Sock designing and weaving integrated system based on three-dimensional digitization

Abstract

The invention relates to the technical field of sock production monitoring, in particular to a sock designing and weaving integrated system based on three-dimensional digitization. Acquiring three-dimensional design data of the socks; randomly sampling the produced socks to obtain sock weaving samples; performing three-dimensional scanning on the sock weaving sample by adopting high-resolution scanning equipment to obtain three-dimensional sample data; obtaining standard fuzzing data and denoising the three-dimensional sample data to obtain preprocessed three-dimensional sample data; constructing a sock defect identification model to identify the preprocessed three-dimensional sample data to obtain a defect area and a defect coefficient; marking the defect area and the defect coefficient on the three-dimensional design data to obtain defect distribution three-dimensional data; and clustering defect areas and defect coefficients on the defect distribution three-dimensional data through a clustering algorithm to obtain abnormal areas and abnormal coefficients. According to the method, early warning is accurately carried out on the three-dimensional digital sock design and weaving process through the obtained abnormal areas and abnormal coefficients.

Description

Technical Field

[0001] The invention relates to the technical field of sock production monitoring, and in particular to a sock design and weaving integrated system based on three-dimensional digitization. Background Art

[0002] Three-dimensional digital sock design and weaving refers to the use of data technology to create sock models, and then use the sock models to guide the weaving process to achieve automation of sock design and weaving. Compared with traditional sock design and production technology, the three-dimensional digital design of socks can be improved in real time according to the personalized needs of consumers and market orientation, shortening the development cycle of socks and improving the efficiency of iterative upgrades; guiding the weaving process through digital sock models can omit many steps from sock design to sock production, optimize the production process, improve product production efficiency, and be more competitive.

[0003] In the three-dimensional digital design and weaving of socks, the design and production of socks products are effectively optimized to improve production efficiency; however, when defects appear in the three-dimensional digital design plan or weaving process of socks products, if the weaving production process is not strictly reviewed, the defects will be caused by the rapid automated production process, resulting in huge economic losses.

[0004] Therefore, during the three-dimensional digital design and weaving process of socks, it is necessary to conduct defect detection on the design scheme and weaving process to ensure the quality of the socks products.

[0005] In the quality inspection process of socks, the appearance, material, structure and other properties of socks are inspected through equipment and instruments; with the development of image recognition technology, the efficiency of inspection can be improved by obtaining images of socks for identification; and the three-dimensional digital design and weaving of socks involve more microscopic data, while the existing inspection technology mainly conducts quality inspection and identification from a macro level, resulting in low recognition accuracy.

[0006] Therefore, a sock design and weaving integrated system based on three-dimensional digitization was proposed. Summary of the invention

[0007] The purpose of the present invention is to provide a sock design and weaving integrated system based on three-dimensional digitization; to obtain three-dimensional design data of socks; to randomly sample the produced socks to obtain sock weaving samples; to use high-resolution scanning equipment to three-dimensionally scan the sock weaving samples to obtain three-dimensional sample data; to obtain standard fuzzing data to denoise the three-dimensional sample data to obtain pre-processed three-dimensional sample data; to construct a sock defect recognition model to recognize the pre-processed three-dimensional sample data to obtain defect areas and defect coefficients; to mark the defect areas and defect coefficients on the three-dimensional design data to obtain defect distribution three-dimensional data; to cluster the defect areas and defect coefficients on the defect distribution three-dimensional data by the DBSCAN clustering algorithm to obtain abnormal areas and abnormal coefficients; and to accurately warn the three-dimensional digitized sock design and weaving process by the abnormal areas and abnormal coefficients.

[0008] To achieve the above object, the present invention provides the following technical solutions: A three-dimensional digitalized hosiery design and weaving integrated system, comprising: A sock sample acquisition module is used to acquire three-dimensional design data of socks; and to randomly sample socks produced according to the three-dimensional design data to obtain sock weaving samples; The sample data acquisition module uses a high-resolution scanning device to perform three-dimensional scanning on the sock knitting sample to obtain three-dimensional sample data; The sample data processing module matches the historical sock design data according to the three-dimensional design data of the socks to obtain the raised control socks; scans the raised control socks to obtain the standard raised data; and denoises the three-dimensional sample data according to the standard raised data to obtain the pre-processed three-dimensional sample data; A sample defect recognition module is used to construct a hosiery defect recognition model, wherein the hosiery defect recognition model recognizes the pre-processed three-dimensional sample data according to the three-dimensional design data, and obtains a defect area and a defect coefficient of the pre-processed three-dimensional sample data according to the difference of the three-dimensional data of the corresponding area; A defect data processing module obtains defect areas and defect coefficients of all sock knitting samples and marks them on the three-dimensional design data to obtain three-dimensional defect distribution data; The defect data identification module clusters the defect areas and defect coefficients on the three-dimensional data of defect distribution through the DBSCAN clustering algorithm to obtain the abnormal area; the abnormal coefficient is obtained based on the defect area and defect coefficient in the abnormal area; The abnormal warning module can issue warnings on the three-dimensional digital sock design and weaving process based on abnormal areas and abnormal coefficients.

[0009] The process of performing three-dimensional scanning on the sock knitting sample to obtain three-dimensional sample data is as follows: Acquire three-dimensional design data of socks, obtain cavity data of socks according to the three-dimensional design data, and obtain a cavity mold of socks according to the cavity data of socks; Putting a sock knitting sample onto the sock cavity mold; The sock knitting sample worn on the sock cavity mold is three-dimensionally scanned by a high-resolution scanning device to obtain three-dimensional sample data.

