A sketch recognition-based method and system for rapidly generating a three-dimensional clothing model

Abstract

The application relates to a sketch-recognition-based fast generation method and system of a garment three-dimensional model, and relates to the technical field of three-dimensional model generation, which comprises the following steps: collecting a garment sketch uploaded by a user and demand input information; generating a garment image according to the garment sketch; generating a garment initial three-dimensional model according to the garment features extracted from the garment image; calling input types and demand contents based on the demand input information; extracting the demand contents based on the input types to obtain input fabrics and input colors; combining the input fabrics and the input colors, optimizing the garment initial three-dimensional model to generate a garment final three-dimensional model, and outputting the garment final three-dimensional model. The application has the effect of improving the generation efficiency of the garment three-dimensional model.

Description

Technical Field

[0001] This invention relates to the field of 3D model generation technology, and in particular to a method and system for rapidly generating 3D clothing models based on sketch recognition. Background Technology

[0002] 3D model generation technology uses computer algorithms combined with input data such as images, point clouds, and text descriptions to construct digital models with three-dimensional spatial coordinates, textures, materials, and other attributes in virtual space through processes such as modeling, rendering, and optimization. It is widely used in fields such as games, film and television, industrial design, and medicine, and can efficiently realize the transformation from concept to visualized 3D entity.

[0003] In the process of generating a 3D clothing model from a sketch, the AI ​​algorithm first analyzes the clothing sketch uploaded by the user, extracts structural details and decorative elements such as outline, dividing lines, neckline, and cuffs, and then maps these two-dimensional features to the corresponding 3D clothing base template. Combined with preset fabric properties and clothing color, the physics engine automatically generates a complete 3D clothing model.

[0004] Currently, when generating 3D clothing models through sketch recognition, preset fabric properties and clothing colors are generally used. However, different users have different needs for fabric properties and clothing colors, which requires users to make secondary adjustments to the generated 3D clothing models, resulting in low efficiency in generating 3D clothing models. Summary of the Invention

[0005] To improve the efficiency of generating 3D clothing models, this invention provides a method and system for rapidly generating 3D clothing models based on sketch recognition.

[0006] In a first aspect, the present invention provides a method for rapidly generating 3D clothing models based on sketch recognition, employing the following technical solution:

[0007] A method for rapidly generating 3D clothing models based on sketch recognition, comprising:

[0008] S1: Collect clothing sketches and requirement input information uploaded by users;

[0009] S2: Generate a garment image based on the garment sketch;

[0010] S3: Extract clothing features from clothing images to generate an initial 3D model of the clothing;

[0011] S4: Retrieve input type and requirement content based on requirement input information;

[0012] S5: Extract the required content based on the input type to obtain the input fabric and input color;

[0013] S6: Combining the input fabric and input color, optimize the initial 3D model of the garment to generate the final 3D model of the garment, and output the final 3D model of the garment.

[0014] By adopting the above technical solution, the initial 3D model of the garment is generated from the garment sketch and the input information of the requirements. The input fabric and input color are extracted from the input information of the requirements. The initial 3D model of the garment is then optimized to generate the final 3D model of the garment and output. This ensures that the model restores the core design of the sketch and accurately matches the input fabric, color and other requirements uploaded by the user. This greatly improves the targeting and efficiency of model generation and reduces the deviation rate between design and model.

[0015] Optionally, methods for generating clothing images include:

[0016] S21: When the clothing sketch is an image, preprocess the clothing sketch to obtain a black and white line image;

[0017] S22: Obtain the initial image by removing isolated points from the black and white line image;

[0018] S23: Extract line contours based on the initial image;

[0019] S24: Filter valid lines based on line outlines to obtain a clear image;

[0020] S25: Smooth the image based on the clear image to obtain the clothing image.

[0021] By adopting the above technical solution, through step-by-step operations such as preprocessing, isolated point removal, contour extraction, effective line selection and smoothing, the image quality is gradually optimized to obtain clothing images with clear lines and accurate contours. This provides a high-quality data foundation for subsequent extraction of clothing features to generate an initial 3D model and reduces model distortion caused by image quality issues.

[0022] Optional methods for filtering valid lines include:

[0023] S241: Determine the line thickness characteristics, line texture characteristics, and line intersections based on the line outline;

[0024] S242: Determine the sketch style by combining the characteristics of line thickness and line texture;

[0025] S243: Determine the line classification attributes based on the sketch style;

[0026] S244: Classify line outlines according to line classification attributes to obtain line categories;

[0027] S245: Determine the number of intersections and their positions based on the intersection points of the lines;

[0028] S246: Select line contours by combining line type, number of intersections, and intersection position to obtain selected lines as valid lines.

[0029] By adopting the above technical solution, the sketch style is determined and classified by combining line thickness and texture characteristics. At the same time, the lines are selected by combining the number of intersections and their positions. This can accurately identify the core design lines under different sketch styles, eliminate invalid interference lines, and ensure that the extracted effective lines fit the design intent, laying the foundation for generating accurate clothing images and subsequent models.

[0030] Optional methods for determining the input fabric and color include:

[0031] S51: Determine whether the input type is the preset text type;

[0032] S52: If yes, semantic recognition is performed based on the requirements to obtain semantic fabric and semantic color, and semantic fabric is used as input fabric and semantic color is used as input color.

[0033] S53: If not, then identify the shadow area and the bright area based on the content requirements and adjust them to obtain a unified image;

[0034] S54: Based on the unified image, take color points at different locations to obtain the input color;

[0035] S55: Identify texture features based on the required sample image to obtain the sample image texture;

[0036] S56: Extract effective reflective features from the required sample image to obtain the fabric reflective intensity and fabric reflective texture;

[0037] S57: Determine the material properties of the fabric based on its reflectivity and reflective texture;

[0038] S58: Combine the sample image texture, fabric material properties, and input color to determine the estimated fabric, and use the estimated fabric as the input fabric.

[0039] By adopting the above technical solution, different recognition strategies are adapted to the input type. Text types are directly obtained through semantic recognition, while non-text types are determined through sample image processing, color point sampling, and texture and reflective feature extraction. This achieves compatibility and adaptation for different input types, ensuring the accuracy and comprehensiveness of input fabric and color recognition, and providing precise parameters for model optimization.

[0040] Optional methods for determining the fabric type include:

[0041] S581: Determine the fabric density based on the texture of the sample image;

[0042] S582: Determine the basic fabric structure by combining fabric density and fabric material properties;

[0043] S583: Retrieves color saturation, color uniformity, and color transmittance based on the input color;

[0044] S584: Determine the color prediction fabric by combining color saturation, color uniformity and color transmittance;

[0045] S585: Select the predicted fabric from the color prediction fabrics based on the fabric structure.

