A method and system for acquiring three-dimensional human body data based on big data and artificial intelligence

Abstract

The present invention proposes a human body three-dimensional data acquisition system based on big data and artificial intelligence, which includes a face value evaluation subsystem and a clothing setting subsystem; at the same time, the system and its method extract the user's face, body information and clothing information to obtain multiple parameters corresponding to the face value and the human body clothing, establish a human body 3D model based on the above parameters, select the corresponding clothing model, and accurately paste the 3D clothing image on the corresponding part of the 3D human body model to show the 3D effect when the user puts on the clothes. Using the above system and method, users can easily try on clothes and customize clothes online, which improves the user's experience and saves the user's time and energy; at the same time, it reduces the time and clothing resources of clothing stores occupied by customers during the fitting process, and alleviates the congestion of clothing stores.

Description

Technical Field

[0001] The present invention belongs to the field of intelligent image recognition, and in particular relates to a method and system for acquiring three-dimensional human body data based on big data and artificial intelligence. Background Art

[0002] The existing fitting room technology is not mature, and the system is not perfect. Most fitting systems still use 2D systems, which cover the one-sided pictures of clothing on the photos of models to complete fitting. This can only show the 2D plane effect of clothing, and cannot achieve the ideal fitting and purchasing effect. In addition, some fitting systems with 3D functions are not very perfect, and cannot generate 3D clothing objects through the combination of various clothing material pictures. Users can only try on the clothes provided by the fitting system, but cannot try on all the clothes that users like, which makes the user's selection range quite narrow and cannot truly experience the superiority of the fitting system. At the same time, fitting systems often use some inherent models. Although some fitting systems support real-life photo transmission, it is difficult to show the effect of real-life fitting because the expressions, angles, and colors of the photos are relatively simple.

[0003] In addition, in the traditional clothing purchase process, customers need to take off their clothes to try on new clothes, which takes up a lot of their time. Due to time and clothing resource constraints, sometimes they cannot buy a piece of clothing that suits them after several hours. The customer's fitting process also takes up a lot of time and clothing resources of clothing stores, causing clothing stores to lose a lot of profits. Especially during the golden holidays, when there are many users, clothing stores are often crowded and work efficiency is low. At the same time, due to spatial location constraints, clothing stores cannot provide a variety of clothing for users to refer to, which also limits the development of clothing stores. Summary of the invention

[0004] In order to solve the above technical problems, the present invention proposes a human body three-dimensional data acquisition system based on big data and artificial intelligence, which includes an appearance evaluation subsystem and a clothing formulation subsystem;

[0005] The appearance evaluation subsystem includes:

[0006] A face image acquisition module is used to capture a frontal face image of a user;

[0007] A facial feature extraction module is used to recognize the facial contour based on the frontal facial image and extract specific facial features;

[0008] A face parameter analysis module, used to determine a first predicted face value and a corresponding confidence level in a face image through a classification model;

[0009] The face parameter analysis module is further used to determine the second predicted appearance value and the corresponding weight of the face in the face image through a regression model;

[0010] an appearance calculation module, configured to determine the appearance of the face based on the first predicted appearance, the confidence level, the second predicted appearance, and the weight;

[0011] The clothing setting subsystem includes:

[0012] A human body information collection module is used to scan and photograph the user's body to obtain the user's body image;

[0013] A human feature extraction module, used to extract human contour information, human body information and clothing information based on the body image acquired by the human information acquisition module;

[0014] A human body model building module, used to generate a 3D human body model according to the specific parameter values ​​corresponding to the human body information of the user and the user's facial value, and store the 3D human body model in the user model database;

[0015] A clothing model extraction module is used to extract clothing models corresponding to clothing information;

[0016] A 3D clothing image generation module, used to generate a 3D clothing image corresponding to the clothing model according to the clothing model obtained by the clothing model extraction module and the style, category, color, fabric, auxiliary material, and accessories of the user's clothing information;

[0017] The space positioning module is used to position the 3D clothing image generated by the 3D clothing image generating module in three-dimensional space based on the 3D human body model generated by the human body model establishing module.