[0010] The acquisition process of the raised control socks is as follows: Matching historical sock design data according to the three-dimensional design data of the socks, and determining standard socks according to the similarity of the design data; the historical sock data includes the three-dimensional design data of the historical socks; The standard socks are sampled to obtain standard socks samples, quality inspection is performed on the standard socks samples, and the standard socks samples with qualified quality are used as fuzzing control socks.

[0011] The denoising process of the three-dimensional sample data according to the standard fuzz data is as follows: According to the method of performing three-dimensional scanning on the sock knitting sample to obtain three-dimensional sample data, performing three-dimensional scanning on the raising characteristics of the raising control socks to obtain standard raising data; The three-dimensional sample data is identified according to the standard raising data, and the three-dimensional data whose similarity with the standard raising data meets a threshold value in the three-dimensional sample data is deleted to obtain pre-processed three-dimensional sample data.

[0012] The hosiery defect recognition model includes a design data recognition unit, a sample data recognition unit, a data difference recognition unit, and a defect area recognition unit; The design data identification unit is used to identify the three-dimensional design data and divide the three-dimensional design data into a plurality of design sub-areas; The sample data identification unit is used to identify the pre-processed three-dimensional sample data, and divide the pre-processed three-dimensional sample data into a plurality of sample sub-regions according to the way in which the three-dimensional design data is divided into design sub-regions; wherein the design sub-regions and the sample sub-regions are in a one-to-one correspondence; The data difference identification unit is used to identify the difference between the design sub-region and the sample sub-region in a corresponding relationship, and obtain the difference coefficient according to the data difference between the design sub-region and the sample sub-region; The defect area identification unit is used to judge the difference coefficient according to the difference threshold, take the difference coefficient greater than the difference threshold as the defect coefficient, and take the corresponding sample sub-area as the defect area.

[0013] The method for obtaining the difference threshold is: Obtaining qualified socks produced according to the three-dimensional design data of the socks; scanning the qualified socks by a high-resolution scanning device to obtain three-dimensional data of the qualified socks; The data similarity between the qualified socks three-dimensional data and the three-dimensional design data is taken as the qualified similarity; and the difference threshold is determined according to the qualified similarity.

[0014] The process of acquiring the three-dimensional defect distribution data is as follows: The defective areas and defective coefficients of the sock weaving samples are obtained in turn, and the corresponding design sub-areas on the three-dimensional design data of the socks are obtained according to the defective area identification; the defective areas and defective coefficients of the sock weaving samples are marked according to the position distribution of the design sub-areas in the three-dimensional design data; after the defective areas and defective coefficients of all the sock weaving samples are marked, the three-dimensional data of defect distribution is obtained.

[0015] The identification process of the DBSCAN clustering algorithm is: Acquire three-dimensional data of defect distribution, establish a three-dimensional coordinate system based on the three-dimensional data of defect distribution, and obtain three-dimensional coordinate data of defect areas; obtain distance data between defect areas based on the three-dimensional coordinate data of defect areas; then obtain defect weights based on defect coefficients of defect areas; perform clustering by combining the distance data and the defect weights through the DBSCAN clustering algorithm to obtain cluster areas, and use the cluster areas as abnormal areas.

[0016] The defect weight is obtained by: obtaining the defect coefficients of all sock knitting samples to obtain the defect average value; and taking the ratio of the defect coefficient to the defect average value as the defect weight; The calculation formula of the defect weight is: ; in, represents the defect weight; represents the defect coefficient; Indicates the average defect value.

[0017] The calculation formula of the abnormal coefficient is: ; in, represents the anomaly coefficient; Indicates the abnormal area The defect coefficient corresponding to each defect area; Indicates the number of defective areas contained in the abnormal area.

[0018] Compared with the prior art, the present invention has the following beneficial effects: 1. The present invention matches historical sock design data according to the three-dimensional design data of socks, determines standard socks according to the similarity of the data; uses standard sock samples with qualified quality as fuzzing control socks to identify and obtain standard fuzzing data; then identifies the three-dimensional sample data through the standard fuzzing data, deletes normal fuzzing data in the three-dimensional sample data, so as to eliminate the interference of normal fuzzing data on data identification, and effectively improves the accuracy of three-dimensional sample data identification.

[0019] 2. The present invention constructs a sock defect recognition model based on a deep neural network, and uses the sock defect recognition model to segment the three-dimensional design data and the pre-processed three-dimensional sample data to obtain one-to-one corresponding design sub-regions and sample sub-regions; then, the corresponding design sub-regions and sample sub-regions are difference-identified to obtain difference coefficients; the defect coefficients and defect regions are obtained according to the difference coefficients, and the defect data of the pre-processed three-dimensional sample data are accurately identified.

[0020] 3. The present invention is based on three-dimensional design data, and marks the identified defect areas and defect coefficients to obtain three-dimensional data of defect distribution; the defect weight is calculated according to the defect coefficient in the three-dimensional data of defect distribution, and the distance data between the defect areas is obtained according to the three-dimensional coordinate data; clustering is performed according to the defect weight and the distance data between the defect areas to identify the abnormal areas, which can accurately identify the areas where defects occur in the sock knitting samples. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 It is a structural schematic diagram of a hosiery design and weaving integrated system based on three-dimensional digitization of the present invention; Figure 2 A schematic diagram of a denoising process for three-dimensional design data of socks according to the present invention; Figure 3 It is a structural schematic diagram of the hosiery defect recognition model of the present invention; Figure 4 This is a schematic diagram of the three-dimensional design data of the product C sample of the present invention. DETAILED DESCRIPTION

[0022] 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.