[0046] By adopting the above technical solution, the fabric density is obtained through the texture of the sample image, and the fabric structure is constructed by combining the fabric material properties. Then, the color saturation, color uniformity, and color transparency are adjusted by inputting the color to determine the color prediction fabric. Finally, the estimated fabric is obtained by selecting the fabric structure basis, realizing accurate fabric prediction from multiple dimensions of texture, material, and color, ensuring that the estimated fabric is highly consistent with the requirements, and improving the fabric reproduction accuracy of the subsequent model.

[0047] Optional methods for generating the final 3D model of the clothing include:

[0048] S61: Determine the natural wrinkle information and drape effect information of the fabric based on the estimated fabric;

[0049] S62: Generate adjustment points on the initial 3D model of the garment based on the information of natural fabric wrinkles and fabric drape.

[0050] S63: Determine the garment's posture based on the garment pattern;

[0051] S64: Adjust the initial 3D model of the garment according to the garment posture and adjustment points to obtain the garment posture model;

[0052] S65: The clothing pose model is fused based on the input color to obtain the clothing color model, and the clothing color model is used as the final 3D model of the clothing.

[0053] By adopting the above technical solution, and by setting adjustment points based on the estimated folds and drape of the fabric, and then coloring the model after adjusting the garment posture, the model not only restores the design outline, but also has a natural shape and accurate posture that conforms to the characteristics of the fabric. At the same time, it accurately presents the input color, significantly improving the realism and design fidelity of the final model.

[0054] Optional methods for determining clothing posture include:

[0055] S631: Identify joint locations and adjustable points in the garment based on the garment pattern scan;

[0056] S632: Determine joint gravity parameters by combining the joint location points of the garment with the estimated fabric;

[0057] S633: Determine the locally adjustable points of the model based on the initial 3D model of the garment;

[0058] S634: Determine whether the adjustable points of the clothing are consistent with the adjustable points of the model;

[0059] S635: If not, determine the natural posture of the garment based on the joint gravity parameters and use the natural posture of the garment as the garment posture.

[0060] S636: If yes, then determine the posture adjustment parameters based on the local adjustable points of the model and the local adjustable points of the clothing;

[0061] S637: Determine the posture of the local adjustment point based on the posture adjustment parameters and joint gravity parameters, and use the posture of the local adjustment point as the posture of the clothing.

[0062] By adopting the above technical solution, the joint position points and local adjustable points of the garment are identified, the joint gravity parameters are determined by combining the estimated fabric, and the local adjustable points of the garment are determined by the three-dimensional model of the garment. Then, by comparing the consistency between the local adjustable points of the model and the local adjustable points of the garment, the natural posture or the posture of the local adjustment points is determined and used as the posture of the garment. This ensures that the posture conforms to the physical properties of the fabric and adapts to the design requirements, avoids posture distortion, and makes the posture of the model more closely match the actual wearing effect.

[0063] Optionally, after outputting the final 3D model of the garment, the following may also be included:

[0064] S71: Collect user-uploaded images of people wearing tight clothing as reference images for users;

[0065] S72: Perform feature recognition based on the user reference image to obtain a user feature image;

[0066] S73: Determine specific body data based on user profile images;

[0067] S74: Identify and record body characteristics based on specific body data;

[0068] S75: Adjust the final 3D model of the clothing based on body characteristics.

[0069] By adopting the above technical solution, body characteristics are obtained by collecting and analyzing images of users' tight-fitting clothing, and then the final 3D model of the clothing is adjusted based on the body characteristics, thereby improving the personalized adaptation capability of the model and providing a precise body shape matching basis for subsequent virtual try-on scenarios.

[0070] Optionally, adjustments to the final 3D model of the clothing based on body characteristics include:

[0071] S751: Determine model features based on the final 3D model of the garment;

[0072] S752: Based on the deviation between model features and body features, the feature deviation value is obtained;

[0073] S753: Determine the feature position and feature adjustment value based on the feature deviation value;

[0074] S754: Adjust the final 3D model of the garment based on feature location and feature adjustment value.

[0075] By adopting the above technical solution, the model features are determined through the final three-dimensional model of the clothing, and the feature positions and feature adjustment values ​​are determined by the deviation between the model features and body features, thereby improving the model's adaptation accuracy in key body shape dimensions.

[0076] Secondly, the present invention provides a rapid generation system for 3D clothing models based on sketch recognition, employing the following technical solution:

[0077] A system for rapidly generating 3D clothing models based on sketch recognition, comprising:

[0078] The data acquisition module is used to collect clothing sketches, requirement input information, and user reference images;

[0079] The memory stores a program for implementing a method for rapidly generating a 3D clothing model based on sketch recognition as described in any of the first aspects;

[0080] The processor loads and executes programs stored in memory.

[0081] In summary, the present invention has at least one of the following beneficial technical effects:

[0082] 1. By collecting clothing sketches and requirement input information, an initial 3D model of the clothing is generated from the clothing sketches. The input fabric and input color are extracted from the requirement input information. The initial 3D model of the clothing is then optimized to generate the final 3D model of the clothing and output. This ensures that the model reproduces the core design of the sketch and accurately matches the input fabric, color and other requirements uploaded by the user, which greatly improves the targeting and efficiency of model generation and reduces the deviation rate between design and model.

[0083] 2. The fabric density is obtained through the texture of the sample image. The fabric structure is constructed by combining the fabric material properties. Then, the color saturation, color uniformity and color transparency are adjusted by inputting the color to determine the color prediction fabric. Finally, the estimated fabric is obtained by selecting the fabric structure basis. This achieves accurate fabric prediction from multiple dimensions of texture, material and color, ensuring that the estimated fabric is highly consistent with the requirements and improving the fabric reproduction of the subsequent model.

[0084] 3. Determine model features through the final 3D model of the clothing, and determine feature positions and adjustment values ​​by the deviation between model features and body features, thereby improving the model's adaptation accuracy in key body shape dimensions. Attached Figure Description

[0085] Figure 1 This is a flowchart of a method for rapidly generating 3D clothing models based on sketch recognition;

[0086] Figure 2 This is a flowchart illustrating the method for generating clothing images;

[0087] Figure 3 This is a flowchart illustrating the method for determining the input fabric and color.

[0088] Figure 4 This is a flowchart illustrating the method for generating the final 3D model of the garment. Detailed Implementation

[0089] The present invention will now be described in further detail with reference to the accompanying drawings and embodiments.

[0090] A rapid method for generating 3D clothing models based on sketch recognition is proposed. This method involves collecting clothing sketches and input requirements, preprocessing them, removing isolated points, extracting contours, filtering effective lines, and smoothing to generate clothing images. Then, clothing features are extracted to obtain an initial 3D model. Input fabric and color are extracted according to the requirement type. Combined with fabric fold / drape information and clothing posture adjustments, and coloring, the final model is generated. After output, images of the user's tight-fitting clothing can be collected to extract body features, and key dimensions such as upper arm width can be adjusted to match the model. This ensures that the model accurately reproduces the core design of the sketch while precisely matching the user's uploaded input fabric, color, and other requirements, significantly improving the relevance and efficiency of model generation and reducing the deviation rate between design and model.

