Knitted fabric fuzzing rating method and system based on image processing

Abstract

The invention relates to the technical field of knitted fabric fluffing identification, in particular to a knitted fabric fluffing rating method and system based on image processing. Randomly sampling the same batch of knitted fabrics to obtain an initial sample; identifying the initial sample production process to obtain a processing parameter set; constructing a fabric region division model to identify the processing parameter set to obtain a plurality of initial sub-region sets; constructing a fabric fuzzing recognition model to recognize an initial sub-region image set of the initial sub-region set to obtain first fuzzing data; performing fuzzing test on the initial sample to obtain a test sample; obtaining a test sub-region image set through a test sample; identifying the test sub-region image set through a fabric fluffing identification model to obtain second fluffing data; and constructing a fuzzing anomaly identification model to identify the processing parameter set, the first fuzzing data and the second fuzzing data to obtain a fuzzing anomaly coefficient. According to the method, the fluffing grade of the knitted fabric is accurately evaluated through the abnormal fluffing coefficient.

Description

Technical Field

[0001] The invention relates to the technical field of knitted fabric fuzzing identification, and in particular to a knitted fabric fuzzing rating method and system based on image processing. Background Art

[0002] Knitted fabrics are made by using knitting needles to bend yarn into loops and intertwine them. The raw materials mainly include natural fibers such as cotton, linen, silk, wool, and chemical fibers such as acrylic and polyester. Knitted fabrics have many advantages such as ventilation and breathability, and are widely used in clothing fabrics and home textile products.

[0003] Pilling refers to the formation of fine fiber fluff on the surface of fabrics, which is usually caused by the fibers falling off, moving or opening from the fabric. In the production and manufacturing process of knitted fabrics, due to the influence of yarn characteristics and knitting technology, knitted fabrics will have different degrees of pilling, which not only affects the appearance and comfort of knitted fabrics, but also reduces their durability.

[0004] Therefore, in the production process of knitted fabrics, it is necessary to test and identify the fuzzing characteristics of knitted fabrics in order to adjust the knitted fabric production technology.

[0005] With the development of technology, the fuzzing characteristics on the surface of knitted fabrics can be identified through image recognition technology, and then the fuzzing situation of knitted fabrics can be understood to a certain extent by analyzing the feature data; however, the existing technology only identifies the current fuzzing characteristics of knitted fabrics, and the yarn characteristics and knitting process that are closely related to the fuzzing characteristics are not studied, resulting in inaccurate identification of its fuzzing characteristics.

[0006] Therefore, a knitted fabric fuzzing rating method and system based on image processing is proposed. Summary of the invention

[0007] The object of the present invention is to provide a knitted fabric fuzzing rating method and system based on image processing; randomly sample the same batch of knitted fabrics to obtain initial samples; identify the production process of the initial samples to obtain a processing parameter set; construct a fabric area division model to identify the processing parameter set to obtain multiple initial sub-area sets; construct a fabric fuzzing recognition model to identify the initial sub-area image set of the initial sub-area set to obtain first fuzzing data; perform a fuzzing test on the initial sample to obtain a test sample; obtain a test sub-area image set through the test sample; identify the test sub-area image set through the fabric fuzzing recognition model to obtain second fuzzing data; construct a fuzzing abnormality recognition model to identify the processing parameter set, the first fuzzing data and the second fuzzing data to obtain a fuzzing abnormality coefficient; and accurately evaluate the fuzzing grade of the knitted fabric through the fuzzing abnormality coefficient.

[0008] To achieve the above object, the present invention provides the following technical solutions:

[0009] A knitted fabric fuzzing rating method based on image processing, comprising:

[0010] S10. Randomly sample knitted fabrics produced in the same batch to obtain an initial sample; collect data on the production process of the initial sample to obtain a processing parameter set; the processing parameter set includes yarn data and textile data;

[0011] S20. Constructing a fabric area division model to identify the processing parameter set, dividing the initial sample into regions according to the yarn data and the textile data to obtain initial sub-regions; identifying a processing parameter subset of the initial sub-region, obtaining a regional similarity according to the processing parameter subset, classifying the initial sub-regions, and obtaining a plurality of initial sub-region sets;

[0012] S30. photographing the initial sub-region set to obtain an initial sub-region image set; constructing a fabric fuzzing recognition model, identifying the initial sub-region image set through the fabric fuzzing recognition model, and obtaining first fuzzing data;

[0013] S40. Performing a fuzz test on the initial sample to obtain a test sample;

[0014] S50. According to the division and classification method of the initial sub-regions in the initial sample, the test sample is processed to obtain a test sub-region set; the test sub-region set is photographed to obtain a test sub-region image set; the test sub-region image set is identified by the fabric fuzz identification model to obtain second fuzz data;

[0015] S60. Construct a fuzzing abnormality recognition model to recognize the processing parameter set, the first fuzzing data and the second fuzzing data to obtain a fuzzing abnormality coefficient; and evaluate the fuzzing grade of the knitted fabric according to the fuzzing abnormality coefficient.

[0016] The yarn data includes yarn type, yarn twist, yarn length and yarn diameter; the textile data includes color distribution, structure distribution and density distribution.