[0023] Embodiment 1 The present invention proposes a hosiery design and weaving integrated system based on three-dimensional digitization, the structure of the system is as follows: Figure 1 As shown, it includes: a sock sample acquisition module, a sample data acquisition module, a sample data processing module, a sample defect recognition module, a defect data processing module, a defect data recognition module and an abnormal warning module.

[0024] The sock sample acquisition module acquires the three-dimensional design data of the socks, and randomly samples the socks produced according to the three-dimensional design data to obtain sock weaving samples.

[0025] The three-dimensional design data includes yarn geometric structure data of the hosiery.

[0026] The sample data acquisition module uses a high-resolution scanning device to perform a three-dimensional scan on the sock knitting sample to obtain three-dimensional sample data.

[0027] The process of performing three-dimensional scanning on the sock knitting sample to obtain three-dimensional sample data is as follows: Acquire three-dimensional design data of socks, obtain cavity data of socks according to the three-dimensional design data, and obtain a cavity mold of socks according to the cavity data of socks; Putting a sock knitting sample onto the sock cavity mold; The sock knitting sample worn on the sock cavity mold is three-dimensionally scanned by a high-resolution scanning device to obtain three-dimensional sample data.

[0028] The present invention obtains sock cavity data according to the three-dimensional design data of the socks, and then obtains the sock cavity mold according to the sock cavity data; the shape of the sock weaving sample is fixed according to the sock cavity mold, which can effectively ensure that the three-dimensional sample data obtained by scanning is in a corresponding relationship with the three-dimensional design data, so as to improve the accuracy of data acquisition.

[0029] The sample data processing module matches the historical sock design data according to the three-dimensional design data of the socks to obtain the raised control socks; scans the raised control socks to obtain the standard raised data; and denoises the three-dimensional sample data according to the standard raised data to obtain the pre-processed three-dimensional sample data.

[0030] The process of denoising the three-dimensional design data of the present invention is as follows: Figure 2 As shown, including: Obtain three-dimensional design data of socks; Matching historical sock design data according to the three-dimensional design data of the socks, and determining standard socks according to the similarity of the data, wherein the historical sock data includes the three-dimensional design data of the historical socks; The standard socks are sampled to obtain standard socks samples, quality inspection is performed on the standard socks samples, and the standard socks samples with qualified quality are used as fuzzing control socks.

[0031] The present invention matches historical sock design data according to the three-dimensional design data of the socks, and determines the standard socks through the data similarity between the three-dimensional design data of the socks and the historical sock design data; the historical socks that are sufficiently similar to the target socks can be accurately obtained, and the comparability of the standard socks and the target socks can be effectively ensured.

[0032] According to the method of performing three-dimensional scanning on the sock knitting sample to obtain three-dimensional sample data, performing three-dimensional scanning on the raising characteristics of the raising control socks to obtain standard raising data; The three-dimensional sample data is identified according to the standard raising data, and the three-dimensional data whose similarity with the standard raising data meets a threshold value in the three-dimensional sample data is deleted to obtain pre-processed three-dimensional sample data.

[0033] The present invention performs three-dimensional scanning on the fuzzing characteristics of the fuzzing control socks to obtain standard fuzzing data; then the three-dimensional sample data is identified by the standard fuzzing data, and normal fuzzing data in the three-dimensional sample data is deleted to eliminate the interference of the normal fuzzing data on the identification, thereby effectively improving the accuracy of the identification of the three-dimensional sample data.

[0034] The sample defect recognition module constructs a hosiery defect recognition model, which recognizes the pre-processed three-dimensional sample data according to the three-dimensional design data, and obtains the defect area and defect coefficient of the pre-processed three-dimensional sample data according to the difference of the three-dimensional data of the corresponding area.

[0035] The hosiery defect recognition model is based on a deep neural network, and its structure is as follows: Figure 3 As shown, it includes a design data identification unit, a sample data identification unit, a data difference identification unit, and a defect area identification unit; The design data identification unit is used to identify the three-dimensional design data and divide the three-dimensional design data into a plurality of design sub-areas; The sample data identification unit is used to identify the pre-processed three-dimensional sample data, and divide the pre-processed three-dimensional sample data into a plurality of sample sub-regions according to the way in which the three-dimensional design data is divided into design sub-regions; wherein the design sub-regions and the sample sub-regions are in a one-to-one correspondence; The data difference identification unit is used to identify the difference between the design sub-region and the sample sub-region in a corresponding relationship, and obtain the difference coefficient according to the data difference between the design sub-region and the sample sub-region; The defect area identification unit is used to judge the difference coefficient according to the difference threshold, take the difference coefficient greater than the difference threshold as the defect coefficient, and take the corresponding sample sub-area as the defect area.