[0091] Reference Figure 1 This invention discloses a method for rapidly generating 3D clothing models based on sketch recognition, comprising:

[0092] S1: Collect clothing sketches and requirement input information uploaded by users.

[0093] In this context, a garment sketch refers to a line drawing created by a designer or user to express the core design intent of a garment. Garment sketches can be scanned images obtained after hand-drawing, or they can be created using professional 2D software.

[0094] Requirement input information refers to the text or image information used to indicate the intended clothing style.

[0095] Both clothing sketches and requirement input information are obtained through user input.

[0096] S2: Generate a clothing image based on the clothing sketch.

[0097] Among them, clothing images refer to two-dimensional images that have been processed to have complete line outlines and visual effects.

[0098] By recognizing the format and lines of clothing sketches, when the clothing sketch is drawn using professional 2D software, the lines within the clothing sketch are directly recognized to generate a clothing image.

[0099] When the garment sketch is a scanned image obtained after being hand-drawn, the garment sketch is analyzed and processed to generate a garment image for subsequent use. The specific steps for generating the garment image are described in S21 to S25.

[0100] To further ensure the validity of the clothing images, further separate analysis and calculation of the clothing images are required, which will be explained in detail through the steps shown below.

[0101] Reference Figure 2 The method for generating clothing images includes the following steps:

[0102] S21: When the clothing sketch is an image, preprocess it based on the clothing sketch to obtain a black and white line image.

[0103] Among them, black and white line art refers to images that contain only black clothing lines and a white background.

[0104] When the clothing sketch is an image, it means that the clothing sketch was obtained by scanning after being hand-drawn. Therefore, by using a grayscale algorithm to convert the color image into a grayscale image containing only different grayscale values, and then using an adaptive threshold binarization algorithm to convert the background pixels to white and the pixels that meet the line characteristics to black, a black and white line image is obtained, which is convenient for subsequent use.

[0105] Both the grayscale conversion algorithm and the adaptive threshold binarization algorithm are existing technologies and will not be elaborated further.

[0106] S22: Obtain the initial image by removing isolated points from the black and white line image.

[0107] Isolated points refer to isolated pixels or tiny clusters of pixels in a black-and-white line image that have no connection to the core lines of the garment. Isolated points are caused by interference factors such as pen smudges and scanning noise during sketching. The initial image refers to the image obtained after removing isolated points.

[0108] By performing connected component detection on the black and white line image, counting the number of pixels in each connected pixel cluster, and identifying connected pixel clusters with a pixel count lower than a preset connectivity threshold as isolated points, the isolated point pixels are then converted from black to white through pixel whitening, thereby eliminating these interference points.

[0109] The connectivity baseline threshold refers to the minimum number of connections required for a device to be in a non-isolated state under normal circumstances. The connectivity baseline threshold is preset by the operator.

[0110] S23: Extract line contours based on the initial image.

[0111] Among them, the line outline refers to the set of continuous lines in the initial image that conform to the structural characteristics of clothing.

[0112] By extracting the portion of the initial image excluding the background, the line outline is obtained, which facilitates subsequent use.

[0113] S24: Filter valid lines based on line outlines to obtain a clear image.

[0114] A clear image refers to an image obtained after filtering the lines.

[0115] By filtering the line outlines, auxiliary lines and annotation lines are removed to obtain valid lines. The valid lines are then combined in their original positions to form a clear image.

[0116] To further ensure the rationality of the effective lines, it is necessary to perform further separate analysis and calculation on the effective lines, which will be explained in detail through the steps shown below.

[0117] The method for selecting valid lines includes the following steps:

[0118] S241: Determine the line thickness, texture, and intersection points based on the line outline.

[0119] Among them, line thickness features refer to the width attribute of each line segment in the line outline. Line thickness features include the overall width range of the line and the thickness variation in different areas.

[0120] The textural characteristics of lines refer to the texture and style of brushstrokes. These characteristics include the irregular strokes of hand-drawn lines, the gray-scale gradations of pencil lines, and the sharp edges of pen lines.

[0121] A line intersection point refers to the pixel location formed by the intersection of two or more line segments in a line outline.

[0122] By scanning each line segment in the outline vertically point by point and counting the number of consecutive black pixels as the line width at that position, the overall thickness range is obtained by calculating the mean and the degree of thickness variation is obtained by variance analysis, thus obtaining the line thickness characteristics for subsequent use.

[0123] By identifying the grayscale variation values ​​within each line area and the abrupt grayscale changes at the edges, the texture characteristics of the lines are obtained, facilitating subsequent use. Grayscale variation values ​​refer to the degree of grayscale value change within the line's interior area, while abrupt grayscale changes at the edges refer to the sharpness of the grayscale change at the line's edges.

[0124] By identifying the intersections between each line segment in the line outline, the line intersection points can be obtained, which is convenient for subsequent use.

[0125] S242: Determine the sketch style by combining the characteristics of line thickness and line texture.

[0126] Sketch style refers to the style used when drawing a sketch. Sketch styles include regular mechanical style and irregular hand-drawn style.

[0127] The width and thickness range are retrieved by analyzing line thickness characteristics, and the grayscale variation values ​​within the area and the edge grayscale abrupt changes are retrieved by analyzing line texture characteristics. A regular, mechanical style corresponds to smaller variations in line thickness and grayscale values, but larger edge grayscale abrupt changes. Conversely, an irregular, hand-drawn style corresponds to larger variations in line thickness and grayscale values, but smaller edge grayscale abrupt changes.

[0128] By comparing the width range, thickness variation, grayscale variation value and abrupt change with the corresponding preset benchmark values, when all are satisfied, the regular mechanical style is used as the sketch style; when not all are satisfied, the irregular hand-drawn style is used as the sketch style.

[0129] Preset reference values ​​refer to the various reference values ​​corresponding to the width range, thickness variation, grayscale variation value and sudden change when regularizing the mechanical style. The preset reference values ​​are set in advance by the operator.

[0130] S243: Determine the line classification attributes based on the sketch style.

[0131] Among them, the line classification attribute refers to the functional category identifier of the line based on the sketch style adaptability. The line classification attribute is used to identify different lines.

[0132] Different sketch styles correspond to different line classification attributes. By inputting the sketch style into a preset style database, the line classification attributes are matched and obtained for convenient subsequent use.