[0017] The process of dividing the initial sub-areas by the fabric area division model is as follows:

[0018] Acquire a processing parameter subset of an initial sub-region through a processing parameter set of an initial sample, wherein the processing parameter subset includes yarn data and textile data of the initial sub-region;

[0019] By comparing the yarn data between the initial sub-regions, the yarn similarity is obtained; by comparing the textile data between the initial sub-regions, the textile similarity is obtained;

[0020] The regional similarity between the initial sub-regions is obtained through the yarn similarity and the textile similarity; the initial sub-regions are classified according to the regional similarity to obtain an initial sub-region set.

[0021] The fabric fuzzing identification model includes a fabric data processing layer, a same sample data identification layer, a different sample data identification layer and a fuzzing data identification layer;

[0022] The fabric data processing layer obtains a sub-region image set of the fabric, identifies the sub-region image set; organizes the sub-region images belonging to the same sample into a same sample image set; and organizes the sub-region images of the same position area in different samples into a different sample image set;

[0023] The sample data recognition layer recognizes the sample image set by texture features and color features, and excludes abnormal data by comparing and identifying between samples to obtain the sample fuzz data;

[0024] The abnormal sample data identification layer identifies the abnormal sample image set through texture features and color features, and excludes abnormal data through comparison and identification between the abnormal samples to obtain abnormal sample fuzzing data;

[0025] The fuzz data recognition layer recognizes the sub-region image set by combining the same sample fuzz data and the different sample fuzz data to obtain comprehensive fuzz data.

[0026] The process of the fuzz abnormality identification model identifying the fuzz abnormality coefficient is as follows:

[0027] Obtaining a processing parameter subset of an initial sub-region in an initial sub-region set through a processing parameter set;

[0028] Identify through the processing parameter subset and the first and second fuzzing data of the initial sub-region set to obtain a sub-region abnormality coefficient;

[0029] Obtain the sub-region anomaly coefficients of all initial sub-region sets of the initial sample;

[0030] The sub-region abnormality coefficient is weighted according to the area ratio of the initial sub-region to the initial sample in different initial sub-region sets to obtain the fuzzing abnormality coefficient.

[0031] The process of evaluating the fuzzing grade of knitted fabrics by using the fuzzing abnormality coefficient is as follows:

[0032] Setting a fuzzing abnormality threshold; obtaining a fuzzing abnormality coefficient of the knitted fabric; judging the fuzzing abnormality coefficient by the fuzzing abnormality threshold to obtain a fuzzing grade.

[0033] A knitted fabric fuzzing rating system based on image processing, comprising:

[0034] The parameter acquisition module randomly samples knitted fabrics produced in the same batch to obtain an initial sample; collects data on the production process of the initial sample to obtain a processing parameter set; the processing parameter set includes yarn data and textile data;

[0035] A fabric division module constructs a fabric area division model to identify the processing parameter set, divides the initial sample into regions according to the yarn data and the textile data, and obtains initial sub-regions; identifies a processing parameter subset of the initial sub-region, obtains regional similarity according to the processing parameter subset, classifies the initial sub-regions, and obtains a plurality of initial sub-region sets;

[0036] A first recognition module is used to photograph the initial sub-region set to obtain an initial sub-region image set; a fabric fuzzing recognition model is constructed, and the initial sub-region image set is recognized by the fabric fuzzing recognition model to obtain first fuzzing data;

[0037] A fuzzing test module, performing a fuzzing test on the initial sample to obtain a test sample;

[0038] The second recognition module processes the test sample according to the division and classification method of the initial sub-regions in the initial sample to obtain a test sub-region set; photographs the test sub-region set to obtain a test sub-region image set; and recognizes the test sub-region image set through the fabric fuzz recognition model to obtain second fuzz data;

[0039] The grade assessment module constructs a fuzzing abnormality recognition model to identify the processing parameter set, the first fuzzing data and the second fuzzing data, and obtains the fuzzing abnormality coefficient; and the fuzzing grade of the batch of knitted fabrics is assessed through the fuzzing abnormality coefficient.

[0040] The process of dividing the initial sub-areas by the fabric area division model is as follows:

[0041] Acquire a processing parameter subset of an initial sub-region through a processing parameter set of an initial sample, wherein the processing parameter subset includes yarn data and textile data of the initial sub-region;

[0042] By comparing the yarn data between the initial sub-regions, the yarn similarity is obtained; by comparing the textile data between the initial sub-regions, the textile similarity is obtained;

[0043] The regional similarity between the initial sub-regions is obtained through the yarn similarity and the textile similarity; the initial sub-regions are classified according to the regional similarity to obtain an initial sub-region set.

[0044] The fabric fuzzing identification model includes a fabric data processing layer, a same sample data identification layer, a different sample data identification layer and a fuzzing data identification layer;

[0045] The fabric data processing layer obtains a sub-region image set of the fabric, identifies the sub-region image set, and organizes the sub-region images of the same sample into a same sample image set; and organizes the sub-region images of the same position area in different samples into a different sample image set;

[0046] The sample data recognition layer recognizes the sample image set by texture features and color features, and excludes abnormal data by comparing and identifying between samples to obtain the sample fuzz data;

[0047] The abnormal sample data identification layer identifies the abnormal sample image set through texture features and color features, and excludes abnormal data through comparison and identification between the abnormal samples to obtain abnormal sample fuzzing data;

[0048] The fuzz data recognition layer recognizes the sub-region image set by combining the same sample fuzz data and the different sample fuzz data to obtain comprehensive fuzz data.

