Cloth Piece Information Digitalization System and Its Method

Abstract

The present invention provides a cloth piece information digitization system and a cloth piece information digitization method. The cloth piece information digitization system includes an image capturing device and a computing device. The image capturing device acquires a cloth piece image. The computing device includes an image processing module. The computing device executes the image processing module to analyze the cloth piece image to obtain cloth piece classification information. The computing device inputs the cloth piece image into one of a plurality of neural network modules corresponding to different cloth piece classifications in the image processing module according to the cloth piece classification information to generate a normal layer and a roughness layer. The computing device integrates the cloth piece classification information, the normal layer, and the roughness layer to generate a cloth piece file. The cloth piece information digitization system and method of the present invention can realize the filing operation of cloth piece digital information.

Description

Technical Field

[0001] The present invention relates to an information digitization technology, and particularly to a cloth piece information digitization system and method thereof. Background Art

[0002] At present, the digitization of the textile industry has become the main global industrial trend. However, the bottleneck of textile digitization is how to efficiently convert physical cloth piece information into digital cloth piece data for subsequent application in the textile industry. However, to obtain the digital information of existing physical cloth pieces, a large number of even expensive measuring devices and professional measuring personnel are required to take images of the physical cloth pieces, and the measurement operations of relevant cloth piece specification parameters are carried out manually, resulting in the inability to efficiently convert physical cloth piece information into digital cloth piece data.

[0003] Moreover, the acquisition of the digital information of the aforementioned existing physical cloth pieces can often only obtain simple two-dimensional images and simple cloth piece specification parameters, resulting in too low practical application benefits. Furthermore, for general textile industry practitioners, they do not have the conditions to afford expensive measuring devices and hire professional measuring personnel. Therefore, the current cloth piece digitization means in the textile industry have not been effectively popularized, resulting in a slow process of textile industry digitization. Summary of the Invention

[0004] In view of this, the present invention provides a cloth piece information digitization system and method thereof, which can quickly and conveniently obtain digitized cloth piece information through image analysis and is conducive to the filing operation of cloth piece digital information.

[0005] The cloth piece information digitization system of the present invention includes an image capturing device and a computing device. The image capturing device obtains a cloth piece image. The computing device is coupled to the image capturing device and includes an image processing module. The computing device executes the image processing module to analyze the cloth piece image to obtain cloth piece classification information. The computing device inputs the cloth piece image into one of a plurality of neural network modules corresponding to different cloth piece classifications in the image processing module according to the cloth piece classification information to generate a normal layer and a roughness layer. The computing device integrates the cloth piece classification information, the normal layer, and the roughness layer to generate a cloth piece file.

[0006] The cloth piece information digitization method of the present invention includes the following steps: obtaining a cloth piece image through an image capturing device; analyzing the cloth piece image through a computing device executing an image processing module to obtain cloth piece classification information; inputting the cloth piece image into one of a plurality of neural network modules corresponding to different cloth piece classifications according to the cloth piece classification information through the computing device to generate a normal layer and a roughness layer; and integrating the cloth piece classification information, the normal layer, and the roughness layer through the computing device to generate a cloth piece file.

[0007] Based on the above, the fabric piece information digitization system and method of the present invention can analyze the fabric piece images to quickly obtain multiple digitized fabric piece information and fabric piece layer images, and establish files, so as to realize the rapid and convenient file establishment operation of the digitized fabric piece information.

[0008] To make the above features and advantages of the present invention more obvious and understandable, specific embodiments are hereinafter given and described in detail in conjunction with the accompanying drawings of the specification as follows. Description of the Drawings

[0009] Figure 1 is a schematic diagram of a fabric piece information digitization system according to an embodiment of the present invention;

[0010] Figure 2 is a flowchart of a fabric piece information digitization method according to an embodiment of the present invention;

[0011] Figure 3 is a schematic diagram of an image processing module according to an embodiment of the present invention;

[0012] Figure 4 is a schematic diagram of fabric piece classification performed by a fabric piece classification module according to an embodiment of the present invention;

[0013] Figure 5 is a schematic diagram of generating a normal layer and a roughness layer according to an embodiment of the present invention;

[0014] Figure 6 is a schematic diagram of a fabric piece information digitization system according to another embodiment of the present invention.