[0133] The style database pre-stores a lookup table of different sketch styles and their corresponding line classification attributes. The style database is obtained after the operator pre-inputs the data. When the sketch style is a regular mechanical style, the line classification attributes corresponding to the core structural lines, annotation lines, and auxiliary lines are all preset by the 2D software. When the sketch style is an irregular hand-drawn style, the line classification attributes corresponding to the core structural lines, annotation lines, and auxiliary lines are obtained by the operator performing feature recognition on different hand-drawn images and then inputting the results into a deep learning model.

[0134] S244: Classify the line outlines according to the line classification attributes to obtain the line category.

[0135] Among them, line category refers to the category corresponding to each line segment in the line outline. Line categories include core structure lines, annotation lines, and auxiliary lines.

[0136] By classifying the line outlines according to line classification attributes, line categories are obtained, which facilitates subsequent use.

[0137] S245: Determine the number of intersections and their positions based on the intersection points of the lines.

[0138] The number of intersections refers to the total number of line intersections confirmed by detection within the line outline. The intersection location refers to the position of each intersection point within the garment structure.

[0139] By counting the intersections of the lines and taking the count as the number of intersections, and then taking the position of the intersection point in the garment structure as the intersection position, it is convenient for subsequent use.

[0140] S246: Select line contours by combining line type, number of intersections, and intersection position to obtain selected lines as valid lines.

[0141] Here, "selecting lines" refers to the lines that are selected.

[0142] By selecting the core structural lines corresponding to the line categories and obtaining initial lines, the reference values ​​corresponding to each garment structural position are obtained by calculating the product between the number of intersections and the preset proportion of the garment structural position. Then, the number of intersections of the same line is obtained by counting the number of intersections of the same line. The initial lines with the same number of intersections less than the reference value are selected to obtain selected lines. The selected lines are then used as valid lines, thereby improving the accuracy of obtaining valid lines.

[0143] S25: Smooth the image based on the clear image to obtain the clothing image.

[0144] Smoothing refers to the process of optimizing clothing lines in a clear image.

[0145] By smoothing clear images through filtering, denoising, morphological restoration, and curve fitting, clothing images can be obtained, thus improving the accuracy of the acquired clothing images.

[0146] S3: Extract clothing features from clothing images to generate an initial 3D model of the clothing.

[0147] Among them, clothing characteristics refer to the set of key information extracted from clothing images that reflects the core design and structure of the clothing. Clothing characteristics include the silhouette (such as H-shape, A-shape, and fitted waist), structural details (such as neckline type, cuff style, and seam position), size proportions (such as the relative proportions of garment length, shoulder width, and sleeve length), and parameters of key parts (such as neckline curvature, waistline height, and skirt unfolding angle).

[0148] The initial 3D model of clothing refers to a 3D mesh model generated based on the characteristics of clothing, which has the basic form and structure of clothing.

[0149] By employing contour fitting algorithms (such as polygon approximation + Bézier curve optimization) to analyze the outer contour of the garment image, the overall shape of the garment (such as the narrow-at-the-top and wide-at-the-bottom contour of an A-line dress) is determined. Subsequently, object detection algorithms (such as YOLO and SSD) are used to identify key areas such as the neckline, cuffs, and waistline, classifying and determining the type of these areas (such as round neck, square neck, and bell sleeves). Finally, key parameters (such as neckline diameter, sleeve length ratio, and waistline height) are quantified through pixel coordinate conversion (combined with image resolution and a preset scale), integrating these parameters to form the garment's characteristics. These characteristics are then input into a pre-stored library of basic templates for different garment categories to match corresponding templates. Non-rigid deformation algorithms (such as thin-plate spline interpolation) are used to adjust the template's contour shape, key component dimensions, and structural relationships, ensuring a precise match between the template and the garment's characteristics. This ultimately generates an initial 3D model of the garment with a complete three-dimensional structure, facilitating subsequent use.

[0150] S4: Retrieve input type and requirement content based on requirement input information.

[0151] Input type refers to the format classification of the input information. Input types include text and image types. Requirement content refers to the specific details of the requirements conveyed by the input information. Requirement input information includes both input type and requirement content.

[0152] The input type and required content can be retrieved by inputting the requirements information, which facilitates subsequent use.

[0153] S5: Extract the required content based on the input type to obtain the input fabric and input color.

[0154] Here, "Input Fabric" refers to the fabric requested by the user, and "Input Color" refers to the color requested by the user.

[0155] The input type is used to extract the required information, thereby obtaining the input fabric and input color for convenient subsequent use.

[0156] To further ensure the rationality of the input fabric and color, it is necessary to perform further separate analysis and calculation on the input fabric and color, which will be explained in detail through the steps shown below.

[0157] Reference Figure 3 The method for determining the input fabric and color includes the following steps:

[0158] S51: Determine if the input type is a preset text type. If yes, proceed to S52; if no, proceed to S53.

[0159] Among them, text type refers to the type of plain text, which is obtained after the operator pre-inputs.

[0160] By judging whether the input type is a preset text type, it can be determined whether semantic recognition can be performed directly.

[0161] S52: Based on the requirements, perform semantic recognition to obtain semantic fabric and semantic color, and use semantic fabric as input fabric and semantic color as input color.

[0162] Semantic fabric refers to the type of fabric explicitly requested by the user. Semantic color refers to the color explicitly requested by the user.

[0163] When the input type is a preset text type, it means that semantic recognition can be performed directly. Therefore, the required fabric type is parsed from the text in the requirement content through semantic recognition as semantic fabric, and the required color is parsed as semantic color. The semantic fabric is used as the input fabric, and the semantic color is used as the input color, thereby improving the accuracy of the obtained input fabric and input color.

[0164] S53: Identify shadow and highlight areas based on the required content and adjust them to obtain a unified image.

[0165] In this context, shadow areas refer to the dark areas in an image caused by insufficient lighting or occlusion. High-brightness areas refer to... A uniform image refers to the bright areas in an image caused by excessive lighting.

[0166] When the input type is not the preset text type, it means that semantic recognition cannot be performed directly. Therefore, the required content is processed into grayscale and converted into a single-channel grayscale image to simplify brightness analysis. Then, an adaptive brightness threshold algorithm is used to dynamically set the shadow threshold and high brightness threshold according to the overall brightness distribution. Areas with grayscale values ​​lower than the shadow threshold are identified as shadow areas, and areas with grayscale values ​​higher than the high brightness threshold are identified as high brightness areas. Next, the Gamma correction algorithm is used to brighten the shadow areas and darken the high brightness areas. At the same time, histogram equalization technology is used to optimize the overall contrast of the sample image to ensure that the brightness transition between different areas is natural. Finally, the adjusted images are integrated to generate a uniform sample image with uniform brightness and clear details for convenient subsequent use.

[0167] S54: Take color points at different locations based on the unified image to obtain the input color.

[0168] The input color refers to the color data corresponding to different positions.

[0169] By taking color points at various locations on a unified image and averaging the color data at each location, the average value is used as the input color for convenient subsequent use.