[0018] Optionally, the appearance evaluation subsystem further includes:

[0019] A training module, used to input a set number of image samples containing human faces and the standard appearance of the human faces contained in each image sample into an untrained convolutional neural network, and train the convolutional layer and the fully connected layer of the untrained convolutional neural network;

[0020] A control module, configured to stop training the convolutional neural network and obtain the regression model when it is determined that the weight parameters of the connections between the nodes in the convolutional layer and the fully connected layer meet the first preset condition;

[0021] The control module is also used to stop training the convolutional neural network and obtain the classification model when it is determined that the weight parameters of the connections between the nodes in the convolutional layer and the fully connected layer meet the second preset condition.

[0022] Optionally, the appearance calculation module is specifically used to determine the user appearance based on the first predicted appearance, the confidence, the second predicted appearance, the weight and the standard appearance and the following relationship:

[0023] B=(P1*W+P2*K) / (w1+w2+Bp)

[0024] Wherein, B is the user's appearance, P1 is the first predicted appearance, W is the confidence, P2 is the second predicted appearance, K is the weight, and Bp is the appearance calibration value. The appearance calibration value is pre-set according to the user's gender or clothing type.

[0025] Optionally, a human body model establishment module is used to determine whether the human body information feature values ​​of the user have been stored in the user model database; when the judgment is yes, obtain the parameter values ​​of the height range, face shape range, body shape range, and gender of the specific target person corresponding to the user's human body information from the user model database; when the judgment is no, generate specific parameter values ​​corresponding to the specific target person from the user's human body information according to the SVM algorithm, and store the specific parameter values ​​in the user model database.

[0026] Optionally, the human body information includes the target person's height range, face shape range, body shape range, and gender; the clothing information includes clothing style, category, color, fabric, auxiliary materials, and accessories.

[0027] The present invention also provides a human body three-dimensional data acquisition method of a human body three-dimensional data acquisition system based on big data and artificial intelligence, which specifically includes an appearance evaluation method and a clothing formulation method.

[0028] Optionally, the appearance evaluation method includes:

[0029] (1) Take a frontal facial image of the user;

[0030] (2) Recognize the facial contour based on the frontal face image and extract specific facial features, such as eyebrow size, eye size, nose and mouth shape;

[0031] (3) Determine the first predicted appearance and corresponding confidence level of the face in the face image through the classification model;

[0032] Determine a second predicted appearance value and a corresponding weight of the face in the face image by using a regression model;

[0033] (4) Determining the appearance of the face based on the first predicted appearance, the confidence level, the second predicted appearance, and the weight.

[0034] Optionally, the clothing formulation method specifically includes:

[0035] (1) Scanning and photographing the user's body to obtain the user's body image;

[0036] (2) extracting human body contour information, human body information and clothing information based on the body image acquired by the human body information acquisition module;

[0037] (3) determining whether the user's human body information feature values ​​are stored in a user model database; if the determination is yes, obtaining the parameter values ​​of the height range, face shape range, body shape range, and gender of the specific target person corresponding to the user's human body information from the user model database; if the determination is no, generating specific parameter values ​​corresponding to the specific target person from the user's human body information according to an SVM algorithm, and storing the specific parameter values ​​in the user model database;

[0038] (4) generating a 3D human body model according to the specific parameter values ​​corresponding to the human body information of the user and the facial value of the user, and storing the 3D human body model in the user model database;

[0039] (5) determining whether the clothing information of the user is stored in the clothing model database; if so, extracting the clothing model corresponding to the clothing information; if not, finding parameters that are identical or similar to the above parameters based on three or more parameters in the specific clothing information; then selecting the clothing model corresponding to the clothing information with the most identical or similar parameters;

[0040] (6) generating a 3D clothing image corresponding to the clothing model according to the clothing model obtained by the clothing model extraction module and the style, category, color, fabric, auxiliary materials, and accessories of the user's clothing information;

[0041] (7) Based on the 3D human body model generated by the human body model establishment module, the 3D clothing image generated by the 3D clothing image generation module is positioned in three-dimensional space so that the 3D clothing image is accurately attached to the corresponding part of the 3D human body model, showing the 3D effect when the user puts on the clothes.

