray level run difference matrix of the sampled image tends to be 0, the more the numerical value can indicate that the runs of all the same gray level in the gray level run matrix of each angle corresponding to the sampled image are mutually vertical, namely, the more the types of the textiles to be classified are likely to be warp knitting or mixedAnd (4) spinning.
And (2-2-2) determining the fiber vertical quantity corresponding to each sampling image according to the first gray level run-length difference matrix and the second gray level run-length difference matrix corresponding to each sampling image and each sampling image.
In this embodiment, the fiber verticality corresponding to each sampling image is calculated according to the first gray level run-length difference matrix and the second gray level run-length difference matrix corresponding to each sampling image obtained in step (2-2-1) and the size of each sampling image obtained in step (2-1-3), and the calculation formula is as follows:
wherein the content of the first and second substances,
is a firstiThe amount of fiber verticality corresponding to each sampled image,
is as followsiThe values of the elements with the gray level k and the run length d in the first gray level run difference matrix corresponding to each sampled image,
first, theiThe gray level of a second gray level run-length difference matrix corresponding to each sampling image is K, the run length is d, and K is the number of the elementsiThe number of gray scale levels in the first gray scale run-length difference matrix and the second gray scale run-length difference matrix corresponding to each sampled image, D is the secondiThe maximum value of the run length in the first gray level run difference matrix and the second gray level run difference matrix corresponding to each sampled image, i.e. the firstiThe size of the sample image.
Therefore, each sampling image corresponds to a fiber vertical quantity, and it should be noted that the more the fiber vertical quantity corresponding to the sampling image tends to be 0, the more the number of mutually perpendicular runs with the same gray level in the gray level run matrix of each angle corresponding to the sampling image is illustrated to be balanced, that is, the more the yarn fibers of the textiles to be classified are perpendicular to each other, otherwise, the yarn fibers of the textiles to be classified are unidirectionally distributed, that is, the yarn fibers of the textiles to be classified are not perpendicular.
(3) And obtaining the fabric type of the textile to be classified, and determining the sampling scale weight corresponding to each sampling image according to the fabric type of the textile to be classified and the sampling scale corresponding to each sampling image.
First, when textiles of different fabric types are qualitatively classified, the requirements on the attention degree of the sampling scale corresponding to the sampling image are different.
When the type of the textile of the knitted fabric type is determined, the sampling image with a larger sampling scale is more concerned, namely when the textile of the knitted fabric type is qualitatively classified, the whole sampling image is more concerned, and the vertical information in the sampling image with the larger sampling scale represents the difference between warp knitting and weft knitting; when the type of the textile of the type of the woven fabric is determined, the sampling image with a smaller sampling scale is more concerned, namely when the textile of the type of the woven fabric is qualitatively classified, the detail content in the sampling image is more concerned, and the vertical information in the sampling image with the smaller sampling scale represents the difference between blending and interweaving. In addition, the smaller the sampling scale of the sampling image is, the more the vertical information of the yarn fiber of the textile of the fabric type can be obtained, that is, the smaller the sampling scale of the sampling image is, the clearer and truer the detail content in the sampling image is. Therefore, the fabrics of the textiles to be classified are different in type, the calculation modes of the sampling scale weights corresponding to the sampling images are different, and the specific contents comprise:
(3-1) if the fabric type of the textile to be classified is a knitted fabric, determining a calculation formula of sampling scale weight corresponding to each sampling image:
wherein the content of the first and second substances,
when the fabric type of the textile to be classified is knitted fabriciThe sampling scale weights corresponding to the individual sampled images,
is as followsiThe corresponding sampling scale of each of the sampled images,
is the maximum value in the corresponding sampling scale of each sampling image.
(3-2) if the fabric type of the textile to be classified is woven fabric, determining a calculation formula of sampling scale weight corresponding to each sampling image:
wherein the content of the first and second substances,
when the fabric type of the textile to be classified is woven fabriciThe corresponding sample scale weights for each of the sampled images,
is as followsiThe corresponding sampling scale of each of the sampled images,
is the maximum value in the sampling scale corresponding to each sampling image.
(4) And determining the classification coefficient of the textiles to be classified according to the sampling scale weight and the fiber vertical amount corresponding to each sampling image.
In this embodiment, the classification coefficient of the textile to be classified when the textile to be classified is a different fabric type is calculated according to the sampling scale weight corresponding to each sampling image when the fabric type of the textile to be classified obtained in step (3) is a knitted fabric, the sampling scale weight corresponding to each sampling image when the fabric type of the textile to be classified is a woven fabric, and the fiber vertical amount corresponding to each sampling image obtained in step (2-2-2), and the specific content includes:
(4-1) if the fabric type of the to-be-classified textile is a knitted fabric, determining a calculation formula of the classification coefficient of the to-be-classified textile:
wherein the content of the first and second substances,
the classification coefficient of the textiles to be classified when the types of the textiles to be classified are knitted textiles,
when the fabric type of the textile to be classified is knitted fabriciThe sampling scale weights corresponding to the individual sampled images,
is a firstiThe amount of fiber verticality corresponding to each sampled image,Ithe number of each sampled image.