[0049] The process of the fuzz abnormality identification model identifying the fuzz abnormality coefficient is as follows:

[0050] Obtaining a processing parameter subset of an initial sub-region in an initial sub-region set through a processing parameter set;

[0051] Identify through the processing parameter subset and the first and second fuzzing data of the initial sub-region set to obtain a sub-region abnormality coefficient;

[0052] Obtain the sub-region anomaly coefficients of all initial sub-region sets of the initial sample;

[0053] The sub-region abnormality coefficient is weighted according to the area ratio of the initial sub-region to the initial sample in different initial sub-region sets to obtain the fuzzing abnormality coefficient.

[0054] Compared with the prior art, the present invention has the following beneficial effects:

[0055] 1. The present invention identifies and divides the region of the knitted fabric according to the yarn data and textile data of the knitted fabric to obtain multiple sub-regions; obtains the yarn similarity by comparing the yarn data between the sub-regions; obtains the textile similarity by comparing the textile data between the sub-regions; obtains the regional similarity between the sub-regions through the yarn similarity and the textile similarity; and accurately realizes the division and classification of the knitted fabric through the regional similarity.

[0056] 2. The present invention constructs a fabric fuzzing recognition model to recognize an initial sub-region image set; divides the initial sub-region image set to obtain a same sample image set and a different sample image set; obtains the same sample fuzzing data by recognizing the same sample image set; obtains the different sample fuzzing data by recognizing the different sample image set; obtains the comprehensive fuzzing data based on the same sample fuzzing data and the different sample fuzzing data; and accurately recognizes the fuzzing data of the initial sub-region image set.

[0057] 3. The present invention identifies the first and second fuzzing data of the processing parameter subset and the initial sub-region set to obtain the sub-region anomaly coefficient of the initial sub-region set; then weights the sub-region anomaly coefficient by the area ratio of the initial sub-region to the initial sample in different initial sub-region sets to obtain the fuzzing anomaly coefficient; and accurately identifies the fuzzing condition of the knitted fabric. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] Figure 1 It is a schematic flow chart of a method for grading fuzz of knitted fabrics based on image processing according to the present invention;

[0059] Figure 2 It is a structural schematic diagram of the fabric fuzzing identification model of the present invention;

[0060] Figure 3 It is a structural schematic diagram of a knitted fabric fuzzing rating system based on image processing of the present invention;

[0061] Figure 4 It is a schematic diagram of the area division of the knitted fabric sample of the present invention;

[0062] Figure 4 In FIG. 5 , a, b, c, d, and e represent the initial sub-regions of the knitted fabric sample. DETAILED DESCRIPTION

[0063] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0064] Pilling refers to the formation of fine fiber fluff on the surface of fabrics, which is usually caused by the fibers falling off, moving or opening from the fabric. In the production and manufacturing process of knitted fabrics, due to the influence of yarn characteristics and knitting technology, knitted fabrics will have different degrees of pilling, which not only affects the appearance and comfort of knitted fabrics, but also reduces their durability.

[0065] Therefore, in the production process of knitted fabrics, it is necessary to test and identify the fuzzing characteristics of knitted fabrics in order to adjust the knitted fabric production technology.

[0066] Therefore, a knitted fabric fuzzing rating method and system based on image processing is proposed.

[0067] Embodiment 1

[0068] The present invention proposes a knitted fabric fuzzing rating method based on image processing, the process of which is as follows: Figure 1 As shown, including:

[0069] S10. Randomly sample knitted fabrics produced in the same batch to obtain an initial sample; collect data on the production process of the initial sample to obtain a processing parameter set; the processing parameter set includes yarn data and textile data.

[0070] The yarn data includes yarn type, yarn twist, yarn length and yarn diameter; the textile data includes color distribution, structure distribution and density distribution.

[0071] Yarn is a linear object made from fibers through a spinning process and is the basic component of knitted fabrics. The type of yarn is related to the fibers used. According to the source of the fibers, it can be divided into natural yarn, synthetic yarn, and recycled yarn, etc.

[0072] Among them, natural yarn is made of natural fibers, such as cotton, wool, silk and linen; synthetic yarn is made of synthetic fibers, such as polyester, nylon, acrylic, etc.; recycled yarn is made of fibers obtained by chemical treatment of natural materials, such as viscose and modal.

[0073] Yarn twist refers to the degree of rotation of the fibers in the yarn, that is, the degree of twisting of the fibers along the yarn direction; yarn twist directly affects the structure, strength, softness and surface state of the yarn, and thus affects the fuzzing of knitted fabrics. Yarns with high twist are tighter, and the fibers are more tightly bound, which reduces the loose fibers and thus reduces the probability of fuzzing; the fibers in low twist yarns are loose and easily loosened during friction, resulting in an increase in fuzzing.

[0074] Yarn length refers to the length of a single fiber of the yarn, which affects the structural density and surface smoothness of the fabric. Generally speaking, long-fiber yarns are stronger, have a smoother surface, and a more stable structure. They can effectively reduce the loose fibers in the yarn, thereby reducing the probability of pilling.

[0075] Yarn diameter refers to the cross-sectional diameter of the yarn, which affects the strength, flexibility and final properties of the yarn. Coarse yarns usually have a tighter structure, with stronger connections between fibers, are less likely to fall off, and have relatively less pilling. Fine yarns have a small diameter, so the friction of the fibers is relatively small, making them more likely to fall off and pill.

[0076] Color distribution refers to the color distribution of knitted fabrics; structural distribution refers to the type, size, number and arrangement of structures on knitted fabrics, reflecting the structural composition of yarns; color distribution and structural distribution will affect the image recognition of the fuzzing characteristics of knitted fabrics.