[0015] Description of the Reference Numerals

[0016] 100, 600: Fabric piece information digitization system;

[0017] 110: Image capture device;

[0018] 120: Computing device;

[0019] 121, 321, 622_1: Image processing module;

[0020] 122, 622_2: Fabric piece database;

[0021] 200: Physical fabric piece;

[0022] 301: Fabric piece color information;

[0023] 302: Fabric piece manufacturing information;

[0024] 303: Fabric piece weaving method information;

[0025] 304: Normal layer;

[0026] 305: Roughness layer;

[0027] 3211: Color analysis module;

[0028] 3212: Fabric piece classification module;

[0029] 3212_1: First classification hierarchy;

[0030] 3212_2: Second classification hierarchy;

[0031] 3213, 3213_1~3213_P: Neural network module;

[0032] 3213A: First generative adversarial network model;

[0033] 3213B: Second generative adversarial network model;

[0034] 3214: Image conversion module;

[0035] 3214_1: Color tone conversion module;

[0036] 3214_2: Color clustering conversion module;

[0037] 3214_3: Selection module;

[0038] 322: Fabric piece file;

[0039] 400: Fabric piece image;

[0040] 401: First color conversion image;

[0041] 402: Second color conversion image;

[0042] 403: Input image;

[0043] 610: Electronic device;

[0044] 611: Image capture unit;

[0045] 612, 623: Communication interface;

[0046] 620: Cloud server;

[0047] 621: Processor;

[0048] 622: Storage device;

[0049] A_1~A_N: Fabric piece manufacturing classification;

[0050] B_1~B_M: Fabric piece weaving method classification;

[0051] S210, S220, S230, S240: Steps. Detailed implementation

[0052] Reference will now be made in detail to exemplary embodiments of the present invention, examples of which are illustrated in the accompanying drawings. As far as possible, the same reference numerals are used in the drawings and the description to denote the same or similar parts.

[0053] Figure 1 is a schematic diagram of a cloth piece information digitization system according to an embodiment of the present invention. Refer to Figure 1 , the cloth piece information digitization system 100 includes an image capturing device 110 and a computing device 120. The computing device 120 includes an image processing module 121 and a cloth piece database 122. The image capturing device 110 is coupled to the computing device 120. The image capturing device 110 can be connected to the computing device 120 through wired or wireless communication methods and can transmit data to the computing device 120. In this embodiment, the computing device 120 may include a processor and a memory, and the memory may store data such as the image processing module 121, the cloth piece database 122, image data, and related algorithms or other modules for the processor to access and execute.

[0054] In this embodiment, the image capturing device 110 may be, for example, a flatbed scanner, a smart mobile device (such as a mobile phone or a tablet, etc.) or other forms of scanning devices. It should be noted that the image capturing device 110 may have an image scanning or image capturing function to directly obtain the cloth piece image, or the image capturing device 110 may receive the cloth piece image obtained by other scanning devices or electronic devices with image scanning or image capturing functions to indirectly obtain the cloth piece image. In this embodiment, the image capturing device 110 may transmit the cloth piece image to the computing device 120 so that the computing device 120 analyzes the cloth piece image by executing the image processing module 121. It should be noted that the cloth piece image obtained by the image capturing device 110 may be a two-dimensional image. The image processing module 121 can automatically analyze the cloth piece image to obtain corresponding multiple cloth piece information. In this embodiment, the computing device 120 can integrate these cloth piece information to establish a corresponding cloth piece file. The computing device 120 can directly store the cloth piece file in the cloth piece database 122, or first perform data conversion and / or data compression operations on the cloth piece file, and then store the converted cloth piece file in the cloth piece database 122 to complete the automatic filing operation.