[0042] The technical solution of the present invention enables users to easily try on and customize clothes online, thereby improving the user experience and saving the user's time and energy; at the same time, it reduces the large amount of time and clothing resources of clothing stores occupied by customers in the process of trying on clothes, thereby alleviating congestion in clothing stores.

[0043] Further advantages of the present invention will be further reflected in detail in the specific embodiments section in conjunction with the drawings of the specification. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0045] Figure 1 It is a schematic diagram of a human body three-dimensional data acquisition system;

[0046] Figure 2 It is a flow chart of the appearance evaluation method in the method of acquiring three-dimensional data of the human body;

[0047] Figure 3 It is a flow chart of a clothing formulation method in a method for acquiring three-dimensional data of a human body. DETAILED DESCRIPTION

[0048] The invention is further described below in conjunction with the accompanying drawings and specific implementation methods.

[0049] like Figure 1 As shown, a human body three-dimensional data acquisition system based on big data and artificial intelligence is provided, which includes an appearance evaluation subsystem and a clothing formulation subsystem.

[0050] The appearance evaluation subsystem includes:

[0051] The face image acquisition module is used to capture the user's front face image.

[0052] The facial feature extraction module is used to recognize the facial contour based on the frontal facial image and extract specific facial features, such as eyebrow size, eye size, nose and mouth shape, etc.

[0053] A face parameter analysis module, used to determine a first predicted beauty value and a corresponding confidence level of a face in a face image through a classification model, wherein the first predicted beauty value is a discrete value;

[0054] The face parameter analysis module is further used to determine a second predicted beauty value and a corresponding weight of the face in the face image through a regression model, wherein the second predicted beauty value is a continuous value;

[0055] The appearance calculation module is used to determine the appearance of the face based on the first predicted appearance, the confidence level, the second predicted appearance and the weight.

[0056] The face parameter analysis module is specifically used to input the face contour and facial specific feature parameters obtained in the face feature extraction module into the classification model; the classification model is obtained by training the first convolutional neural network;

[0057] Subsequently, the facial parameter analysis module determines a first predicted appearance value and a confidence level of the face based on the classification model;

[0058] In addition, the face parameter analysis module inputs the face contour and facial specific feature parameters obtained in the face feature extraction module into the regression model; the regression model is obtained by training a second convolutional neural network;

[0059] Subsequently, the facial parameter analysis module determines a second predicted appearance value and weight of the face based on the regression model.

[0060] The appearance evaluation subsystem also includes:

[0061] A training module, used to input a set number of image samples containing human faces and the standard appearance of the human faces contained in each image sample into an untrained convolutional neural network, and train the convolutional layer and the fully connected layer of the untrained convolutional neural network;

[0062] The control module is used to stop training the convolutional neural network and obtain the regression model when it is determined that the weight parameters of the connections between the nodes in the convolutional layer and the fully connected layer meet the first preset condition.

[0063] The control module is also used to stop training the convolutional neural network and obtain the classification model when it is determined that the weight parameters of the connections between the nodes in the convolutional layer and the fully connected layer meet the second preset condition.

[0064] The appearance calculation module is specifically used to determine the user appearance based on the first predicted appearance, the confidence, the second predicted appearance, the weight and the standard appearance and the following relationship:

[0065] B=(P1*W+P2*K) / (w1+w2+Bp)

[0066] Wherein, B is the user's appearance, P1 is the first predicted appearance, W is the confidence, P2 is the second predicted appearance, K is the weight, and Bp is the appearance calibration value. The appearance calibration value is pre-set according to the user's gender or clothing type.

[0067] The clothing setting subsystem includes:

[0068] The human body information acquisition module is used to scan and photograph the user's body to obtain the user's body image.

[0069] The human feature extraction module is used to extract human contour information, human information and clothing information based on the body image acquired by the human information acquisition module. The human information includes the height range, face shape range, body shape range, and gender of the target person; the clothing information includes clothing style, category, color, fabric, auxiliary materials, and accessories.