[0077] Yarn density refers to the number of yarns per unit length or unit area; yarn density determines the tightness of the fabric, which in turn affects the fuzzing performance of the fabric; high-density fabrics have tightly arranged yarns, strong friction between yarns, fibers are not easy to loosen, the surface is flatter and smoother, and fuzzing is less likely to occur; low-density fabrics have loosely arranged yarns, fibers are easy to loosen, and fuzzing is easily formed during friction.

[0078] The present invention collects yarn parameters used in knitted fabrics from the perspectives of yarn type, yarn twist, yarn length and yarn diameter to obtain yarn data; collects textile parameters of knitted fabrics from the perspectives of color distribution, structure distribution and density distribution to obtain textile data; and the processing parameters of knitted fabrics can be accurately identified through yarn data and textile data.

[0079] S20. Construct a fabric area division model to identify the processing parameter set, divide the initial sample into regions according to the yarn data and the textile data, and obtain initial sub-regions; identify the processing parameter subset of the initial sub-region, obtain the regional similarity according to the processing parameter subset, classify the initial sub-regions, and obtain multiple initial sub-region sets.

[0080] The fabric area division model is obtained based on a deep neural network model, and the division process of the initial sub-areas is as follows:

[0081] Acquire a processing parameter subset of an initial sub-region through a processing parameter set of an initial sample, wherein the processing parameter subset includes yarn data and textile data of the initial sub-region;

[0082] By comparing the yarn data between the initial sub-regions, the yarn similarity is obtained; by comparing the textile data between the initial sub-regions, the textile similarity is obtained;

[0083] The regional similarity between the initial sub-regions is obtained through the yarn similarity and the textile similarity; the initial sub-regions are classified according to the regional similarity to obtain an initial sub-region set.

[0084] The present invention identifies and divides the area of ​​the knitted fabric according to the yarn data and textile data of the knitted fabric to obtain a plurality of sub-areas; the yarn data between the sub-areas are compared to obtain the yarn similarity; the textile similarity is obtained by comparing the textile data between the sub-areas; the regional similarity between the sub-areas is obtained by the yarn similarity and the textile similarity; the knitted fabric is accurately divided and classified by the regional similarity.

[0085] S30. Photographing the initial sub-region set to obtain an initial sub-region image set; constructing a fabric fuzzing recognition model, and identifying the initial sub-region image set by using the fabric fuzzing recognition model to obtain first fuzzing data.

[0086] The fabric fuzz recognition model is obtained based on a deep neural network model, and its structure is as follows: Figure 2 As shown, it includes a fabric data processing layer, a same sample data recognition layer, a different sample data recognition layer and a fuzzing data recognition layer;

[0087] The fabric data processing layer obtains a sub-region image set of the fabric, identifies the sub-region image set; organizes the sub-region images belonging to the same sample into a same sample image set; and organizes the sub-region images of the same position area in different samples into a different sample image set;

[0088] The sample data recognition layer recognizes the sample image set by texture features and color features, and excludes abnormal data by comparing and identifying between samples to obtain the sample fuzz data;

[0089] The abnormal sample data identification layer identifies the abnormal sample image set through texture features and color features, and excludes abnormal data through comparison and identification between the abnormal samples to obtain abnormal sample fuzzing data;

[0090] The fuzz data recognition layer recognizes the sub-region image set by combining the same sample fuzz data and the different sample fuzz data to obtain comprehensive fuzz data. The comprehensive fuzz data includes texture feature data and color feature data.

[0091] In the recognition of the initial sub-region image set, the obtained comprehensive hair raising data is used as the first hair raising data.

[0092] In order to verify the recognition accuracy of the fabric fuzz identification model for the fuzz data of the sub-region image set, the present invention conducts a control experiment on it. The control experiment includes four models in total; among them, model 1 is the fabric fuzz identification model, which performs comprehensive recognition through the sub-region image set and its same sample fuzz data and different sample fuzz data to obtain the first comprehensive fuzz data;

[0093] Model 2 is based on Model 1, but removes the recognition of the hair raising data of the outlier samples, and only performs comprehensive recognition through the sub-region image set and its hair raising data of the same sample to obtain the second comprehensive hair raising data;

[0094] Model 3 is based on Model 1, but removes the recognition of the same sample hair raising data, and only performs comprehensive recognition through the sub-region image set and its heterogeneous sample hair raising data to obtain the third comprehensive hair raising data;

[0095] Model 4 is based on Model 1, but removes the recognition of the same sample raising data and the different sample raising data, and only recognizes through the sub-region image set to obtain the fourth comprehensive raising data.

[0096] The data similarity between the comprehensive fuzzing data obtained by recognition and the standard fuzzing data is taken as the recognition accuracy. The higher the data similarity, the higher the recognition accuracy of the model. The recognition accuracy verification data of the fabric fuzzing recognition model is shown in Table 1.

[0097] Table 1 Verification data of fabric fuzzing recognition model recognition accuracy

[0098] Model Same identification Identification of outliers Recognition accuracy Model 1 Include Include 0.9572 Model 2 Include Not included 0.9135 Model 3 Not included Include 0.9248 Model 4 Not included Not included 0.8879

[0099] It can be seen from the data in Table 1 that the fabric fuzzing recognition model described in the present invention has the highest recognition accuracy.