[0055] In this embodiment, the aforementioned processor may be, for example, a Central Processing Unit (CPU), a Graphic Process Unit (GPU), or other programmable general-purpose or special-purpose microprocessors, Digital Signal Processors (DSPs), programmable controllers, Application Specific Integrated Circuits (ASICs), Programmable Logic Devices (PLDs), other similar processing devices, or a combination of these devices. The memory may be, for example, Dynamic Random Access Memory (DRAM), Flash memory, or Non-Volatile Random Access Memory (NVRAM), etc.

[0056] Figure 2 is a flowchart of a method for digitizing patch information according to an embodiment of the present invention. Refer to Figure 1 and Figure 2 , the patch information digitization system 100 can execute the following steps S210 to S240 to achieve patch information digitization. In this embodiment, a user can operate the image capture device 110. In step S210, the image capture device 110 acquires a patch image (two-dimensional image). The image capture device 110 can transmit the patch image to the computing device 120. In step S220, the computing device 120 executes the image processing module 121 to analyze the patch image to obtain patch classification information. In step S230, the computing device 120 inputs the patch image into one of a plurality of neural network modules corresponding to different patch classifications in the image processing module 121 according to the patch classification information to generate a Normal map and a Roughness map. In step S240, the computing device 120 integrates the patch classification information, the Normal map, and the Roughness map to generate a patch file. In this embodiment, the computing device 120 can combine the patch classification information, the Normal map, and the Roughness map to generate a patch file, and can store the patch file in the patch database 122 to complete the filing operation.

[0057] Therefore, the fabric piece information digitization system 100 and the fabric piece information digitization method of this embodiment can digitize the fabric piece information quickly and effectively, and can automatically generate a normal layer and a roughness layer that can be used for three-dimensional model (stereoscopic virtual image) modeling according to the two-dimensional image of the physical fabric piece. The fabric piece information digitization system 100 and the fabric piece information digitization method of this embodiment can automatically collect at least the aforementioned digitized information and layer data for filing to establish a fabric piece database 122. In addition, the implementation details of each step of this embodiment will be described in detail by the following multiple embodiments.

[0058] Figure 3 is a schematic diagram of an image processing module according to an embodiment of the present invention. Refer to Figure 3 , the image processing modules of the various embodiments of the present invention can implement the architecture of the image processing module 321 as shown in Figure 3 . In this embodiment, the image processing module 321 includes a color analysis module 3211, a fabric piece classification module 3212, and a plurality of neural network modules 3213_1 to 3213_P, where P is a positive integer. The image processing module 321 can input the fabric piece image 400 into the color analysis module 3211 and the fabric piece classification module 3212. It should be noted that, in one embodiment, before the image processing module 321 inputs the fabric piece image 400 into the color analysis module 3211 and the fabric piece classification module 3212, the image processing module 321 can first perform an image segmentation operation on the fabric piece image 400 to remove the non-fabric part in the image. And, the image processing module 321 can also perform color correction on the fabric piece color presented by the fabric piece image 400. For example, as shown in Figure 1 , the image capture device 110 can first read the color readings on a specific color paper, and the image processing module 321 can then use the difference between the reading result and the correct color reading of this specific color paper as a correction parameter to input it into a correction algorithm (such as a matrix calculation with limited interpolation). Therefore, the image processing module 321 can correct the pixel values of multiple pixels of the overall image of the fabric piece image 400 through this correction algorithm, so that the corrected fabric piece image 400 can have a color performance result closer to the real color.

[0059] In this embodiment, the color analysis module 3211 can analyze the fabric piece image 400 to generate fabric piece color information 301. Specifically, the color analysis module 3211 can generate an overall color histogram of the fabric piece image 400, and determine the color numbers in the fabric piece color information of the fabric piece image 400 according to at least one cluster peak of the overall color histogram. In other words, the color analysis module 3211 can pre-establish multiple color numbers (fabric color numbers) corresponding to multiple colors and multiple color combinations. Then, the color analysis module 3211 can determine at least one main color (or main color distribution) of the fabric piece image 400 through the results of multiple pixel values of the overall pixels statistically in the fabric piece image 400, and find the corresponding color number according to this at least one main color as the fabric piece color information 301.