[0036] The present invention constructs a sock defect recognition model based on a deep neural network, and uses the sock defect recognition model to segment three-dimensional design data and pre-processed three-dimensional sample data to obtain one-to-one corresponding design sub-areas and sample sub-areas; then, the corresponding design sub-areas and sample sub-areas are difference-identified to obtain difference coefficients; defect coefficients and defect areas are obtained according to the difference coefficients, and the defect data of the pre-processed three-dimensional sample data are accurately identified.

[0037] The method for obtaining the difference threshold is as follows: Obtaining qualified socks produced according to the three-dimensional design data of the socks; scanning the qualified socks by a high-resolution scanning device to obtain three-dimensional data of the qualified socks; The data similarity between the qualified socks three-dimensional data and the three-dimensional design data is taken as the qualified similarity; and the difference threshold is determined according to the qualified similarity.

[0038] The present invention obtains qualified socks produced according to three-dimensional design data of socks, performs three-dimensional scanning on the qualified socks to obtain three-dimensional data of the qualified socks; identifies and obtains qualified similarity between the three-dimensional data of the qualified socks and the three-dimensional design data of the socks, and uses the obtained qualified similarity as a difference threshold, so as to accurately determine the identified difference coefficient.

[0039] The defect data processing module obtains the defect areas and defect coefficients of all sock knitting samples, and marks them on the three-dimensional design data to obtain three-dimensional data of defect distribution.

[0040] The process of acquiring the three-dimensional defect distribution data is as follows: The defect areas and defect coefficients of the sock knitting samples are obtained in sequence, and the corresponding design sub-areas on the three-dimensional design data of the socks are obtained according to the defect area identification; According to the position distribution of the design sub-areas in the three-dimensional design data, the defect areas and defect coefficients of the sock knitting samples are marked; After marking the defect areas and defect coefficients of all sock knitting samples, three-dimensional defect distribution data are obtained.

[0041] The present invention is based on three-dimensional design data, takes the included design sub-areas as positioning standards, marks the defect areas and defect coefficients obtained by pre-processing three-dimensional sample data, and obtains three-dimensional defect distribution data; accurately obtains the defect data distribution of all sock weaving samples.

[0042] The defect data identification module clusters the defect areas and defect coefficients on the three-dimensional data of defect distribution through the DBSCAN clustering algorithm to obtain the abnormal area; and obtains the abnormal coefficient based on the defect area and defect coefficient in the abnormal area.

[0043] The identification process of the DBSCAN clustering algorithm is: Acquire three-dimensional defect distribution data, establish a three-dimensional coordinate system according to the three-dimensional defect distribution data, and obtain three-dimensional coordinate data of the defect area; The distance data between the defective areas is calculated based on the three-dimensional coordinate data of the defective areas; and the defect weight is obtained based on the defect coefficient of the defective area; Clustering is performed by combining the distance data and the defect weights using a DBSCAN clustering algorithm to obtain a clustering region, which is used as an abnormal region.

[0044] The defect weight is obtained by: obtaining the defect coefficients of all sock knitting samples to obtain the defect average value; and taking the ratio of the defect coefficient to the defect average value as the defect weight; The calculation formula of the defect weight is: ; in, represents the defect weight; represents the defect coefficient; Indicates the average defect value; Then, clustering is performed according to the distance data and the defect weights using a DBSCAN clustering algorithm to obtain a clustering region, and the clustering region is used as an abnormal region.

[0045] The present invention obtains defect weights according to defect coefficients in three-dimensional data of defect distribution, obtains distance data between defective areas according to three-dimensional coordinate data identification, and then performs clustering according to the defect weights and the distance data between defective areas to identify abnormal areas, thereby accurately identifying areas where defects occur in a sock knitting sample concentration.

[0046] The formula for calculating the anomaly coefficient based on the abnormal area is: ; in, represents the anomaly coefficient; Indicates the abnormal area The defect coefficient corresponding to each defect area; Indicates the number of defective areas contained in the abnormal area.

[0047] The present invention obtains defect areas and defect coefficients in the abnormal area, and takes the average value of the defect coefficients of all defect areas as the abnormal coefficient, so as to accurately measure the abnormal degree of the abnormal area.

[0048] The abnormal warning module can issue warnings on the three-dimensional digital sock design and weaving process based on abnormal areas and abnormal coefficients.

[0049] The present invention obtains three-dimensional design data of socks, randomly samples the produced socks, and obtains sock weaving samples; uses a high-resolution scanning device to perform three-dimensional scanning on the sock weaving samples to obtain three-dimensional sample data; obtains standard fuzzing data to perform denoising on the three-dimensional sample data to obtain pre-processed three-dimensional sample data; constructs a sock defect recognition model to recognize the pre-processed three-dimensional sample data to obtain defect areas and defect coefficients; marks the defect areas and defect coefficients on the three-dimensional design data to obtain defect distribution three-dimensional data; clusters the defect areas and defect coefficients on the defect distribution three-dimensional data through a DBSCAN clustering algorithm to obtain abnormal areas and abnormal coefficients; and accurately warns the three-dimensional digital sock design and weaving process through the abnormal areas and abnormal coefficients.

[0050] Embodiment 2 Sock weaving factory P is engaged in the design and production of sock products. In order to optimize production efficiency and improve industry competitiveness, sock weaving factory P introduced digital technology to achieve three-dimensional digitalization of sock design and weaving, which greatly improved the efficiency of sock design and weaving. However, with the improvement of sock design and weaving efficiency, the previous sock quality inspection methods and processes have become insufficient, which in turn increases the losses caused by sock defects; for this reason, sock weaving factory P adopts a sock design and weaving integrated system based on three-dimensional digitalization described in the present invention.