[0100] The present invention constructs a fabric fuzzing recognition model to recognize an initial sub-region image set; divides the initial sub-region image set to obtain a same sample image set and a different sample image set; obtains same sample fuzzing data through the same sample image set recognition; obtains different sample fuzzing data through the different sample image set recognition; obtains comprehensive fuzzing data based on the same sample fuzzing data and the different sample fuzzing data recognition; and accurately recognizes the fuzzing data of the initial sub-region image set.

[0101] S40. Perform a fuzzing test on the initial sample to obtain a test sample.

[0102] The test method standards for pilling of clothing and textile products are: GB / T 4802.1-2008 "Circular Locus Method", GB / T4802.2-2008 "Modified Martindale Method", GB / T 4802.3-2008 "Pilling Box Method" and GB / T 4802.4-2020 "Random Tumbling Method".

[0103] S50. According to the division and classification method of the initial sub-regions in the initial sample, the test sample is processed to obtain a test sub-region set; the test sub-region set is photographed to obtain a test sub-region image set; the test sub-region image set is identified by the fabric fuzzing identification model to obtain second fuzzing data.

[0104] In the recognition of the test sub-region image set, the comprehensive fuzzing data obtained by the fabric fuzzing recognition model is used as the second fuzzing data.

[0105] S60. Construct a fuzzing abnormality recognition model to recognize the processing parameter set, the first fuzzing data and the second fuzzing data to obtain a fuzzing abnormality coefficient; and evaluate the fuzzing grade of the knitted fabric according to the fuzzing abnormality coefficient.

[0106] The fuzzing abnormality recognition model is obtained based on a deep neural network model, and the process of identifying the fuzzing abnormality coefficient is as follows:

[0107] Obtaining a processing parameter subset of an initial sub-region in an initial sub-region set through a processing parameter set;

[0108] Identify through the processing parameter subset and the first and second fuzzing data of the initial sub-region set to obtain a sub-region abnormality coefficient;

[0109] Obtain the sub-region anomaly coefficients of all initial sub-region sets of the initial sample;

[0110] The sub-region abnormality coefficient is weighted according to the area ratio of the initial sub-region to the initial sample in different initial sub-region sets to obtain the fuzzing abnormality coefficient.

[0111] The present invention identifies the first and second fuzzing data of the processing parameter subset and the initial sub-region set to obtain the sub-region anomaly coefficient of the initial sub-region set; then weights the sub-region anomaly coefficient by the area ratio of the initial sub-region to the initial sample in different initial sub-region sets to obtain the fuzzing anomaly coefficient; and accurately identifies the fuzzing condition of the knitted fabric.

[0112] The process of evaluating the fuzzing grade of knitted fabrics by using the fuzzing abnormality coefficient is as follows:

[0113] Setting a fuzzing abnormality threshold; obtaining a fuzzing abnormality coefficient of the knitted fabric; judging the fuzzing abnormality coefficient by the fuzzing abnormality threshold to obtain a fuzzing grade.

[0114] The present invention randomly samples the same batch of knitted fabrics to obtain initial samples; identifies the production process of the initial samples to obtain a processing parameter set; constructs a fabric area division model to identify the processing parameter set to obtain multiple initial sub-area sets; constructs a fabric fuzzing recognition model to identify the initial sub-area image set of the initial sub-area set to obtain first fuzzing data; performs a fuzzing test on the initial sample to obtain a test sample; obtains a test sub-area image set through the test sample; identifies the test sub-area image set through the fabric fuzzing recognition model to obtain second fuzzing data; constructs a fuzzing abnormality recognition model to identify the processing parameter set, the first fuzzing data and the second fuzzing data to obtain a fuzzing abnormality coefficient; and accurately evaluates the fuzzing grade of the knitted fabric through the fuzzing abnormality coefficient.

[0115] Embodiment 2

[0116] Textile Factory X mainly produces knitted fabrics. In order to understand the fuzzing characteristics of knitted fabrics and to upgrade industrial technology, Textile Factory X adopts the knitted fabric fuzzing rating system based on image processing described in the present invention.

[0117] The structure of the system is as follows Figure 3 As shown, it includes: a parameter acquisition module, a fabric classification module, a first identification module, a fuzzing test module, a second identification module and a grade assessment module;

[0118] The parameter acquisition module randomly samples knitted fabrics produced in the same batch to obtain an initial sample; collects data on the production process of the initial sample to obtain a processing parameter set; the processing parameter set includes yarn data and textile data.

[0119] Taking the production of textile factory X as an example, random sampling is performed on knitted fabrics produced in the same batch to obtain initial samples; the initial samples include initial sample A, initial sample B and initial sample C; and the processing parameter set of the initial samples is collected.

[0120] The yarn data includes yarn type, yarn twist, yarn length and yarn diameter; the textile data includes color distribution, structure distribution and density distribution.

[0121] Among them, the yarn type is related to the fiber used, and can be divided into natural yarn, synthetic yarn and regenerated yarn according to the fiber source; yarn twist refers to the degree of rotation of the fiber in the yarn, that is, the degree of twisting of the fiber along the yarn direction; yarn length refers to the length of a single fiber of the yarn; yarn diameter refers to the cross-sectional diameter of the yarn.

[0122] Color distribution refers to the color distribution of knitted fabrics; structural distribution refers to the type, size, number and arrangement of structures on knitted fabrics; yarn density refers to the number of yarns per unit length or unit area.

[0123] The fabric division module constructs a fabric area division model to identify the processing parameter set, divides the initial sample into regions according to the yarn data and the textile data, and obtains initial sub-regions; identifies the processing parameter subsets of the initial sub-regions, obtains the regional similarity according to the processing parameter subsets, classifies the initial sub-regions, and obtains multiple initial sub-region sets.