[0060] In this embodiment, the fabric piece classification module 3212 can analyze the fabric piece image 400 to obtain fabric piece classification information, where the fabric piece classification information can include fabric piece manufacturing information 302 and fabric piece weaving method information 303. In this embodiment, the fabric piece manufacturing information 302 can be, for example, used to represent fabric piece manufacturing classifications such as woven fabric pieces (Woven) and knitted fabrics (Knit), etc., but the present invention is not limited thereto. In this embodiment, the fabric piece weaving method information 303 can be, for example, used to represent weaving methods such as poplin weave, twill weave, satin weave, single jersey, or ponte-de-roma, etc., but the present invention is also not limited thereto. In one embodiment, the definitions and classifications of the fabric piece manufacturing method and the fabric piece weaving method can be selectively defined according to different classification requirements.

[0061] In this embodiment, the neural network modules 3213_1 to 3213_P respectively correspond to different fabric piece classifications. The neural network modules 3213_1 to 3213_P can be respectively trained in advance using multiple sample images corresponding to different fabric piece manufacturing methods and different fabric piece weaving methods so that they can generate corresponding normal layers and roughness layers. For example, if the fabric piece image 400 is determined to be a cotton and linen fabric and is made by a plain weave method, the image processing module 321 can input the fabric piece image 400 into the neural network module corresponding to the cotton and linen fabric and the plain weave method, so that this neural network module can correspondingly generate a normal layer 304 and a roughness layer 305 that can faithfully reflect the characteristics of the corresponding physical fabric piece.

[0062] In this embodiment, the fabric piece color information 301, the fabric piece manufacturing information 302, the fabric piece weaving method information 303, the normal layer 304, and the roughness layer 305 can be integrated into a fabric piece file 322, and the fabric piece file 322 can be stored in such as Figure 1In the fabric piece database 122 shown, the fabric piece file 322 may further include the storage of the corresponding original fabric piece image 400. In addition, in one embodiment, the image processing module 321 may further mark the scale information of the fabric piece image 400 according to a preset image format, and may record the scale information in the fabric piece file 322. The scale information may be, for example, fabric piece size data such as fabric piece length, width, etc., and the present invention is not limited thereto. In another embodiment, the image processing module 321 may also automatically identify the text patterns in the fabric piece image 400 through natural language processing (Natural Language Understanding, NLU) and semantic understanding, etc., to obtain the corresponding text information, and record the text information in the fabric piece file 322. The text information may be, for example, the manufacturer information of the fabric piece manufacturer, the fabric piece number or the fabric piece name, etc., and the present invention is not limited thereto.

[0063] Figure 4 FIG. is a schematic diagram of fabric piece classification performed by a fabric piece classification module according to an embodiment of the present invention. Refer to Figure 3 and Figure 4 , in this embodiment, the fabric piece classification module 3212 may include a support vector machine (Support Vector Machine, SVM) classification model. The support vector machine classification model may be pre-trained to identify respective multiple image features in multiple sample images corresponding to different fabric piece manufacturing methods and different fabric piece weaves. As Figure 4 shown, the support vector machine classification model of this embodiment may, for example, implement two classification stages. The first classification level 3212_1 may, for example, classify the fabric manufacturing method based on the fabric piece features that are clearly distinguishable in the appearance of the physical fabric piece, and the second classification level 3212_2 may, for example, classify the weave details of the physical fabric piece. In this regard, when the fabric piece image 400 is input into the fabric piece classification module 3212, in the first classification level 3212_1, the fabric piece classification module 3212 may first identify and classify the fabric manufacturing method of the fabric piece image 400 to determine that the fabric piece image 400 can be classified into one of the fabric manufacturing classifications A_1 to A_N, and output the corresponding fabric manufacturing information 302, where N is a positive integer. Then, in the second classification level 3212_2, the fabric piece classification module 3212 may then identify and classify the fabric piece weave of the fabric piece image 400 to determine that the fabric piece image 400 can be classified into one of the fabric piece weave classifications B_1 to B_M, and output the corresponding fabric piece weave information 303, where M is a positive integer.