(4-2) if the fabric type of the textiles to be classified is woven fabric, determining a calculation formula of the classification coefficient of the textiles to be classified:
wherein the content of the first and second substances,
the classification coefficient of the textiles to be classified is determined when the types of the textiles to be classified are woven fabrics,
is to be divided intoWhen the fabric type of the textile-like product is woven fabriciThe sampling scale weights corresponding to the individual sampled images,
is as followsiThe amount of fiber verticality corresponding to each sampled image,Ithe number of each sampled image.
And determining the classification coefficient of the textiles to be classified according to the surface image fabric types of the textiles to be classified. It should be noted that the classification coefficient of the textiles to be classified is the classification coefficient of the textiles to be classified when the fabric type of the textiles to be classified is the knitted fabric
And the classification coefficient of the textiles to be classified when the fabric types of the textiles to be classified are woven fabrics
Are all normalized values.
(5) And determining the type of the textiles to be classified according to the classification coefficient of the textiles to be classified.
In this embodiment, the type of the textile to be classified is determined according to the classification coefficient of the textile to be classified when the type of the textile to be classified obtained in step (4) is a knitted fabric and the classification coefficient of the textile to be classified when the type of the textile to be classified is a woven fabric, and the specific contents are as follows:
(5-1) when the fabric type of the to-be-classified textile is a knitted fabric, if the classification coefficient of the to-be-classified textile of the knitted fabric is greater than or equal to a first preset classification coefficient and less than or equal to a second preset classification coefficient, the to-be-classified textile is a warp knitted fabric textile, and if the classification coefficient of the to-be-classified textile of the knitted fabric is greater than the second preset classification coefficient and less than or equal to a third preset classification coefficient, the to-be-classified textile is a weft knitted fabric textile.
(5-2) when the fabric type of the textile to be classified is woven fabric, if the classification coefficient of the textile to be classified of the woven fabric is greater than or equal to a first preset classification coefficient and less than or equal to a second preset classification coefficient, the textile to be classified is a blended machine textile, and if the classification coefficient of the textile to be classified of the woven fabric is greater than the second preset classification coefficient and less than or equal to a third preset classification coefficient, the textile to be classified is an interweaving machine textile.
In this embodiment, the first preset classification coefficient, the second preset classification coefficient, and the third preset classification coefficient are sequentially increased, the first preset classification coefficient is set to 0, the second preset classification coefficient is set to 0.5, and the third preset classification coefficient is set to 1. Thus, this time: if it is
∈[0,0.5]The textile to be classified is a warp-knitted textile; if it is
∈(0.5,1]The textile to be classified is a weft knitted textile. If it is
∈[0,0.5]The textile to be classified is the textile of the blended machine fabric; if it is
∈(0.5,1]The textile to be classified is an interwoven machine fabric textile.
According to the invention, each sampling image is obtained by applying electronic equipment identification, the vertical fiber amount of each sampling image and the sampling scale weight of each sampling image corresponding to different fabric types are determined, the classification coefficient of the textiles to be classified corresponding to different fabric types is further determined, and the types of the textiles to be classified are determined according to the classification coefficient of the textiles to be classified. The textile qualitative classification method based on artificial intelligence provided by the invention can be suitable for qualitative classification of textiles of different fabric types, simplifies the textile qualitative classification process, improves the applicability of textile qualitative classification, and can be used for development of application software such as artificial intelligence systems and computer vision software in the production field.
The above-mentioned embodiments are only used to illustrate the technical solutions of the present application, and not to limit the same; although the present application has been described in detail with reference to the foregoing embodiments, it should be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; such modifications and substitutions do not substantially depart from the spirit and scope of the embodiments of the present application and are intended to be included within the scope of the present application.
Claims (10)
1. A textile qualitative classification method based on artificial intelligence is characterized by comprising the following steps:
acquiring a surface image of a textile to be classified, and performing multi-scale down sampling on the surface image to obtain each sampling image and the sampling scale thereof;
determining a gray level run matrix of each angle corresponding to each sampling image according to each sampling image, and further determining a fiber vertical quantity corresponding to each sampling image;
acquiring the fabric type of the textile to be classified, and determining the sampling scale weight corresponding to each sampling image according to the fabric type of the textile to be classified and the sampling scale corresponding to each sampling image;
determining a classification coefficient of the textiles to be classified according to the sampling scale weight and the fiber vertical quantity corresponding to each sampling image;
and determining the type of the textiles to be classified according to the classification coefficient of the textiles to be classified.