[0051] The system comprises: a sock sample acquisition module, a sample data acquisition module, a sample data processing module, a sample defect recognition module, a defect data processing module, a defect data recognition module and an abnormality warning module.

[0052] The sock sample acquisition module acquires the three-dimensional design data of socks of a sock knitting factory P, and randomly samples the socks produced according to the three-dimensional design data to obtain sock knitting samples.

[0053] The three-dimensional design data includes yarn geometric structure data of the socks, through which the spatial geometric structure and yarn type data of the socks can be identified.

[0054] The sample data acquisition module uses a high-resolution scanning device to perform a three-dimensional scanning of the sock weaving samples of the sock weaving factory P to obtain three-dimensional sample data.

[0055] The process of performing three-dimensional scanning on the sock knitting sample to obtain three-dimensional sample data is as follows: Acquire three-dimensional design data of socks, obtain sock cavity data according to the three-dimensional design data, and obtain a sock cavity mold according to the sock cavity data; wear a sock weaving sample on the sock cavity mold; perform three-dimensional scanning on the sock weaving sample worn on the sock cavity mold by a high-resolution scanning device to obtain three-dimensional sample data.

[0056] According to the sock cavity data and the sock cavity mold, it can be ensured that the scanned 3D sample data is in a corresponding relationship with the 3D design data, which is helpful for the subsequent defect identification of the sock weaving sample of the sock weaving factory P.

[0057] The sample data processing module matches the historical sock design data according to the three-dimensional design data of the socks to obtain the raised control socks; scans the raised control socks to obtain the standard raised data; and denoises the three-dimensional sample data according to the standard raised data to obtain the pre-processed three-dimensional sample data.

[0058] In the design and weaving of socks, the pilling of socks is inevitable, and reasonable pilling is an important component of measuring the quality of socks. In order to avoid the interference of normal pilling on the three-dimensional sample data, the pilling data needs to be identified and processed.

[0059] The acquisition process of the raised control socks is as follows: The historical sock design data of the sock knitting factory P are matched according to the three-dimensional design data of the socks, and the standard socks are determined according to the similarity of the data, wherein the historical sock data includes the historical three-dimensional design data of the socks; thereby, data in the historical sock design data of the sock knitting factory P that is similar to the three-dimensional design data of the target socks is identified.

[0060] The standard socks are sampled to obtain standard socks samples, quality inspection is performed on the standard socks samples, and the standard socks samples with qualified quality are used as fuzzing control socks.

[0061] Standard socks are determined by the data similarity between the three-dimensional design data of the socks and the historical socks design data; historical socks that are sufficiently similar to the target socks can be accurately obtained, effectively ensuring the comparability of the standard socks and the target socks.

[0062] The denoising process of the three-dimensional sample data is as follows: According to the method of performing three-dimensional scanning on the sock knitting sample to obtain three-dimensional sample data, performing three-dimensional scanning on the raising characteristics of the raising control socks to obtain standard raising data; The three-dimensional sample data is identified according to the standard raising data, and the three-dimensional data whose similarity with the standard raising data meets a threshold value in the three-dimensional sample data is deleted to obtain pre-processed three-dimensional sample data.

[0063] The standard raising data is obtained by three-dimensionally scanning the raising characteristics of the raising control socks; and the three-dimensional sample data is then identified and denoised using the standard raising data, thereby effectively improving the accuracy of the three-dimensional sample data identification.

[0064] In order to verify the effect of denoising the three-dimensional sample data, different types of socks produced by the sock weaving factory P are selected for verification experiments; the selected sock products are shown in Table 1: Table 1 Product category data of socks knitting factory P

[0065] Among them, wool yarn and cotton yarn are natural fibers with more short fibers and scales on the surface, which are easy to be pulled out and often pilling during the friction process; polyester yarn and nylon yarn are synthetic fibers with relatively stable structures, and are not easily rubbed to produce loose fibers, and the pilling phenomenon is relatively rare.

[0066] Obtain the three-dimensional design data of socks product A, product B, product C and product D, and perform weaving production according to the three-dimensional design data, sample and obtain weaving samples, and perform quality inspection on the weaving samples to obtain qualified weaving samples, and obtain corresponding product A samples, product B samples, product C samples and product D samples; Among them, the 3D design data of product C sample is as follows Figure 4 As shown; Perform three-dimensional scanning on product A sample, product B sample, product C sample and product D sample to obtain corresponding initial three-dimensional sample data; adopt the denoising method of the present invention to denoise the initial three-dimensional sample data of product A sample, product B sample, product C sample and product D sample respectively to obtain pre-processed three-dimensional sample data of product A sample, product B sample, product C sample and product D sample; The similarity between the initial three-dimensional sample data and the corresponding three-dimensional design data of the socks product is obtained as the first similarity; the similarity between the preprocessed three-dimensional sample data and the corresponding three-dimensional design data of the socks product is obtained as the second similarity; the effect of data denoising is verified according to the first similarity and the second similarity; the recognition results are shown in Table 2.