[0064] It should be noted that in this embodiment, each of the fabric sheet manufacturing classifications A_1 to A_N may correspond to the same multiple fabric sheet weaving classifications B_1 to B_M, but the present invention is not limited thereto. In one embodiment, each of the fabric sheet manufacturing classifications A_1 to A_N may correspond to different fabric sheet weaving classifications. However, the fabric sheet classification module 3212 of the present invention may also not be limited to the classification method and the number of classification levels as Figure 4 shown. Even in another embodiment, the fabric sheet classification module 3212 may also be implemented using other machine learning models, and the number of classifications and the classification method thereof are not limited either.

[0065] Figure 5 is a schematic diagram of generating a normal map layer and a roughness map layer according to an embodiment of the present invention. Refer to Figure 3 and Figure 5 , the image processing module 321 may further include an image conversion module 3214 to perform color conversion on the fabric sheet image 400. In this embodiment, the image conversion module 3214 may include a Color Hue Transformation module 3214_1, a Color Cluster Transformation module 3214_2, and a selection module 3214_3. In this embodiment, the Color Hue Transformation module 3214_1 may convert the fabric sheet image 400 to generate a first color conversion image 401. The Color Cluster Transformation module 3214_2 may convert the fabric sheet image 400 to generate a second color conversion image 402. The Color Hue Transformation module 3214_1 and the Color Cluster Transformation module 3214_2 may input the first color conversion image 401 and the second color conversion image 402 to the selection module 3214_3.

[0066] Specifically, in order to eliminate the color deviation caused by the textile color, the color tone conversion module 3214_1 can convert the cloth piece image 400 from the RGB (Red-Green-Blue) color space to the HSV (Hue-Saturation-Value) color space. Moreover, the color tone conversion module 3214_1 can set the hue channel of the image after converting the color space to a single value (i.e., the hue value of each pixel in the entire image is adjusted to a single value), and normalize the values of the saturation channel and the value channel of the image after converting the color space respectively (i.e., the value and saturation values of each pixel in the entire image are normalized respectively). In other words, the color tone conversion module 3214_1 can remove the background color in the cloth piece image 400, so that the cloth surface texture features of the physical cloth piece can be effectively highlighted, which is beneficial for the subsequent Generative Adversarial Network (GAN) model to effectively compare the normal map and roughness map of the drawing accuracy.

[0067] However, in some cases, when two color patterns in the cloth piece image 400 have the same hue but different saturation and value, these two color patterns cannot be fused after the aforementioned color tone conversion, so incorrect image conversion results may occur. Therefore, the color clustering conversion module 3214_2 can convert the cloth piece image 400 from the RGB color space to the HSV color space. Moreover, the color clustering conversion module 3214_2 can perform k-means clustering operations on the values of the hue channel, saturation channel, and value channel of the image after converting the color space respectively to separate different color groups, and then normalize the values of these channels respectively. In other words, the color clustering conversion module 3214_2 can use another method to remove the background color in the cloth piece image 400, so that the cloth surface texture features of the physical cloth piece can be effectively highlighted.

[0068] In this embodiment, the selection module 3214_3 can compare the overall color histograms of the first color conversion image 401 and the second color conversion image 402 to select the one with the highest cluster peak and / or the smallest number of cluster peaks. In other words, the selection module 3214_3 can determine which image has clearer texture or higher resolution characteristics by comparing the results of the pixel values of the overall pixels of the first color conversion image 401 and the second color conversion image 402 after statistics respectively. Then, the selection module 3214_3 can use one of the first color conversion image 401 and the second color conversion image 402 selected by the above judgment method as the input image 403, and input the input image 403 into the first generative adversarial network model 3213A and the second generative adversarial network model 3213B of the neural network module 3213 respectively. In this embodiment, the first generative adversarial network model 3213A can draw and output the normal layer 304, and the second generative adversarial network model 3213B can draw and output the roughness layer 305.