[0170] S55: Identify texture features based on the required sample image to obtain the sample image texture.

[0171] Texture features refer to the regular visual representation of the fabric surface in the sample image. Texture features include the repetition pattern of the texture (such as the repetition period of checks and stripes), directional distribution (such as the angle of vertical stripes and diagonal checks), thickness (such as the pixel spacing of coarse linen texture and fine cotton texture), and grayscale / color variation patterns. Sample image texture refers to standardized texture data obtained by extracting texture features.

[0172] By segmenting the texture region of the sample image and removing non-fabric areas at the edges, the Gray-Level Co-occurrence Matrix (GLCM) is used to calculate statistical features such as contrast, correlation, and energy of the texture, reflecting its coarseness and uniformity. Local Binary Pattern (LBP) is used to extract local structural features of the texture, capturing the uneven texture of the fabric surface (such as the fiber texture of cotton and linen). Fourier transform is used to analyze the frequency characteristics of the texture, determining the period and direction of repeating patterns (such as the repetition spacing of checks and the tilt angle of stripes). Finally, the extracted statistical features, local structural features, frequency features, and other multi-dimensional features are integrated to generate standardized texture image blocks (such as 256×256 pixel repeating texture units) and feature parameter sets, which constitute the sample image texture.

[0173] S56: Extract effective reflective features based on the required sample image to obtain the fabric reflective intensity and fabric reflective texture.

[0174] Effective reflective features refer to the set of visual features that truly reflect the inherent reflective properties of a fabric. These features include the brightness distribution, area ratio, gradient changes, and spectral reflectance patterns of reflective areas. Fabric reflectivity refers to the intensity of light reflected from the fabric. Fabric reflective texture refers to the stylistic characteristics of the fabric's reflectivity. Fabric reflective textures include specular reflection, soft diffuse reflection, micro-pearl reflection, and matte reflection.

[0175] By extracting effective reflective features such as brightness distribution, area ratio, gradient change, and spectral reflectance from the required sample image, the reflective intensity of the fabric is calculated through brightness distribution. The brightness distribution, area ratio, gradient change, and spectral reflectance are combined to calculate the brightness variance, reflective area ratio, edge gradient value, and spectral reflectance ratio. These are then input into a preset reflective texture classification model to output the reflective texture of the fabric for subsequent use.

[0176] The reflective texture classification model is obtained by pre-inputting different brightness variances, reflective area proportions, edge gradient values, and spectral reflectance ratios to output the reflective texture of the fabric.

[0177] S57: Determine the material properties of the fabric based on its reflectivity and reflective texture.

[0178] Among them, fabric material attributes refer to the inherent material category of the fabric (such as silk, cotton, leather, chemical fiber, etc.).

[0179] By inputting the fabric reflectivity and reflective texture into a preset fabric material database, the fabric material properties are matched and obtained for convenient subsequent use.

[0180] The fabric material database has a pre-stored table of different fabric reflectivity, fabric reflectivity texture and corresponding fabric material properties. The fabric material database is obtained by the operator in advance by detecting the reflectivity of different fabric material properties and pre-inputting the corresponding fabric reflectivity texture.

[0181] S58: Combine the sample image texture, fabric material properties, and input color to determine the estimated fabric, and use the estimated fabric as the input fabric.

[0182] Among them, the estimated fabric refers to the fabric corresponding to the fabric in the sample picture after estimation.

[0183] By combining and analyzing the sample image texture, fabric material properties, and input color, the estimated fabric is determined and used as the input fabric, thereby improving the accuracy of the obtained input fabric.

[0184] To further ensure the reasonableness of the estimated fabric, it is necessary to perform further separate analysis and calculation on the estimated fabric, which will be explained in detail through the steps shown below.

[0185] The method for determining the fabric to be estimated includes the following steps:

[0186] S581: Determine the fabric density based on the texture of the sample image.

[0187] Fabric density refers to the number of yarns interlacing per unit area of ​​the fabric (for woven fabrics) or the density of loop arrangement (for knitted fabrics).

[0188] By extracting the overall spacing and number of textures from the sample image texture, and then calculating the ratio between the overall spacing and the number of textures, the calculation result is used as the fabric density for subsequent use.

[0189] S582: Determine the basic fabric structure by combining fabric density and fabric material properties.

[0190] Among them, the basic structure of fabric refers to the combination system of the core structural types of fabric (such as woven structure, knitted structure, and nonwoven structure) and corresponding structural parameters (such as the weave type of woven fabric, the loop structure of knitted fabric, and the bonding method of nonwoven fabric).

[0191] By inputting the fabric density and fabric material properties into a preset structural database, a fabric structural basis is obtained for easy subsequent use.

[0192] The structural foundation database pre-stores a table of different fabric densities, fabric material properties, and corresponding fabric structural foundations. The structural foundation database is accessed after the operator pre-inputs the information.

[0193] The structural database pre-stores the fabric structural basics corresponding to different densities for materials such as cotton, silk, denim, leather, wool, and chemical fibers.

[0194] S583: Retrieves color saturation, color uniformity, and color transparency based on the input color.

[0195] Color saturation refers to the quantitative indicator of the ratio of pure color components to gray components in a color. Color uniformity refers to the consistency of the distribution of the input color across a fabric area. Color transmittance refers to the degree to which light is allowed to pass through the fabric.

[0196] By converting the input color to the HSV color space, the S channel value is used as the color saturation, and the V channel value is used as the color transmittance. Then, the color uniformity is obtained by calculating the distribution dispersion of the input color, which is convenient for subsequent use.

[0197] The higher the V channel value, the higher the color transmittance; the lower the V channel value, the lower the color transmittance. The greater the distribution dispersion, the lower the color uniformity; the smaller the distribution dispersion, the higher the color uniformity.

[0198] S584: Determine the color prediction fabric by combining color saturation, color uniformity and color transparency.

[0199] Color prediction fabrics refer to fabric categories that are adapted to color performance.

[0200] By inputting color saturation, color uniformity, and color transparency into a preset color prediction database, color prediction fabrics are obtained for convenient subsequent use.

[0201] The color prediction database pre-stores a table of different color saturation, color uniformity, color transparency, and corresponding color prediction fabrics. The color prediction database is stored after the operator has pre-tested the color saturation, color uniformity, and color transparency of different color prediction fabrics.

[0202] S585: Select the predicted fabric from the color prediction fabrics based on the fabric structure.

[0203] Specifically, by inputting the color prediction fabric and the basic fabric structure into a preset structure matching database to obtain a structure matching value, and selecting the color prediction fabric with the largest structure matching value as the estimated fabric, the accuracy of the obtained estimated fabric is improved.

[0204] The structural fit value refers to the degree of fit between the color prediction fabric and the fabric structure base when the fabric is manufactured. The higher the structural fit value, the better the match between the fabric structure base and the color prediction fabric.