[0070] The human body model building module is used to determine whether the human body information feature value of the user has been stored in the user model database. If it is determined to be yes, the parameter values ​​of the height range, face shape range, body shape range, and gender of the specific target person corresponding to the human body information of the user are obtained from the user model database; if it is determined to be no, the specific parameter values ​​corresponding to the specific target person are generated from the human body information of the user according to the SVM algorithm, and the specific parameter values ​​are stored in the user model database.

[0071] Subsequently, the human body model building module generates a 3D human body model according to the specific parameter values ​​corresponding to the user's human body information and the user's facial appearance, and stores the 3D human body model in the user model database.

[0072] The clothing model extraction module is used to determine whether the clothing information of the user is stored in the clothing model database. If yes, the clothing model corresponding to the clothing information is extracted. If not, based on three or more parameters in the specific clothing information (clothing style, category, color, fabric, auxiliary material, accessories), the same or similar parameters as the above parameters are found, such as the same style, similar color, similar fabric material, etc. Subsequently, the clothing model extraction module selects the clothing model corresponding to the clothing information with the most identical or similar parameters. The clothing model is usually a 2D clothing image.

[0073] The 3D clothing image generation module is used to generate a 3D clothing image corresponding to the clothing model based on the clothing model obtained by the clothing model extraction module and multiple parameters such as style, category, color, fabric, auxiliary materials and accessories in the user's clothing information.

[0074] The spatial positioning module is used to position the 3D clothing image generated by the 3D clothing image generation module in three-dimensional space based on the 3D human body model generated by the human body model establishment module, so that the 3D clothing image is accurately attached to the corresponding part of the 3D human body model, showing the 3D effect when the user puts on the clothes.

[0075] The present invention also provides a method for acquiring three-dimensional human body data based on big data and artificial intelligence, specifically including an appearance evaluation method and a clothing formulation method.

[0076] like Figure 2 As shown, the appearance evaluation method specifically includes:

[0077] (1) Take a frontal facial image of the user.

[0078] (2) Recognize the facial contour based on the frontal face image and extract specific facial features, such as eyebrow size, eye size, nose and mouth shape, etc.

[0079] (3) determining a first predicted beauty value and a corresponding confidence level of a face in the face image by using a classification model, wherein the first predicted beauty value is a discrete value;

[0080] The second predicted appearance value and the corresponding weight of the face in the face image are determined by a regression model, wherein the second predicted appearance value is a continuous value.

[0081] (4) Determining the appearance of the face based on the first predicted appearance, the confidence level, the second predicted appearance, and the weight.

[0082] Step (3) specifically comprises: inputting the facial contour and facial specific feature parameters obtained in the facial feature extraction module into the classification model; the classification model is obtained by training a first convolutional neural network;

[0083] Based on the classification model, a first predicted appearance and confidence level of the face is determined.

[0084] In addition, the facial contour and facial specific feature parameters obtained in the facial feature extraction module are input into the regression model; the regression model is obtained by training a second convolutional neural network.

[0085] Subsequently, based on the regression model, a second predicted appearance value and weight of the face are determined.

[0086] Step (3) specifically includes: inputting a set number of image samples containing human faces and the standard appearance of the human faces contained in each image sample into an untrained convolutional neural network, and training the convolutional layer and the fully connected layer of the untrained convolutional neural network;

[0087] Subsequently, when it is determined that the weight parameters of the connections between the nodes in the convolutional layer and the fully connected layer meet the first preset condition, the training of the convolutional neural network is stopped to obtain the regression model;

[0088] At the same time, when it is determined that the weight parameters of the connections between the nodes in the convolutional layer and the fully connected layer meet the second preset condition, the training of the convolutional neural network is stopped to obtain the classification model.

[0089] Step (3) specifically includes: determining the user appearance based on the first predicted appearance, the confidence level, the second predicted appearance, the weight and the standard appearance and the following relationship:

[0090] B=(P1*W+P2*K) / (w1+w2+Bp)

[0091] Wherein, B is the user's appearance, P1 is the first predicted appearance, W is the confidence, P2 is the second predicted appearance, K is the weight, and Bp is the appearance calibration value. The appearance calibration value is pre-set according to the user's gender or clothing type.