[0069] It should be noted that Figure 5 The neural network module 3213 can be Figure 3 One of the neural network modules 3213_1 to 3213_P. It should be noted that the multiple first generative adversarial network models and the multiple second generative adversarial network models of the neural network modules 3213_1 to 3213_P can be pre-trained using multiple sample images corresponding to different fabric manufacturing methods and different fabric weaves respectively, so that the generative networks (Generative Network) in the multiple first generative adversarial network models and the multiple second generative adversarial network models can generate different normal layers and different roughness layers corresponding to different fabric manufacturing methods and different fabric weaves, and these normal layers and these roughness layers generated by the generative network can be output through the identification of the discriminating networks (Discriminating Network) of the multiple first generative adversarial network models and the multiple second generative adversarial network models respectively. Therefore, the image processing module 321 of this embodiment can automatically generate the light and shadow features (normal layer 304) and grayscale features (roughness layer 305) that can faithfully reflect the characteristics of the physical fabric.

[0070] Figure 6 It is a schematic diagram of a fabric information digitization system according to another embodiment of the present invention. Refer to Figure 6 , the fabric information digitization system 600 includes an electronic device 610 and a cloud server 620. First, it should be noted that Figure 6 can be Figure 1 An application example of an embodiment, so Figure 1 The image capture device 110 of the embodiment can be implemented asFigure 6 an electronic device 610, and Figure 1 the computing device 120 of the embodiment can implement as Figure 6 a cloud server 620. In this embodiment, the electronic device 610 includes an image capturing unit 611 and a communication interface 612. The image capturing unit 611 is coupled to the communication interface 612. The cloud server 620 includes a processor 621, a storage device 622, and a communication interface 623. The processor 621 is coupled to the storage device 622 and the communication interface 623.

[0071] In this embodiment, the communication interface 612 and the communication interface 623 may each include a wired or wireless communication module and circuit, and the communication type of the present invention is not limited thereto. In this embodiment, the storage device 622 may store an image processing module 622_1 and a fabric piece database 622_2. The image processing module 622_1 can implement as Figure 1 the image processing module 121 of the embodiment. The fabric piece database 622_2 can implement as Figure 1 the fabric piece database 122 of the embodiment. In this embodiment, the electronic device 610 may be, for example, a flatbed scanner, and the image capturing unit 611 may be, for example, a CMOS image sensor (CIS) or a charge-coupled device (CCD).

[0072] For example, a user (which may refer to a person in the textile industry or an ordinary person) can place a physical fabric piece on the flatbed scanner (electronic device 610) so that the image capturing unit 611 can scan the physical fabric piece 200 to generate a two-dimensional fabric piece image 400. The flatbed scanner can directly send this fabric piece image 400 to the cloud server 620 through the communication interface 612. Alternatively, the user can operate, for example, a smartphone to connect to the flatbed scanner (the communication interface 612 of the electronic device 610) to obtain the fabric piece image 400 and send the fabric piece image 400 to the cloud server 620. Then, the processor 621 of the cloud server 620 can receive this fabric piece image 400 through the communication interface 623. The processor 621 of the cloud server 620 can automatically execute the image processing module 622_1 to perform image analysis and image processing operations as in the above embodiments to generate a corresponding fabric piece file. The processor 621 of the cloud server 620 can store the fabric piece file in the fabric piece database 622_2 to complete the filing operation of digital fabric piece information. In other words, the fabric piece information digitization system 600 of the present invention can automatically establish a corresponding file with fabric piece digital information with only one two-dimensional image. From another perspective, the user only needs to simply operate the flatbed scanner to achieve the automated filing operation of fabric piece digitized information.