[0205] The structure adaptation database pre-stores a table of different color prediction fabrics, fabric structure bases, and corresponding structure adaptation values. The structure adaptation database is obtained after the operator pre-inputs the data.

[0206] S6: Combining the input fabric and input color, optimize the initial 3D model of the garment to generate the final 3D model of the garment, and output the final 3D model of the garment.

[0207] Among them, the final 3D model of clothing refers to a 3D mesh model that has the characteristics of clothing fabric and color.

[0208] By optimizing the initial 3D model of the garment based on the input fabric and color, the final 3D model of the garment is generated and output. This ensures that the model accurately reproduces the core design of the sketch and precisely matches the user's uploaded input fabric, color, and other requirements, greatly improving the relevance and efficiency of model generation and reducing the deviation rate between design and model.

[0209] To further ensure the rationality of the final 3D model of the garment, it is necessary to perform further separate analysis and calculations on the final 3D model of the garment, which will be explained in detail through the steps shown below.

[0210] Reference Figure 4 The method for generating the final 3D model of the clothing includes the following steps:

[0211] S61: Determine the natural folds and drape of the fabric based on the estimated fabric size.

[0212] Among them, fabric natural wrinkle information refers to the set of wrinkle features describing the fabric when it is naturally suspended or under slight stress. Fabric drape effect information refers to the set of drape shape features of the fabric under the action of gravity.

[0213] By inputting the estimated fabric into a preset fabric database, information on the fabric's natural wrinkles and drape can be obtained for subsequent use.

[0214] The fabric database pre-stores a table showing the correspondence between different estimated fabrics and their corresponding information on natural fabric wrinkles and drape. The fabric database is accessed after the operator has pre-entered the information.

[0215] S62: Generate adjustment points on the initial 3D model of the garment based on the information of natural fabric wrinkles and fabric drape.

[0216] Among them, adjustment points refer to the set of key three-dimensional coordinate points on the surface of the initial three-dimensional model of the garment, which are used to precisely control the shape of the folds and the draping trend.

[0217] By selecting key areas such as the neckline, cuffs, hem, and waistline from the initial 3D model of the garment, and using preset adjustment reference intervals to select positions as initial adjustment points, the number of initial adjustment points corresponding to each area is determined and used as the initial value. Then, the fabric's natural wrinkle information, fabric drape information, and key areas are input into a preset initial reference database to match and obtain the initial reference value. When the initial value is less than the initial reference value, the initial adjustment point is used as the adjustment point. When the initial value is not less than the initial reference value, adjustment points are selected based on the initial reference value according to the principle of increasing the spacing and used as adjustment points for convenient subsequent use.

[0218] The adjustment reference interval distance refers to the reference distance between each adjustment point during adjustment. The adjustment reference interval distance is preset by the operator according to requirements. The initial reference database pre-stores information on different fabric natural wrinkles, fabric drape effects, and a table mapping key parts to their corresponding initial reference values. The initial reference database is preset by the operator according to actual needs.

[0219] S63: Determine the posture of the garment based on the garment pattern.

[0220] Among them, the posture of clothing refers to the spatial shape of clothing presented in the sample picture, the relative position of key parts, and the angle of placement.

[0221] By analyzing the garment pattern, the garment's posture can be determined, facilitating its subsequent use.

[0222] To further ensure the rationality of the garment's posture, it is necessary to conduct a more detailed separate analysis and calculation of the garment's posture, which will be explained in detail through the steps shown below.

[0223] The method for determining clothing posture includes the following steps:

[0224] S631: Identify joint locations and adjustable points in the garment based on the garment pattern scan.

[0225] Among them, the joint position points of clothing refer to the key coordinate points in the clothing pattern that precisely correspond to the core joints of the human body. Clothing joint position points include the shoulder joint point (corresponding to the human shoulder joint), the elbow joint point (corresponding to the human elbow joint), and the waist joint point (corresponding to the human lumbar spine joint), etc.

[0226] Adjustable points in clothing refer to the coordinate points corresponding to design elements on clothing that have size / shape adjustment functions.

[0227] By scanning and identifying the joint positions from the garment pattern and using them as the garment joint positions, and by matching the garment pattern with preset local adjustment features, local adjustable points of the garment can be obtained for convenient subsequent use.

[0228] Local adjustment features refer to features corresponding to size / shape adjustment functions. Local adjustment features include features such as waist drawstring, collar buttons, and cuff Velcro. Local adjustment features are obtained through pre-input by the operator.

[0229] S632: Determine joint gravity parameters by combining the joint location points of the garment with the estimated fabric.

[0230] Among them, joint gravity parameters refer to the core parameters that quantify the influence of gravity on the shape of clothing joints. Joint gravity parameters include gravity influence coefficient (reflecting the intensity of gravity on the fabric at the joint), deformation threshold (the maximum allowable deformation of the fabric at the joint due to gravity), and recovery coefficient (the ability of the fabric to return to its initial shape after the action of gravity).

[0231] By inputting the joint position points of the garment and the estimated fabric into a preset joint gravity database, joint gravity parameters are obtained for easy subsequent use.

[0232] The joint gravity database pre-stores a table of different garment joint locations, estimated fabrics, and corresponding joint gravity parameters. The joint gravity database is obtained by the operator in advance by detecting parameters such as fabric weight, fiber stiffness, structural density, and structural basis at each garment joint location of different estimated fabrics and then calculating the corresponding joint gravity parameters.

[0233] S633: Determine the locally adjustable points of the model based on the initial 3D model of the garment.

[0234] Among them, the locally adjustable points of the model refer to the adjustable positions in the initial three-dimensional model of the clothing.

[0235] By retrieving the neckline, cuffs, shoulder fit, waist fit, and hip fit from the initial 3D model of the garment and using them as adjustable points in the model, it becomes easier to use them later.

[0236] S634: Determine whether the adjustable points of the clothing are consistent with the adjustable points of the model. If not, proceed to S635; if yes, proceed to S636.

[0237] In this process, the system determines whether the posture can be directly determined based on the joint gravity parameters by judging whether the adjustable points of the clothing are consistent with the adjustable points of the model.

[0238] S635: Determine the natural posture of the garment based on the joint gravity parameters, and use the natural posture of the garment as the garment posture.

[0239] Among them, the natural posture of clothing refers to the posture of clothing when it is only affected by gravity.

[0240] When the adjustable points of the garment are inconsistent with the adjustable points of the model, it means that the posture can be determined directly based on the joint gravity parameters. Therefore, by inputting the joint gravity parameters into the preset gravity posture model to output the natural posture of the garment, and taking the natural posture of the garment as the garment posture, the accuracy of the obtained garment posture can be improved.