2. The method of claim 1, wherein the step of determining the fiber verticality of each sampled image comprises:
determining a first gray level run difference matrix and a second gray level run difference matrix corresponding to each sampling image according to the gray level run matrix of each angle corresponding to each sampling image;
and determining the fiber vertical quantity corresponding to each sampling image according to the first gray level run-length difference matrix and the second gray level run-length difference matrix corresponding to each sampling image and each sampling image.
3. The method for qualitatively classifying textiles according to claim 2, characterised in that the formula for determining the vertical fiber content corresponding to each sample image is:
wherein the content of the first and second substances,
is as followsiThe amount of fiber verticality corresponding to each sampled image,
is as followsiThe values of the elements with the gray level k and the run length d in the first gray level run difference matrix corresponding to each sample image,
first, theiThe gray level of the second gray level run-length difference matrix corresponding to each sampling image is K, and the value of the element with the run length of d is KiThe number of gray scale levels in the first gray scale run-length difference matrix and the second gray scale run-length difference matrix corresponding to each sampled image, D is the secondiAnd the maximum value of the run length in the first gray level run difference matrix and the second gray level run difference matrix corresponding to each sampling image.
4. The method for qualitatively classifying textiles based on artificial intelligence as claimed in claim 1, wherein if the type of the textile to be classified is a knitted textile, the formula for calculating the sampling scale weight corresponding to each sampling image is determined:
wherein, the first and the second end of the pipe are connected with each other,
when the fabric type of the textile to be classified is knitted fabriciThe sampling scale weights corresponding to the individual sampled images,
is as followsiThe corresponding sampling scale of each of the sampled images,
is the maximum value in the corresponding sampling scale of each sampling image.
5. The method for qualitatively classifying textiles based on artificial intelligence as claimed in claim 1 is characterized in that if the type of the textile to be classified is woven fabric, the calculation formula of the sampling scale weight corresponding to each sampling image is determined:
wherein the content of the first and second substances,
when the fabric type of the textile to be classified is woven fabriciThe corresponding sample scale weights for each of the sampled images,
is as followsiThe corresponding sampling scale of each of the sampled images,
is the maximum value in the corresponding sampling scale of each sampling image.
6. The method for qualitatively classifying textiles based on artificial intelligence as claimed in claim 1, wherein if the type of the textile to be classified is a knitted textile, the calculation formula of the classification coefficient of the textile to be classified is determined:
wherein the content of the first and second substances,
the classification coefficient of the textile to be classified when the type of the textile to be classified is the knitted fabric,
when the fabric type of the textile to be classified is knitted fabriciThe sampling scale weights corresponding to the individual sampled images,
is as followsiThe amount of fiber verticality corresponding to each sampled image,Ithe number of each sampled image.
7. The method for qualitatively classifying textiles based on artificial intelligence as claimed in claim 1, characterised in that if the type of the textile to be classified is woven, the calculation formula of the classification coefficient of the textile to be classified is determined:
wherein the content of the first and second substances,
the classification coefficient of the textiles to be classified is determined when the types of the textiles to be classified are woven fabrics,
when the fabric type of the textiles to be classified is woven fabriciThe corresponding sample scale weights for each of the sampled images,
is as followsiThe amount of fiber verticality corresponding to each sampled image,Ithe number of each sampled image.
8. The method for qualitatively classifying textiles according to claim 1, wherein the step of determining the type of textiles to be classified comprises:
when the fabric type of the textiles to be classified is knitted fabrics, if the classification coefficient of the textiles to be classified of the knitted fabrics is larger than or equal to a first preset classification coefficient and smaller than or equal to a second preset classification coefficient, the textiles to be classified are warp-knitted textiles, and if the classification coefficient of the textiles to be classified of the knitted fabrics is larger than the second preset classification coefficient and smaller than or equal to a third preset classification coefficient, the textiles to be classified are weft-knitted textiles;
when the fabric type of the textile to be classified is woven fabric, if the classification coefficient of the textile to be classified of the woven fabric is greater than or equal to a first preset classification coefficient and less than or equal to a second preset classification coefficient, the textile to be classified is a blended machine textile, and if the classification coefficient of the textile to be classified of the woven fabric is greater than the second preset classification coefficient and less than or equal to a third preset classification coefficient, the textile to be classified is an interweaving machine textile.
9. The method for qualitatively classifying textiles according to claim 1, wherein the step of determining the gray level run matrix for each angle corresponding to each sampled image comprises:
performing graying processing on each sampling image according to each sampling image to obtain each sampling image after graying processing;
determining the gray level corresponding to each gray value in each sampling image according to each sampling image after the graying processing;
and determining a gray run matrix of each angle corresponding to each sampling image according to the gray level corresponding to each gray value in each sampling image and each sampling image.