[0073] In this way, in another extended implementation scenario, the same user or another user can, for example, read multiple fabric piece files in the fabric piece database 622_2 through any electronic device to, for example, perform online fabric piece management or online fabric selection operations. In this regard, the cloud server 620 can further include 3D model modeling software. The user can connect to the cloud server 620 by operating specific software or a web interface through any electronic device, and the cloud server 620 can execute the 3D model modeling software. Then, the 3D model modeling software can perform 3D model modeling based on the normal layer and roughness layer in each fabric piece file to generate a realistic 3D fabric piece model, enabling the user to view, through the display of any electronic device, a 3D fabric piece model that can present colors close to reality and has a high degree of realism. Moreover, the 3D fabric piece model can also be paired with its corresponding fabric piece color information, fabric piece manufacturing information, fabric piece weaving method information, etc., to enable the user to perform efficient and reliable online fabric piece management or online fabric selection operations.

[0074] In addition, the image processing operations and filing operations of this embodiment can refer to the descriptions of the above Figures 1 to 5 embodiments to obtain sufficient teachings, suggestions, and implementation descriptions, so they will not be elaborated here.

[0075] In summary, for the fabric piece information digitization system and method of the present invention, after obtaining a two-dimensional fabric piece image through an image capture device, relevant digitized information of the physical fabric piece can be automatically generated via a computing device, so as to achieve an automated fabric piece digital information filing operation. More importantly, since the fabric piece information digitization system and method of the present invention can easily obtain a two-dimensional fabric piece image through an image capture device such as a platform scanner and then automatically generate relevant fabric piece digitized information, without the need for complex and expensive equipment and cumbersome processes, general textile industry practitioners or the general public can adopt the fabric piece information digitization system and method of the present invention to achieve a cost-effective, efficient, and reliable fabric piece information digitization result. Moreover, the fabric piece information digitization system and method of the present invention can automatically generate a normal layer and a roughness layer based on the two-dimensional fabric piece image. In this regard, the 3D model modeling software generates a three-dimensional fabric piece model that can faithfully reflect the characteristics of the physical fabric piece based on the normal layer and the roughness layer, and can be further applied to, for example, online fabric piece management or online fabric selection operations. Therefore, the fabric piece information digitization system and method of the present invention are suitable for realizing relevant backend applications in textile industry digitization.

[0076] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements on some or all of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A cloth piece information digitization system, characterized in that, it includes: an image capture device for obtaining a cloth piece image; and an arithmetic device coupled to the image capture device and including an image processing module, wherein the arithmetic device executes the image processing module to analyze the cloth piece image to obtain cloth piece classification information, and the arithmetic device inputs the cloth piece image into one of a plurality of neural network modules corresponding to different cloth piece classifications according to the cloth piece classification information to generate a normal layer and a roughness layer, wherein the arithmetic device integrates the cloth piece classification information, the normal layer and the roughness layer to generate a cloth piece file, wherein the image processing module performs color conversion on the cloth piece image and inputs the color-converted cloth piece image into a first generative adversarial network model and a second generative adversarial network model of one of the plurality of neural network modules, so that the first generative adversarial network model outputs the normal layer and the second generative adversarial network model outputs the roughness layer, wherein the image processing module includes: a color tone conversion module for converting the cloth piece image to generate a first color conversion image; a color clustering conversion module for converting the cloth piece image to generate a second color conversion image; and a selection module for comparing two overall color histograms of the first color conversion image and the second color conversion image to select one with the highest clustering peak and / or the smallest number of clustering peaks to input into the first generative adversarial network model and the second generative adversarial network model of one of the plurality of neural network modules.

2. The cloth piece information digitization system according to claim 1, characterized in that, the image processing module further includes a color analysis module, and the color analysis module is used to analyze the cloth piece image to generate cloth piece color information, wherein the arithmetic device records the cloth piece color information in the cloth piece file.

3. The cloth piece information digitization system according to claim 2, characterized in that, the color analysis module is used to generate an overall color histogram of the cloth piece image, and the color analysis module judges the color number in the cloth piece color information of the cloth piece image according to at least one clustering peak of the overall color histogram.