[0241] The gravity posture model is obtained by deep learning training, which takes the gravity parameters of each joint as input and outputs the natural posture of the clothing.

[0242] S636: Determine the posture adjustment parameters based on the locally adjustable points of the model and the locally adjustable points of the clothing.

[0243] Among them, posture adjustment parameters refer to the core parameters for adjusting the fit of clothing.

[0244] When the adjustable points of the clothing and the adjustable points of the model are the same, it means that the posture cannot be determined directly based on the joint gravity parameters. Therefore, the image posture parameters are retrieved through the adjustable points of the clothing, and the model posture parameters are retrieved through the adjustable points of the model. The deviation between the image posture parameters and the model posture parameters is then used as the posture adjustment parameters.

[0245] Image pose parameters refer to parameters such as the spatial shape of clothing in the image, the relative positions of key parts, and the placement angle. Model pose parameters refer to parameters such as the spatial shape of clothing in the model, the relative positions of key parts, and the placement angle.

[0246] S637: Determine the posture of the local adjustment point based on the posture adjustment parameters and joint gravity parameters, and use the posture of the local adjustment point as the posture of the clothing.

[0247] Among them, the pose of the local adjustment point refers to the pose corresponding to the model after local adjustments are made.

[0248] By inputting posture adjustment parameters and joint gravity parameters into a preset local posture model to output the posture of the local adjustment point, and using the posture of the local adjustment point as the posture of the clothing, the accuracy of the obtained clothing posture is improved.

[0249] The local pose model is obtained by deep learning training after inputting various pose adjustment parameters and joint gravity parameters and outputting the pose of local adjustment points.

[0250] S64: Adjust the initial 3D model of the garment according to the garment posture and adjustment points to obtain the garment posture model.

[0251] Among them, the clothing posture model refers to the clothing model after posture adjustment.

[0252] By adjusting the corresponding adjustment points on the initial 3D model of the garment according to the spatial shape, relative position of key parts, and placement angle of the garment posture, a garment posture model is obtained, which is convenient for subsequent use.

[0253] S65: The clothing pose model is fused based on the input color to obtain the clothing color model, and the clothing color model is used as the final 3D model of the clothing.

[0254] Among them, the clothing coloring model refers to the clothing model after coloring.

[0255] By blending and coloring the input colors onto the clothing pose model, a colored clothing model is obtained, which is then used as the final 3D model of the clothing, thus improving the accuracy of the final 3D model of the clothing.

[0256] To further ensure the rationality of the final 3D model of the garment, it is necessary to perform further separate analysis and calculations on the final 3D model of the garment, which will be explained in detail through the following steps.

[0257] After outputting the final 3D model of the clothing, the following steps are also included:

[0258] S71: Collect user-uploaded images of people wearing tight clothing as reference images for users.

[0259] Among them, user reference images refer to pictures uploaded by users wearing tight clothing.

[0260] User reference images are obtained by obtaining photos of users wearing tight clothing after authorization and uploading them.

[0261] S72: Perform feature recognition based on the user's reference image to obtain a user feature image.

[0262] User feature images refer to a collection of images representing the characteristics of a user's body shape. These characteristics include the circumferences of key body parts (shoulder width, waist circumference, hip circumference, upper arm width), body contour curves, and the relative positions of key body parts (shoulder line, waist line, hip line).

[0263] By performing feature recognition on key areas such as the shoulders, waist, and upper arm width of user reference images, these images are used as user feature images for convenient subsequent use.

[0264] S73: Determine specific body data based on user feature images.

[0265] Among them, specific body data refers to a set of quantitative physical parameters that reflect the user's actual body shape.

[0266] The user's body shape is determined by identifying the corresponding body parts from the user's feature image. Then, the coordinates of the corresponding features in the image are extracted, and the corresponding data are calculated to obtain the specific body data corresponding to each location.

[0267] For example, when the user's feature image is the shoulder, the shoulder width is calculated by extracting the coordinates corresponding to the two endpoints of the outermost edge of the shoulder, calculating the distance between the two coordinates, and then comparing it with a preset image scale. The shoulder line is calculated by extracting the coordinates corresponding to each endpoint of the shoulder.

[0268] S74: Identify and record body characteristics based on specific body data.

[0269] Among them, body characteristics refer to a set of descriptions that reflect the core attributes of a user's body shape and personalized adaptation needs.

[0270] This process involves retrieving parameters such as upper body height, shoulder width, hip circumference, waist circumference, and upper arm width from specific body data. The ratio between shoulder width and hip circumference is calculated as the shoulder-hip ratio, and the ratio between waist circumference and hip circumference is calculated as the waist-hip ratio. Based on these ratios, body type is determined, including A-type, H-type, O-type, and X-type. The ratio between upper body height and shoulder width is calculated as the body-shoulder ratio, which is then used to determine the height proportion, including slender, standard, or robust types. Finally, the body type is combined with the height proportion and upper arm width to form a comprehensive body characteristic for subsequent use.

[0271] S75: Adjust the final 3D model of the clothing based on body characteristics.

[0272] In this process, the accuracy of the final 3D model of the clothing is improved by adjusting the positions of the shoulders, length, waist, and upper arm width of the garment based on body features.

[0273] To further ensure the rationality of adjusting the final 3D model of clothing based on body characteristics, it is necessary to conduct further separate analysis and calculation on the adjustment of the final 3D model of clothing based on body characteristics, which will be explained in detail through the following steps.

[0274] Adjusting the final 3D model of clothing based on body characteristics includes the following steps:

[0275] S751: Determine model features based on the final 3D model of the clothing.

[0276] Among them, model features refer to the set of quantitative physical parameters of the body shape corresponding to the final three-dimensional model of the clothing.

[0277] By retrieving parameters such as upper body height, shoulder width, hip circumference, waist circumference, and upper arm width from the final 3D model of the garment, the body type and body height-to-body ratio corresponding to the final 3D model of the garment can be calculated. Then, the upper body height, upper arm width, and other parameters of the final 3D model of the garment are combined as model features for convenient subsequent use.

[0278] S752: Based on the deviation between model features and body features, obtain the feature deviation value.

[0279] Among them, the feature deviation value refers to the degree of deviation between the model features and the body features.

[0280] By comparing the model features with the body features using the same type of parameters, when inconsistencies exist, the ratio between the two parameters of the same type is calculated to obtain the feature deviation value, which is convenient for subsequent use.

[0281] For example, when the upper arm width corresponding to the model feature is 10 cm and the upper arm width corresponding to the body feature is 15 cm, the feature bias value is approximately 0.67. When the upper body height corresponding to the model feature is 80 cm and the upper body height corresponding to the body feature is 100 cm, the feature bias value is approximately 0.8.

[0282] S753: Determine the feature position and feature adjustment value based on the feature deviation value.