[0067] Table 2 De-noising verification data table of three-dimensional sample data of socks products

[0068] According to the data in Table 2, the second similarities of product A sample, product B sample, product C sample and product D sample are respectively greater than the corresponding first data similarities, which means that the denoised three-dimensional sample data is closer to the three-dimensional design data than the non-denoised three-dimensional sample data, and the data error is smaller. In the defect recognition of the three-dimensional sample data of socks, the interference of normal fuzzing data on the recognition can be eliminated, and the accuracy of the recognition of the three-dimensional sample data can be effectively improved.

[0069] The sample defect recognition module constructs a hosiery defect recognition model, which recognizes the pre-processed three-dimensional sample data according to the three-dimensional design data, and obtains the defect area and defect coefficient of the pre-processed three-dimensional sample data according to the difference of the three-dimensional data of the corresponding area.

[0070] The sock defect recognition model is obtained based on a deep neural network, and includes a design data recognition unit, a sample data recognition unit, a data difference recognition unit, and a defect area recognition unit; The design data identification unit is used to identify the three-dimensional design data and divide the three-dimensional design data into a plurality of design sub-areas; The sample data identification unit is used to identify the pre-processed three-dimensional sample data, and divide the pre-processed three-dimensional sample data into a plurality of sample sub-regions according to the way in which the three-dimensional design data is divided into design sub-regions; wherein the design sub-regions and the sample sub-regions are in a one-to-one correspondence; The data difference identification unit is used to identify the difference between the design sub-region and the sample sub-region in a corresponding relationship, and obtain the difference coefficient according to the data difference between the design sub-region and the sample sub-region; The defect area identification unit is used to judge the difference coefficient according to the difference threshold, take the difference coefficient greater than the difference threshold as the defect coefficient, and take the corresponding sample sub-area as the defect area.

[0071] The present invention constructs a sock defect recognition model based on a deep neural network, and uses the sock defect recognition model to segment three-dimensional design data and pre-processed three-dimensional sample data to obtain one-to-one corresponding design sub-areas and sample sub-areas; then, the corresponding design sub-areas and sample sub-areas are difference-identified to obtain difference coefficients; defect coefficients and defect areas are obtained according to the difference coefficients, and the defect data of the pre-processed three-dimensional sample data are accurately identified.

[0072] In the process of identifying defects, the present invention converts the product defects of the socks into changes in yarn geometric structure data, thereby achieving accurate identification of the defects.

[0073] In order to verify the recognition accuracy of the defect area of ​​the socks defect recognition model, it is compared with the image recognition technology; Product E of a sock knitting factory P is selected as an experimental object to obtain a sample of product E; defects are present in the sample of product E, and a defective area is marked to obtain a defective marked area; Acquire the three-dimensional design data and pre-processed three-dimensional sample data of the product E sample, input the three-dimensional design data and pre-processed three-dimensional sample data of the product E sample into the hosiery defect recognition model, and use the identified defect area as the first defect recognition area; The image recognition models are trained, including VGG-16, YOLOv8 and FasterRCNN; all product E samples are imaged to obtain product E sample image data; the VGG-16, YOLOv8 and FasterRCNN models are used to recognize the product E sample image data to obtain the corresponding second defect recognition area, third defect recognition area and fourth defect recognition area; The F1 score is obtained by measuring the overlapping areas of the defect marking area with the first defect identification area, the second defect identification area, the third defect identification area and the fourth defect identification area. The obtained data are shown in Table 3: Table 3 Socks defect recognition model validation data table

[0074] According to Table 3, the socks defect recognition model has the best recognition effect on defective areas.

[0075] The method for obtaining the difference threshold is: Obtaining qualified socks produced according to the three-dimensional design data of the socks; scanning the qualified socks by a high-resolution scanning device to obtain three-dimensional data of the qualified socks; The similarity between the qualified socks three-dimensional data and the three-dimensional design data is obtained; and a difference threshold is determined according to the similarity.

[0076] Qualified socks produced according to the three-dimensional design data of the socks are obtained, and the qualified socks are three-dimensionally scanned to obtain the three-dimensional data of the qualified socks; the qualified similarity between the three-dimensional data of the qualified socks and the three-dimensional design data of the socks is identified and used as the difference threshold to accurately determine the identified difference coefficient.

[0077] The defect data processing module obtains the defect areas and defect coefficients of all sock knitting samples, and marks them on the three-dimensional design data to obtain three-dimensional data of defect distribution.

[0078] The process of acquiring the three-dimensional defect distribution data is as follows: The defect areas and defect coefficients of the sock knitting samples are obtained in sequence, and the corresponding design sub-areas on the three-dimensional design data of the socks are obtained according to the defect area identification; According to the position distribution of the design sub-areas in the three-dimensional design data, the defect areas and defect coefficients of the sock knitting samples are marked; After marking the defect areas and defect coefficients of all sock knitting samples, three-dimensional defect distribution data are obtained.

[0079] The defect data identification module clusters the defect areas and defect coefficients on the three-dimensional data of defect distribution through the DBSCAN clustering algorithm to obtain the abnormal area; and obtains the abnormal coefficient based on the defect area and defect coefficient in the abnormal area.