[0124] The process of dividing the initial sub-areas by the fabric area division model is as follows:

[0125] Acquire a processing parameter subset of an initial sub-region through a processing parameter set of an initial sample, wherein the processing parameter subset includes yarn data and textile data of the initial sub-region;

[0126] By comparing the yarn data between the initial sub-regions, the yarn similarity is obtained; by comparing the textile data between the initial sub-regions, the textile similarity is obtained;

[0127] The regional similarity between the initial sub-regions is obtained through the yarn similarity and the textile similarity; the initial sub-regions are classified according to the regional similarity to obtain an initial sub-region set.

[0128] The area division of the initial sample of knitted fabrics from textile mill X is as follows Figure 4 As shown, multiple initial sub-regions are identified through the processing parameter set, which are numbered as a, b, c, d and e.

[0129] Initial sample A includes initial sub-regions Aa, Ab, Ac, Ad and Ae; initial sample B includes initial sub-regions Ba, Bb, Bc, Bd and Be; initial sample C includes initial sub-regions Ca, Cb, Cc, Cd and Ce. The regional similarity is identified and classified through the processing parameter subsets of the initial sub-regions, where the regional similarity of initial sub-regions a and e meets the requirements, and the first initial sub-region set Aa, Ae, Ba, Be, Ca and Ce is obtained; the regional similarity of initial sub-regions b and d meets the requirements, and the second initial sub-region set Ab, Ad, Bb, Bd, Cb and Cd is obtained; the initial sub-region c does not identify the initial sub-region that meets the regional similarity requirements, and the third initial sub-region set Ac, Bc and Cc is obtained.

[0130] The present invention identifies and divides the area of ​​the knitted fabric according to the yarn data and textile data of the knitted fabric to obtain a plurality of sub-areas; the yarn data between the sub-areas are compared to obtain the yarn similarity; the textile similarity is obtained by comparing the textile data between the sub-areas; the regional similarity between the sub-areas is obtained by the yarn similarity and the textile similarity; the knitted fabric is accurately divided and classified by the regional similarity.

[0131] The first recognition module shoots the initial sub-region set to obtain the initial sub-region image set; constructs a fabric fuzzing recognition model, and recognizes the initial sub-region image set through the fabric fuzzing recognition model to obtain the first fuzzing data.

[0132] The fabric fuzzing identification model includes a fabric data processing layer, a same sample data identification layer, a different sample data identification layer and a fuzzing data identification layer;

[0133] The fabric data processing layer obtains a sub-region image set of the fabric, identifies the sub-region image set; organizes the sub-region images belonging to the same sample into a same sample image set; and organizes the sub-region images of the same position area in different samples into a different sample image set;

[0134] The sample data recognition layer recognizes the sample image set by texture features and color features, and excludes abnormal data by comparing and identifying between samples to obtain the sample fuzz data;

[0135] The abnormal sample data identification layer identifies the abnormal sample image set through texture features and color features, and excludes abnormal data through comparison and identification between the abnormal samples to obtain abnormal sample fuzzing data;

[0136] The fuzz data recognition layer recognizes the sub-region image set by combining the same sample fuzz data and the different sample fuzz data to obtain comprehensive fuzz data.

[0137] In the fabric data processing stage, taking the first initial sub-region set Aa, Ae, Ba, Be, Ca and Ce as an example, Aa and Ae belong to the initial sample A, so the images of Aa and Ae are taken as the first sample images; similarly, based on the images of Ba and Be, the second sample image is obtained; based on the images of Ca and Ce, the third sample image is obtained; the sample image set includes the first sample image, the second sample image and the third sample image.

[0138] Aa, Ba and Ca are the initial sub-regions of the same position area in different samples, so their images are taken as the first different sample images. Similarly, the images of Ae, Be and Ce are taken as the second different sample images. The different sample image set includes the first different sample image and the second different sample image. Among them, the third initial sub-region set Ac, Bc and Cc cannot obtain the same sample image, so only its different sample image is identified.

[0139] The present invention constructs a fabric fuzzing recognition model to recognize an initial sub-region image set; divides the initial sub-region image set to obtain a same sample image set and a different sample image set; obtains same sample fuzzing data through the same sample image set recognition; obtains different sample fuzzing data through the different sample image set recognition; obtains comprehensive fuzzing data based on the same sample fuzzing data and the different sample fuzzing data recognition; and accurately recognizes the fuzzing data of the initial sub-region image set.

[0140] The fuzzing test module performs a fuzzing test on the initial sample to obtain a test sample.

[0141] Textile Factory X uses GB / T4802.4-2020 "Random Tumbling Method" to conduct the pilling test; under the rotation of the stainless steel blade in the independent stainless steel sample test room, the initial sample is continuously and randomly rubbed against the cork lining to obtain the test sample.

[0142] The second recognition module processes the test sample according to the division and classification method of the initial sub-regions in the initial sample to obtain a test sub-region set; photographs the test sub-region set to obtain a test sub-region image set; and recognizes the test sub-region image set through the fabric fuzzing recognition model to obtain second fuzzing data.

[0143] The grade assessment module constructs a fuzzing abnormality recognition model to identify the processing parameter set, the first fuzzing data and the second fuzzing data, and obtains the fuzzing abnormality coefficient; and the fuzzing grade of the batch of knitted fabrics is assessed through the fuzzing abnormality coefficient.