[0283] Here, the feature location refers to the body part corresponding to the feature deviation value. The feature adjustment value refers to the adjustment ratio required when adjusting the feature location based on the feature deviation value.

[0284] By retrieving the location of the feature deviation value and using it as the feature position, and by calculating the quotient between 1 and the feature deviation value, the feature adjustment value is obtained, which is convenient for subsequent use.

[0285] For example, if the feature deviation value corresponds to the upper arm width and is specifically 0.67, the feature location is the upper arm, and the feature adjustment value is 1.5. If the feature deviation value corresponds to the upper body height and is specifically 0.8, the feature location is the upper body, and the feature adjustment value is 1.25.

[0286] S754: Adjust the final 3D model of the garment based on feature location and feature adjustment value.

[0287] The accuracy of the final 3D model of the clothing is improved by adjusting the feature positions corresponding to the final 3D model of the clothing according to the feature adjustment value.

[0288] Based on the same inventive concept, embodiments of the present invention provide a rapid generation system for 3D clothing models based on sketch recognition, comprising:

[0289] The data acquisition module is used to collect clothing sketches, requirement input information, and user reference images;

[0290] The memory stores a program for implementing a method for rapidly generating 3D clothing models based on sketch recognition, as described above.

[0291] The processor loads and executes programs stored in memory.

[0292] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional modules is used as an example. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. The specific working process of the system, device, and unit described above can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0293] The above description is merely a preferred embodiment of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principles of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. A method for rapidly generating 3D clothing models based on sketch recognition, characterized in that, include: S1: Collect clothing sketches and requirement input information uploaded by users; S2: Generate a garment image based on the garment sketch; S3: Extract clothing features from clothing images to generate an initial 3D model of the clothing; S4: Retrieve input type and requirement content based on requirement input information; S5: Extract the required content based on the input type to obtain the input fabric and input color; S6: Combining the input fabric and input color, optimize the initial 3D model of the garment to generate the final 3D model of the garment, and output the final 3D model of the garment. Methods for generating clothing images include: S21: When the clothing sketch is an image, preprocess the clothing sketch to obtain a black and white line image; S22: Obtain the initial image by removing isolated points from the black and white line image; S23: Extract line contours based on the initial image; S24: Filter valid lines based on line outlines to obtain a clear image; S25: Smooth the image based on the clear image to obtain the clothing image; Methods for selecting valid lines include: S241: Determine the line thickness characteristics, line texture characteristics, and line intersections based on the line outline; S242: Determine the sketch style by combining the characteristics of line thickness and line texture; S243: Determine the line classification attributes based on the sketch style; S244: Classify line outlines according to line classification attributes to obtain line categories; S245: Determine the number of intersections and their positions based on the intersection points of the lines; S246: Select line contours by combining line type, number of intersections, and intersection position to obtain selected lines as valid lines.

2. The method for rapid generation of clothing 3D models based on sketch recognition according to claim 1, characterized in that, The methods for determining the input fabric and color include: S51: Determine whether the input type is the preset text type; S52: If yes, semantic recognition is performed based on the requirements to obtain semantic fabric and semantic color, and semantic fabric is used as input fabric and semantic color is used as input color. S53: If not, then identify the shadow area and the bright area based on the content requirements and adjust them to obtain a unified image; S54: Based on the unified image, take color points at different locations to obtain the input color; S55: Identify texture features based on the required sample image to obtain the sample image texture; S56: Extract effective reflective features from the required sample image to obtain the fabric reflective intensity and fabric reflective texture; S57: Determine the material properties of the fabric based on its reflectivity and reflective texture; S58: Combine the sample image texture, fabric material properties, and input color to determine the estimated fabric, and use the estimated fabric as the input fabric.

3. The method for rapid generation of clothing 3D models based on sketch recognition according to claim 2, characterized in that, Methods for determining the fabric type include: S581: Determine the fabric density based on the texture of the sample image; S582: Determine the basic fabric structure by combining fabric density and fabric material properties; S583: Retrieves color saturation, color uniformity, and color transmittance based on the input color; S584: Determine the color prediction fabric by combining color saturation, color uniformity and color transmittance; S585: Select the predicted fabric from the color prediction fabrics based on the fabric structure.

4. The method for rapid generation of clothing 3D models based on sketch recognition according to claim 1, characterized in that, Methods for generating the final 3D model of clothing include: S61: Determine the natural wrinkle information and drape effect information of the fabric based on the estimated fabric; S62: Generate adjustment points on the initial 3D model of the garment based on the information of natural fabric wrinkles and fabric drape. S63: Determine the garment's posture based on the garment pattern; S64: Adjust the initial 3D model of the garment according to the garment posture and adjustment points to obtain the garment posture model; S65: The clothing pose model is fused based on the input color to obtain the clothing color model, and the clothing color model is used as the final 3D model of the clothing.

5. The method for rapid generation of clothing 3D models based on sketch recognition according to claim 4, characterized in that, Methods for determining clothing posture include: S631: Identify joint locations and adjustable points in the garment based on the garment pattern scan; S632: Determine joint gravity parameters by combining the joint location points of the garment with the estimated fabric; S633: Determine the locally adjustable points of the model based on the initial 3D model of the garment; S634: Determine whether the adjustable points of the clothing are consistent with the adjustable points of the model; S635: If not, determine the natural posture of the garment based on the joint gravity parameters and use the natural posture of the garment as the garment posture. S636: If yes, then determine the posture adjustment parameters based on the local adjustable points of the model and the local adjustable points of the clothing; S637: Determine the posture of the local adjustment point based on the posture adjustment parameters and joint gravity parameters, and use the posture of the local adjustment point as the posture of the clothing.

6. The method for rapidly generating a 3D clothing model based on sketch recognition according to claim 1, characterized in that, After outputting the final 3D model of the clothing, the following is also included: S71: Collect user-uploaded images of people wearing tight clothing as reference images for users; S72: Perform feature recognition based on the user reference image to obtain a user feature image; S73: Determine specific body data based on user profile images; S74: Identify and record body characteristics based on specific body data; S75: Adjust the final 3D model of the clothing based on body characteristics.

7. The method for rapid generation of clothing 3D models based on sketch recognition according to claim 6, characterized in that, Adjusting the final 3D model of clothing based on body characteristics includes: S751: Determine model features based on the final 3D model of the garment; S752: Based on the deviation between model features and body features, the feature deviation value is obtained; S753: Determine the feature position and feature adjustment value based on the feature deviation value; S754: Adjust the final 3D model of the garment based on feature location and feature adjustment value.

8. A rapid generation system for 3D clothing models based on sketch recognition, characterized in that, include: The data acquisition module is used to collect clothing sketches, requirement input information, and user reference images; The memory stores a program for implementing a method for rapidly generating a 3D clothing model based on sketch recognition as described in any one of claims 1 to 7; The processor loads and executes programs stored in memory.

Images