4. The cloth piece information digitization system according to claim 1, characterized in that, the image processing module further includes a cloth piece classification module, and the cloth piece classification module is used to analyze the cloth piece image to obtain the cloth piece classification information, wherein the cloth piece classification information includes cloth piece manufacturing information and cloth piece weaving method information.

5. The cloth piece information digitization system according to claim 4, characterized in that, the cloth piece classification module includes a support vector machine classification model.

6. The cloth piece information digitization system according to claim 1, characterized in that, the image processing module marks the scale information of the cloth piece image according to the image format, and the arithmetic device records the scale information in the cloth piece file.

7. The fabric piece information digitization system according to claim 1, wherein, the image processing module is further configured to identify a text pattern in the fabric piece image to obtain corresponding text information, and the computing device records the text information in the fabric piece file.

8. The fabric piece information digitization system according to claim 1, wherein, the image capturing device is a flatbed scanner, and the computing device is a cloud server.

9. A fabric piece information digitization method, wherein, it includes: obtaining a fabric piece image through an image capturing device; performing, by a computing device, an image processing module to analyze the fabric piece image to obtain fabric piece classification information; inputting, by the computing device, the fabric piece image into one of a plurality of neural network modules corresponding to different fabric piece classifications in the image processing module according to the fabric piece classification information to generate a normal map layer and a roughness map layer; and integrating, by the computing device, the fabric piece classification information, the normal map layer, and the roughness map layer to generate a fabric piece file, wherein the steps of generating the normal map layer and the roughness map layer include: performing, by the computing device, the image processing module to perform color conversion on the fabric piece image, and inputting the color-converted fabric piece image into a first generative adversarial network model and a second generative adversarial network model of one of the plurality of neural network modules, so that the first generative adversarial network model outputs the normal map layer, and the second generative adversarial network model outputs the roughness map layer, wherein the steps of performing color conversion on the fabric piece image and inputting the color-converted fabric piece image into one of the plurality of neural network modules include: performing, by the computing device, a color tone conversion module of the image processing module to convert the fabric piece image to generate a first color-converted image; performing, by the computing device, a color clustering conversion module of the image processing module to convert the fabric piece image to generate a second color-converted image; and performing, by the computing device, a selection module of the image processing module to compare two overall color histograms of the first color-converted image and the second color-converted image, and select one with the highest clustering peak and / or the smallest number of clustering peaks to input into the first generative adversarial network model and the second generative adversarial network model of one of the plurality of neural network modules.

10. The fabric piece information digitization method according to claim 9, wherein, it further includes: performing, by the computing device, a color analysis module of the image processing module to analyze the fabric piece image to obtain fabric piece color information; and recording, by the computing device, the fabric piece color information in the fabric piece file.

11. The fabric piece information digitization method according to claim 10, wherein, the step of analyzing the fabric piece image to obtain the fabric piece color information includes: Generate an overall color histogram of the fabric piece image through the color analysis module, and determine the color number in the fabric piece color information of the fabric piece image according to at least one cluster peak of the overall color histogram.

12. The method for digitizing fabric piece information according to claim 9, wherein, The step of obtaining the fabric piece classification information includes: Analyze the fabric piece image through the fabric piece classification module of the image processing module by the computing device to obtain the fabric piece classification information, where the fabric piece classification information includes fabric piece manufacturing information and fabric piece weaving method information.

13. The method for digitizing fabric piece information according to claim 12, wherein, The fabric piece classification module includes a support vector machine classification model.

14. The method for digitizing fabric piece information according to claim 9, wherein, It further includes: Mark the scale information of the fabric piece image according to the image format through the computing device executing the image processing module; and Record the scale information into the fabric piece file through the computing device.

15. The method for digitizing fabric piece information according to claim 9, wherein, It further includes: Identify the text pattern in the fabric piece image through the computing device executing the image processing module to obtain the corresponding text information; and Record the text information into the fabric piece file through the computing device.

16. The method for digitizing fabric piece information according to claim 9, wherein, The image capture device is a flatbed scanner, and the computing device is a cloud server.

Images