[0144] The process of the fuzz abnormality identification model identifying the fuzz abnormality coefficient is as follows:

[0145] Obtaining a processing parameter subset of an initial sub-region in an initial sub-region set through a processing parameter set; identifying through the processing parameter subset and the first fuzzing data and the second fuzzing data of the initial sub-region set to obtain a sub-region abnormality coefficient;

[0146] Obtain the sub-region anomaly coefficients of all initial sub-region sets of the initial sample;

[0147] The sub-region abnormality coefficient is weighted according to the area ratio of the initial sub-region to the initial sample in different initial sub-region sets to obtain the fuzzing abnormality coefficient.

[0148] The length of the initial sample A is 10 meters, the width is 4 meters, and the total area is 40 square meters; the ratio of the area of ​​its initial sub-area to the total area is shown in Table 2.

[0149] Table 2 Initial sub-region area data table of initial sample A

[0150] Initial sub-area area Area ratio a 8.00㎡ 0.20 b 8.86㎡ 0.22 c 6.28㎡ 0.16 d 8.86㎡ 0.22 e 8.00㎡ 0.20

[0151] The present invention identifies the first and second fuzzing data of the processing parameter subset and the initial sub-region set to obtain the sub-region anomaly coefficient of the initial sub-region set; then weights the sub-region anomaly coefficient by the area ratio of the initial sub-region to the initial sample in different initial sub-region sets to obtain the fuzzing anomaly coefficient; and accurately identifies the fuzzing condition of the knitted fabric.

[0152] The process of evaluating the fuzzing grade of knitted fabrics by using the fuzzing abnormality coefficient is as follows: setting a fuzzing abnormality threshold; obtaining the fuzzing abnormality coefficient of the knitted fabric; judging the fuzzing abnormality coefficient by using the fuzzing abnormality threshold to obtain the fuzzing grade.

[0153] By means of the knitted fabric fuzzing rating system based on image processing described in the present invention, the textile factory X can accurately identify and evaluate the fuzzing characteristics of the knitted fabrics it produces.

[0154] The present invention randomly samples the same batch of knitted fabrics to obtain initial samples; identifies the production process of the initial samples to obtain a processing parameter set; constructs a fabric area division model to identify the processing parameter set to obtain multiple initial sub-area sets; constructs a fabric fuzzing recognition model to identify the initial sub-area image set of the initial sub-area set to obtain first fuzzing data; performs a fuzzing test on the initial sample to obtain a test sample; obtains a test sub-area image set through the test sample; identifies the test sub-area image set through the fabric fuzzing recognition model to obtain second fuzzing data; constructs a fuzzing abnormality recognition model to identify the processing parameter set, the first fuzzing data and the second fuzzing data to obtain a fuzzing abnormality coefficient; and accurately evaluates the fuzzing grade of the knitted fabric through the fuzzing abnormality coefficient.

[0155] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A knitted fabric fuzz rating method based on image processing, characterized in that: include: S10. Randomly sample knitted fabrics produced in the same batch to obtain an initial sample; Collecting data on the production process of the initial sample to obtain a processing parameter set; the processing parameter set includes yarn data and textile data; S20. Constructing a fabric area division model to identify the processing parameter set, dividing the initial sample into regions according to the yarn data and the textile data to obtain initial sub-regions; identifying a processing parameter subset of the initial sub-region, obtaining a regional similarity according to the processing parameter subset, classifying the initial sub-regions, and obtaining a plurality of initial sub-region sets; S30. photographing the initial sub-region set to obtain an initial sub-region image set; constructing a fabric fuzzing recognition model, identifying the initial sub-region image set through the fabric fuzzing recognition model, and obtaining first fuzzing data; S40. Performing a fuzz test on the initial sample to obtain a test sample; S50. According to the division and classification method of the initial sub-regions in the initial sample, the test sample is processed to obtain a test sub-region set; the test sub-region set is photographed to obtain a test sub-region image set; the test sub-region image set is identified by the fabric fuzz identification model to obtain second fuzz data; S60. Construct a fuzzing abnormality recognition model to recognize the processing parameter set, the first fuzzing data and the second fuzzing data to obtain a fuzzing abnormality coefficient; and evaluate the fuzzing grade of the knitted fabric according to the fuzzing abnormality coefficient.

2. The method for grading fuzz of knitted fabrics based on image processing according to claim 1, characterized in that: The yarn data includes yarn type, yarn twist, yarn length and yarn diameter; the textile data includes color distribution, structure distribution and density distribution.

3. The method for grading fuzz of knitted fabrics based on image processing according to claim 1, characterized in that: The process of dividing the initial sub-areas by the fabric area division model is as follows: Acquire a processing parameter subset of an initial sub-region through a processing parameter set of an initial sample, wherein the processing parameter subset includes yarn data and textile data of the initial sub-region; By comparing the yarn data between the initial sub-regions, the yarn similarity is obtained; by comparing the textile data between the initial sub-regions, the textile similarity is obtained; The regional similarity between the initial sub-regions is obtained through the yarn similarity and the textile similarity; the initial sub-regions are classified according to the regional similarity to obtain an initial sub-region set.