[0080] The identification process of the DBSCAN clustering algorithm is: Acquire three-dimensional defect distribution data, establish a three-dimensional coordinate system according to the three-dimensional defect distribution data, and obtain three-dimensional coordinate data of the defect area; The distance data between the defective areas is calculated based on the three-dimensional coordinate data of the defective areas; and the defect weight is obtained based on the defect coefficient of the defective area; The defect weight is obtained by: obtaining the defect coefficients of all sock knitting samples to obtain the defect average value; and taking the ratio of the defect coefficient to the defect average value as the defect weight; There are differences in defect coefficients between different defect areas, and the defect coefficient reflects the severity of defects in the sock knitting samples; therefore, when performing clustering identification, it is necessary to assign weights to different defect areas according to the defect coefficient, so that the clustering algorithm pays more attention to areas with higher weights, ensuring that these areas have a greater impact on the results during the clustering process, so as to form a tighter and more effective cluster.

[0081] The calculation formula of the defect weight is: ; in, represents the defect weight; represents the defect coefficient; Indicates the average defect value; Then, clustering is performed according to the distance data and the defect weights using a DBSCAN clustering algorithm to obtain a clustering region, and the clustering region is used as an abnormal region.

[0082] The defect weight is calculated based on the defect coefficient in the three-dimensional data of defect distribution, and the distance data between the defective areas is identified based on the three-dimensional coordinate data. Then, clustering is performed based on the defect weight and the distance data between the defective areas to identify the abnormal areas, and accurately identify the areas where defects occur in the sock knitting samples.

[0083] In order to verify the recognition effect of DBSCAN clustering algorithm in clustering sock defect data, verification experiments were carried out based on several common clustering algorithms.

[0084] The DBSCAN clustering algorithm used in the present invention is a density-based clustering algorithm that does not require the number of clusters to be specified in advance and is suitable for data sets with noise and for situations where clusters of arbitrary shapes can be found. Common clustering algorithms also include K-means, spectral clustering, and Gaussian mixture models; K-means: The core idea is to group data points into K clusters by iteratively optimizing the position of cluster centers to minimize the sum of squared errors within the cluster so that the points within the cluster are as similar as possible and the points between clusters are as different as possible.

[0085] Spectral clustering: It is a clustering method based on graph theory, suitable for discovering clusters of complex shapes and non-spherical clusters; it relies on the similarity matrix of the data and uses the spectrum of the data to reduce the dimension.

[0086] Gaussian mixture model: It is a clustering algorithm based on a probability model, assuming that all data points are generated from a mixture of a finite number of Gaussian distributions, and provides each data point with the probability of belonging to each cluster.

[0087] The present invention collects and constructs a verification data set based on sock products woven and produced by a sock weaving factory P, and conducts a verification experiment of a clustering algorithm. The verification data set includes verification three-dimensional data and a verification area of ​​a verification sample, wherein the verification three-dimensional data includes a defect area and a defect coefficient, and the verification area is an area where an abnormality exists.

[0088] The verification three-dimensional data is identified by the clustering algorithm to obtain a test area, and the algorithm recognition effect is measured according to the area data of the verification area, the test area and the verification sample; The area data of the verification sample is the area of ​​the outer surface of the verification sample, and the area data of the verification area and the test area are the areas of their projection areas on the outer surface of the verification sample.

[0089] The present invention uses the F1 score as a measure of the recognition effect of the clustering algorithm, which is calculated based on the precision and recall rate. The calculation formula is: ; in, represents the F1 score; It represents the accuracy, which is the ratio of the overlapping area of ​​the verification area and the test area to the area of ​​the test area; Represents the recall rate, which is the ratio of the area of ​​overlap between the verification area and the test area to the area of ​​the verification area.

[0090] The data obtained through the verification experiment are shown in Table 4.

[0091] Table 4 Clustering algorithm verification data table

[0092] From the F1 scores in Table 4, it can be seen that the DBSCAN clustering algorithm has the best recognition effect.

[0093] After the abnormal area is identified from the three-dimensional data of defect distribution by DBSCAN clustering algorithm, the abnormal coefficient is calculated based on the defect area and defect coefficient in the abnormal area. The calculation formula of the abnormal coefficient is: ; in, represents the anomaly coefficient; Indicates the abnormal area The defect coefficient corresponding to each defect area; Indicates the number of defective areas contained in the abnormal area.

[0094] The abnormal warning module can issue warnings on the three-dimensional digital sock design and weaving process based on abnormal areas and abnormal coefficients.

[0095] Through the three-dimensional digitalized sock design and weaving integrated system described in the present invention, the sock weaving factory P realizes the abnormal monitoring of its sock digital design and weaving production, and timely discovers defects in the design and weaving process, which is helpful for its technical feedback improvement and sock cost control.

[0096] 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 three-dimensional digitalized hosiery design and weaving integrated system, characterized in that: include: Sock sample acquisition module, which acquires the three-dimensional design data of socks; and randomly sampling the socks produced according to the three-dimensional design data to obtain sock knitting samples; The sample data acquisition module uses a high-resolution scanning device to perform three-dimensional scanning on the sock knitting sample to obtain three-dimensional sample data; The sample data processing module matches the historical sock design data with the three-dimensional design data of the socks to obtain the raised control socks; Scanning the raised control socks to obtain standard raised data; De-noising the three-dimensional sample data according to the standard raising data to obtain pre-processed three-dimensional sample data; A sample defect recognition module is used to construct a hosiery defect recognition model, wherein the hosiery defect recognition model recognizes the pre-processed three-dimensional sample data according to the three-dimensional design data, and obtains a defect area and a defect coefficient of the pre-processed three-dimensional sample data according to the difference of the three-dimensional data of the corresponding area; A defect data processing module obtains defect areas and defect coefficients of all sock knitting samples and marks them on the three-dimensional design data to obtain three-dimensional defect distribution data; The defect data identification module clusters the defect areas and defect coefficients on the three-dimensional data of defect distribution through the DBSCAN clustering algorithm to obtain the abnormal area; the abnormal coefficient is obtained based on the defect area and defect coefficient in the abnormal area; The abnormal warning module can issue warnings on the three-dimensional digital sock design and weaving process based on abnormal areas and abnormal coefficients.