[0092] like Figure 3 As shown, the clothing making method specifically includes:

[0093] (1) Scan and photograph the user's body to obtain the user's body image.

[0094] (2) Extracting human body contour information, human body information and clothing information based on the body image acquired by the human body information acquisition module.

[0095] The human body information includes the target person's height range, face shape range, body shape range, and gender; the clothing information includes clothing style, category, color, fabric, auxiliary materials, and accessories.

[0096] (3) Determine whether the user's human body information feature values ​​are stored in the user model database. If the judgment is yes, obtain the parameter values ​​of the height range, face shape range, body shape range, and gender of the specific target person corresponding to the user's human body information from the user model database; if the judgment is no, generate the specific parameter values ​​corresponding to the specific target person from the user's human body information according to the SVM algorithm, and store the specific parameter values ​​in the user model database.

[0097] (4) Generate a 3D human body model according to the specific parameter values ​​corresponding to the user's human body information and the user's facial appearance, and store the 3D human body model in the user model database.

[0098] (5) Determine whether the clothing information of the user is stored in the clothing model database. If yes, extract the clothing model corresponding to the clothing information. If not, find the same or similar parameters as the above parameters based on three or more parameters in the specific clothing information (clothing style, category, color, fabric, auxiliary material, accessories), such as the same style, similar color, similar fabric material, etc. Then, select the clothing model corresponding to the clothing information with the most same or similar parameters. The clothing model is usually a 2D clothing image.

[0099] (6) Generate a 3D clothing image corresponding to the clothing model based on the clothing model obtained by the clothing model extraction module and multiple parameters such as style, category, color, fabric, auxiliary materials, and accessories in the user's clothing information.

[0100] (7) Based on the 3D human body model generated by the human body model establishment module, the 3D clothing image generated by the 3D clothing image generation module is positioned in three-dimensional space so that the 3D clothing image is accurately attached to the corresponding part of the 3D human body model, showing the 3D effect when the user puts on the clothes.

[0101] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A human body three-dimensional data acquisition system based on big data and artificial intelligence, the system includes a face value evaluation subsystem and a clothing setting subsystem; in, The appearance evaluation subsystem includes: A face image acquisition module is used to capture a frontal face image of a user; A facial feature extraction module is used to recognize the facial contour based on the frontal facial image and extract specific facial features; A face parameter analysis module, used to determine a first predicted face value and a corresponding confidence level in a face image through a classification model; The face parameter analysis module is further used to determine the second predicted appearance value and the corresponding weight of the face in the face image through a regression model; an appearance calculation module, configured to determine the appearance of the face based on the first predicted appearance, the confidence level, the second predicted appearance, and the weight; The clothing setting subsystem includes: A human body information collection module is used to scan and photograph the user's body to obtain the user's body image; A human feature extraction module, used to extract human contour information, human body information and clothing information based on the body image acquired by the human information acquisition module; A human body model building module, used to generate a 3D human body model according to the specific parameter values ​​corresponding to the human body information of the user and the user's facial value, and store the 3D human body model in the user model database; A clothing model extraction module is used to extract clothing models corresponding to clothing information; A 3D clothing image generation module, used to generate a 3D clothing image corresponding to the clothing model according to the clothing model obtained by the clothing model extraction module and the style, category, color, fabric, auxiliary materials, accessories and multiple parameters in the user's clothing information; A spatial positioning module, used to position the 3D clothing image generated by the 3D clothing image generation module in three-dimensional space based on the 3D human body model generated by the human body model establishment module; The human body model establishment module is used to determine whether the human body information feature values ​​of the user have been stored in the user model database; when the judgment is yes, the parameter values ​​of the height range, face shape range, body shape range, and gender of the specific target person corresponding to the user's human body information are obtained from the user model database; when the judgment is no, the user's human body information is generated according to the SVM algorithm. Specific parameter values ​​corresponding to the specific target person, and the specific parameter values ​​are stored in the user model database.