4. The method for grading fuzz of knitted fabrics based on image processing according to claim 1, characterized in that: The fabric fuzzing identification model includes a fabric data processing layer, a same sample data identification layer, a different sample data identification layer and a fuzzing data identification layer; The fabric data processing layer obtains a sub-region image set of the fabric, identifies the sub-region image set; organizes the sub-region images belonging to the same sample into a same sample image set; and organizes the sub-region images of the same position area in different samples into a different sample image set; The sample data recognition layer recognizes the sample image set by texture features and color features, and eliminates abnormal data by comparing and identifying between samples to obtain the sample fuzz data; The abnormal sample data identification layer identifies the abnormal sample image set through texture features and color features, and excludes abnormal data through comparison and identification between the abnormal samples to obtain abnormal sample fuzzing data; The fuzz data recognition layer recognizes the sub-region image set by combining the same sample fuzz data and the different sample fuzz data to obtain comprehensive fuzz data.

5. The method for grading fuzz of knitted fabrics based on image processing according to claim 1, characterized in that: The process of the fuzz abnormality identification model identifying the fuzz abnormality coefficient is as follows: Obtaining a processing parameter subset of an initial sub-region in an initial sub-region set through a processing parameter set; Identify through the processing parameter subset and the first and second fuzzing data of the initial sub-region set to obtain a sub-region abnormality coefficient; Obtain the sub-region anomaly coefficients of all initial sub-region sets of the initial sample; The sub-region abnormality coefficient is weighted according to the area ratio of the initial sub-region to the initial sample in different initial sub-region sets to obtain the fuzzing abnormality coefficient.

6. The method for grading fuzz of knitted fabrics based on image processing according to claim 1, characterized in that: The process of evaluating the fuzzing grade of knitted fabrics by using the fuzzing abnormality coefficient is as follows: Setting a fuzzing abnormality threshold; obtaining a fuzzing abnormality coefficient of the knitted fabric; judging the fuzzing abnormality coefficient by the fuzzing abnormality threshold to obtain a fuzzing grade.

7. A knitted fabric fuzz rating system based on image processing, characterized in that: include: The parameter acquisition module randomly samples the knitted fabrics produced in the same batch to obtain the initial sample; Collecting data on the production process of the initial sample to obtain a processing parameter set; the processing parameter set includes yarn data and textile data; A fabric division module constructs a fabric area division model to identify the processing parameter set, divides the initial sample into regions according to the yarn data and the textile data, and obtains initial sub-regions; identifies a processing parameter subset of the initial sub-region, obtains regional similarity according to the processing parameter subset, classifies the initial sub-regions, and obtains a plurality of initial sub-region sets; A first recognition module is used to photograph the initial sub-region set to obtain an initial sub-region image set; a fabric fuzzing recognition model is constructed, and the initial sub-region image set is recognized by the fabric fuzzing recognition model to obtain first fuzzing data; A fuzzing test module is used to perform a fuzzing test on the initial sample to obtain a test sample; The second recognition module processes the test sample according to the division and classification method of the initial sub-regions in the initial sample to obtain a test sub-region set; photographs the test sub-region set to obtain a test sub-region image set; and recognizes the test sub-region image set through the fabric fuzz recognition model to obtain second fuzz data; The grade assessment module constructs a fuzzing abnormality recognition model to identify the processing parameter set, the first fuzzing data and the second fuzzing data, and obtains the fuzzing abnormality coefficient; and the fuzzing grade of the batch of knitted fabrics is assessed through the fuzzing abnormality coefficient.

8. The knitted fabric fuzzing rating system based on image processing according to claim 7, characterized in that: The process of dividing the initial sub-areas by the fabric area division model is as follows: Acquire a processing parameter subset of an initial sub-region through a processing parameter set of an initial sample, wherein the processing parameter subset includes yarn data and textile data of the initial sub-region; By comparing the yarn data between the initial sub-regions, the yarn similarity is obtained; by comparing the textile data between the initial sub-regions, the textile similarity is obtained; The regional similarity between the initial sub-regions is obtained through the yarn similarity and the textile similarity; the initial sub-regions are classified according to the regional similarity to obtain an initial sub-region set.

9. The knitted fabric fuzzing rating system based on image processing according to claim 7, characterized in that: The fabric fuzzing identification model includes a fabric data processing layer, a same sample data identification layer, a different sample data identification layer and a fuzzing data identification layer; The fabric data processing layer obtains a sub-region image set of the fabric, identifies the sub-region image set, and organizes the sub-region images of the same sample into a same sample image set; and organizes the sub-region images of the same position area in different samples into a different sample image set; The sample data recognition layer recognizes the sample image set by texture features and color features, and eliminates abnormal data by comparing and identifying between samples to obtain the sample fuzz data; The abnormal sample data identification layer identifies the abnormal sample image set through texture features and color features, and excludes abnormal data through comparison and identification between the abnormal samples to obtain abnormal sample fuzzing data; The fuzz data recognition layer recognizes the sub-region image set by combining the same sample fuzz data and the different sample fuzz data to obtain comprehensive fuzz data.

10. The knitted fabric fuzz rating system based on image processing according to claim 7, characterized in that: The process of the fuzz abnormality identification model identifying the fuzz abnormality coefficient is as follows: Obtaining a processing parameter subset of an initial sub-region in an initial sub-region set through a processing parameter set; Identify through the processing parameter subset and the first and second fuzzing data of the initial sub-region set to obtain a sub-region abnormality coefficient; Obtain the sub-region anomaly coefficients of all initial sub-region sets of the initial sample; The sub-region abnormality coefficient is weighted according to the area ratio of the initial sub-region to the initial sample in different initial sub-region sets to obtain the fuzzing abnormality coefficient.

Images