2. The three-dimensional digitalized hosiery design and weaving integrated system according to claim 1 is characterized by: The process of performing three-dimensional scanning on the sock knitting sample to obtain three-dimensional sample data is as follows: Acquire three-dimensional design data of socks, obtain cavity data of socks according to the three-dimensional design data, and obtain a cavity mold of socks according to the cavity data of socks; Putting a sock knitting sample onto the sock cavity mold; The sock knitting sample worn on the sock cavity mold is three-dimensionally scanned by a high-resolution scanning device to obtain three-dimensional sample data.

3. The three-dimensional digitalized hosiery design and weaving integrated system according to claim 1 is characterized by: The acquisition process of the raised control socks is as follows: Matching historical sock design data according to the three-dimensional design data of the socks, and determining standard socks according to the similarity of the design data; the historical sock data includes the three-dimensional design data of the historical socks; The standard socks are sampled to obtain standard socks samples, quality inspection is performed on the standard socks samples, and the standard socks samples with qualified quality are used as fuzzing control socks.

4. The three-dimensional digitalized hosiery design and weaving integrated system according to claim 1 is characterized by: The denoising process of the three-dimensional sample data according to the standard fuzz data is as follows: According to the method of performing three-dimensional scanning on the sock knitting sample to obtain three-dimensional sample data, performing three-dimensional scanning on the raising characteristics of the raising control socks to obtain standard raising data; The three-dimensional sample data is identified according to the standard raising data, and the three-dimensional data whose similarity with the standard raising data meets a threshold value in the three-dimensional sample data is deleted to obtain pre-processed three-dimensional sample data.

5. The three-dimensional digitalized hosiery design and weaving integrated system according to claim 1 is characterized by: The hosiery defect recognition model includes a design data recognition unit, a sample data recognition unit, a data difference recognition unit, and a defect area recognition unit; The design data identification unit is used to identify the three-dimensional design data and divide the three-dimensional design data into a plurality of design sub-areas; The sample data identification unit is used to identify the pre-processed three-dimensional sample data, and divide the pre-processed three-dimensional sample data into a plurality of sample sub-regions according to the way in which the three-dimensional design data is divided into design sub-regions; wherein the design sub-regions and the sample sub-regions are in a one-to-one correspondence; The data difference identification unit is used to identify the difference between the design sub-region and the sample sub-region in a corresponding relationship, and obtain the difference coefficient according to the data difference between the design sub-region and the sample sub-region; The defect area identification unit is used to judge the difference coefficient according to the difference threshold, take the difference coefficient greater than the difference threshold as the defect coefficient, and take the corresponding sample sub-area as the defect area.

6. The three-dimensional digitalized hosiery design and weaving integrated system according to claim 5 is characterized by: The method for obtaining the difference threshold is: Obtaining qualified socks produced according to the three-dimensional design data of the socks; scanning the qualified socks by a high-resolution scanning device to obtain three-dimensional data of the qualified socks; The data similarity between the qualified socks three-dimensional data and the three-dimensional design data is used as the qualified similarity; The difference threshold is determined based on the qualified similarity.

7. The three-dimensional digitalized hosiery design and weaving integrated system according to claim 5 is characterized by: The process of acquiring the three-dimensional defect distribution data is as follows: The defect areas and defect coefficients of the sock knitting samples are obtained in sequence, and the corresponding design sub-areas on the three-dimensional design data of the socks are obtained according to the defect area identification; According to the position distribution of the design sub-areas in the three-dimensional design data, the defect areas and defect coefficients of the sock knitting samples are marked; After marking the defect areas and defect coefficients of all sock knitting samples, three-dimensional defect distribution data are obtained.

8. The three-dimensional digitalized hosiery design and weaving integrated system according to claim 1 is characterized by: The identification process of the DBSCAN clustering algorithm is: Acquire three-dimensional defect distribution data, establish a three-dimensional coordinate system according to the three-dimensional defect distribution data, and obtain three-dimensional coordinate data of the defect area; The distance data between the defective areas is calculated based on the three-dimensional coordinate data of the defective areas; and the defect weight is obtained based on the defect coefficient of the defective area; Clustering is performed by combining the distance data and the defect weights using a DBSCAN clustering algorithm to obtain a clustering region, which is used as an abnormal region.

9. The three-dimensional digitalized hosiery design and weaving integrated system according to claim 8, characterized in that: The defect weight is obtained by: obtaining the defect coefficients of all sock knitting samples to obtain the defect average value; and taking the ratio of the defect coefficient to the defect average value as the defect weight; The calculation formula of the defect weight is: ； in, represents the defect weight; represents the defect coefficient; Indicates the average defect value.

10. The three-dimensional digitalized hosiery design and weaving integrated system according to claim 1, characterized in that: The calculation formula of the abnormal coefficient is: ； in, represents the anomaly coefficient; Indicates the abnormal area The defect coefficient corresponding to each defect area; Indicates the number of defective areas contained in the abnormal area.

Images