2. A human body three-dimensional data acquisition system based on big data and artificial intelligence as claimed in claim 1, characterized in that: The appearance evaluation subsystem also includes: A training module, used to input a set number of image samples containing human faces and the standard appearance of the human faces contained in each image sample into an untrained convolutional neural network, and train the convolutional layer and the fully connected layer of the untrained convolutional neural network; A control module, configured to stop training the convolutional neural network and obtain the regression model when it is determined that the weight parameters of the connections between the nodes in the convolutional layer and the fully connected layer meet the first preset condition; The control module is also used to stop training the convolutional neural network and obtain the classification model when it is determined that the weight parameters of the connections between the nodes in the convolutional layer and the fully connected layer meet the second preset condition.

3. The human body three-dimensional data acquisition system based on big data and artificial intelligence as claimed in claim 1, characterized in that: The appearance calculation module is specifically used to determine the user appearance based on the first predicted appearance, the confidence, the second predicted appearance, the weight and the standard appearance and the following relationship: B=(P1*W+P2*K) / (w1+w2+Bp) Among them, B is the user's appearance, P1 is the first predicted appearance, W is the confidence, P2 is the second predicted appearance, K is the weight, and Bp is the appearance calibration value; wherein the appearance calibration value is pre-set according to the user's gender or clothing type.

4. The human body three-dimensional data acquisition system based on big data and artificial intelligence as claimed in claim 1, characterized in that: The human body information includes the target person's height range, face shape range, body shape range, and gender; the clothing information includes clothing style, category, color, fabric, auxiliary materials, and accessories.

5. A method for acquiring three-dimensional human body data based on a human body three-dimensional data acquisition system based on big data and artificial intelligence as described in any one of claims 1 to 4, specifically including an appearance evaluation method and a clothing formulation method.

6. The method for acquiring three-dimensional human body data of the three-dimensional human body data acquisition system based on big data and artificial intelligence as claimed in claim 5, characterized in that: The specific methods for appearance evaluation include: (1) Take a frontal facial image of the user; (2) Recognize the facial contour based on the frontal face image and extract specific facial features, such as eyebrow size, eye size, nose and mouth shape; (3) Determine the first predicted appearance and corresponding confidence level of the face in the face image through the classification model; Determine a second predicted appearance value and a corresponding weight of the face in the face image by using a regression model; (4) Determining the appearance of the face based on the first predicted appearance, the confidence level, the second predicted appearance, and the weight.

7. The method for acquiring three-dimensional human body data of the three-dimensional human body data acquisition system based on big data and artificial intelligence according to claim 6, characterized in that: The clothing making method specifically comprises: (1) Scanning and photographing the user's body to obtain the user's body image; (2) extracting human body contour information, human body information and clothing information based on the body image acquired by the human body information acquisition module; (3) determining whether the user's human body information feature values ​​are stored in a user model database; if the determination is yes, obtaining the parameter values ​​of the height range, face shape range, body shape range, and gender of the specific target person corresponding to the user's human body information from the user model database; if the determination is no, generating specific parameter values ​​corresponding to the specific target person from the user's human body information according to an SVM algorithm, and storing the specific parameter values ​​in the user model database; (4) generating a 3D human body model according to the specific parameter values ​​corresponding to the human body information of the user and the facial value of the user, and storing the 3D human body model in the user model database; (5) determining whether the clothing information of the user is stored in the clothing model database; if so, extracting the clothing model corresponding to the clothing information; if not, finding parameters that are identical or similar to the above parameters based on three or more parameters in the specific clothing information; then selecting the clothing model corresponding to the clothing information with the most identical or similar parameters; (6) generating a 3D clothing image corresponding to the clothing model according to the clothing model obtained by the clothing model extraction module and the style, category, color, fabric, auxiliary materials, and accessories of the user's clothing information; (7) Based on the 3D human body model generated by the human body model establishment module, the 3D clothing image generated by the 3D clothing image generation module is positioned in three-dimensional space so that the 3D clothing image is accurately attached to the corresponding part of the 3D human body model, showing the 3D effect when the user puts on the clothes.

Images