10. The artificial intelligence based qualitative classification method of textile products according to claim 9, characterized in that the step of determining the grey scale level corresponding to each grey scale value in the respective sampled images comprises:
obtaining a gray level histogram of each sampled image according to each sampled image after the graying processing;
obtaining a one-dimensional Gaussian mixture model of each sampled image according to the gray level histogram of each sampled image, and further obtaining each sub-Gaussian model corresponding to each sampled image;
and acquiring the weight of each sub-Gaussian model, and determining the gray level corresponding to each gray value in each sampled image according to the gray histogram of each sampled image, each sub-Gaussian model corresponding to each sampled image and the weight of each sub-Gaussian model.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211022235.1A CN115100481B (en) | 2022-08-25 | 2022-08-25 | Textile qualitative classification method based on artificial intelligence |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211022235.1A CN115100481B (en) | 2022-08-25 | 2022-08-25 | Textile qualitative classification method based on artificial intelligence |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115100481A true CN115100481A (en) | 2022-09-23 |
| CN115100481B CN115100481B (en) | 2022-11-18 |
Family
ID=83301246
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211022235.1A Active CN115100481B (en) | 2022-08-25 | 2022-08-25 | Textile qualitative classification method based on artificial intelligence |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115100481B (en) |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20190205758A1 (en) * | 2016-12-30 | 2019-07-04 | Konica Minolta Laboratory U.S.A., Inc. | Gland segmentation with deeply-supervised multi-level deconvolution networks |
| CN114220012A (en) * | 2021-12-16 | 2022-03-22 | 池明旻 | Textile cotton and linen identification method based on deep self-attention network |
| CN114913178A (en) * | 2022-07-19 | 2022-08-16 | 山东天宸塑业有限公司 | Melt-blown fabric defect detection method and system |
-
2022
- 2022-08-25 CN CN202211022235.1A patent/CN115100481B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20190205758A1 (en) * | 2016-12-30 | 2019-07-04 | Konica Minolta Laboratory U.S.A., Inc. | Gland segmentation with deeply-supervised multi-level deconvolution networks |
| CN114220012A (en) * | 2021-12-16 | 2022-03-22 | 池明旻 | Textile cotton and linen identification method based on deep self-attention network |
| CN114913178A (en) * | 2022-07-19 | 2022-08-16 | 山东天宸塑业有限公司 | Melt-blown fabric defect detection method and system |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115100481B (en) | 2022-11-18 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Silvestre-Blanes et al. | A public fabric database for defect detection methods and results | |
| Hanbay et al. | Fabric defect detection systems and methods—A systematic literature review | |
| CN107870172A (en) | A kind of Fabric Defects Inspection detection method based on image procossing | |
| CN113724241B (en) | Broken filament detection method and device for carbon fiber warp-knitted fabric and storage medium | |
| Pan et al. | Automatic detection of structure parameters of yarn-dyed fabric | |
| CN107274385A (en) | Textile flaw detection method based on class grid pattern and its Gabor characteristic | |
| CN114693677B (en) | Knitted fabric pad dyeing process abnormity detection method | |
| Hanmandlu et al. | Detection of defects in fabrics using topothesy fractal dimension features | |
| CN106770323A (en) | Based on the textile flaw detection method that hierarchical clustering and Gabor are filtered | |
| CN113936001B (en) | Textile surface flaw detection method based on image processing technology | |
| Maroš et al. | Monitoring chenille yarn defects using image processing with control charts | |
| CN114998321A (en) | Textile material surface hairiness degree identification method based on optical means | |
| CN112767384A (en) | Compensation method and device of elastic cloth gram weight online detection system | |
| CN115100481B (en) | Textile qualitative classification method based on artificial intelligence | |
| CN111709429A (en) | Method for identifying structural parameters of woven fabric based on convolutional neural network | |
| CN107945164A (en) | Textile flaw detection method based on peak threshold, rotational alignment and composite character | |
| CN109543733B (en) | Quick production method of yarn-dyed fabric based on cloud platform | |
| Mohanty et al. | Detection and classification of fabric defects in textile using image mining and association rule miner | |
| CN109919028B (en) | Flexible coordinate system establishing and shape identifying method based on fabric weave structure | |
| Nisha et al. | Survey on various defect detection and classification methods in fabric images | |
| Rauf et al. | Fabric weave pattern recognition and classification by machine learning | |
| Chen et al. | Adaptive gabor filtering for fabric defect inspection | |
| CN111968076A (en) | Real-time fabric defect detection method based on S-YOLOV3 | |
| Deng et al. | A new geometric modeling approach for woven fabric based on Frenet frame and Spiral Equation | |
| Niles et al. | A system for analysis, categorisation and grading of fabric defects using computer vision |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |