Female bra recommendation method and system based on LSTM neural network

Abstract

The invention provides a female bra recommendation method and system based on an LSTM neural network, and relates to the field of bra recommendation, and the method comprises the steps: obtaining female human body data, and building a data set; preprocessing the data; according to a statistical analysis result, determining seven characteristic indexes influencing the female chest form; according to a classification result, six breast types are obtained; based on an LSTM algorithm, establishing a chest type discrimination model; according to a pressure test result, determining an optimal bra style of each breast shape; constructing a bra matching model; and obtaining the chest size and dressing preference information of the user, inputting the information into a recommendation system, and outputting a recommendation result with a recommendation size, a recommendation style and a reference picture. And the user size and the dressing preference information are input into the recommendation system, the existing bra style and size on the market are accurately recommended to the corresponding user, and the suitability and practicability of bra recommendation are enhanced.

Description

Technical Field

[0001] The present application relates to the field of bra recommendation, and in particular to a method and system for recommending women's bras based on an LSTM neural network. Background Art

[0002] With the development of information technology, the technology of recommending various personalized information by inputting human body size has gradually entered people's daily life. For example, many clothing companies have developed mass customization for groups and tailor-made recommendation systems for individuals. Major e-commerce platforms have also begun to provide consumers with clothing size matching services. However, behind the extremely high sales volume of online clothing shopping, there is a very high return rate, and most of the reasons are due to ill-fitting. Among them, the ill-fitting problem of bras has caused trouble for most women, and the adverse effects of wearing ill-fitting bras on women in the adolescent development stage are particularly prominent. However, in general, the current technology of information recommendation based on human body size still has many technical problems to be solved. For example, there is no comprehensive user body feature extraction and analysis tool, no appropriate breast shape classification standard has been established, no professional configuration management system that combines user body features to recommend appropriate information, and the combination of human body features and recommended information relies on manual experience, etc.

[0003] In summary, the existing bra recommendation system has technical problems such as the portrayal of a relatively single part, the measurement of human body features is not comprehensive enough, the reliability of the measurement values ​​output by the image is not high, or the limitations of the algorithm. Summary of the invention

[0004] The purpose of the present invention is to provide a method and system for recommending women's bras based on an LSTM neural network to solve the technical problems of existing bra recommendation systems, such as relatively single depicted parts, incomplete measured human features, low reliability of measured values ​​of image output, or limitations of algorithms.

[0005] The above-mentioned purpose of the present application is achieved through the following technical solutions: S1: Obtain female human body data; construct a data set based on the female human body data; S2: preprocess the data set; S3: Conduct correlation analysis and factor analysis on the preprocessed data set to determine the main characteristic indicators that affect the female breast shape; S4: Perform cluster analysis on the data corresponding to the main characteristic indicators in the data set to obtain the breast shape recognition results; S5: Based on the LSTM algorithm, a breast type discrimination model is established; the breast type discrimination model is trained through breast shape recognition results and data sets; S6: Measure the pressure values ​​of subjects with different breast shapes wearing different bra styles and build a bra matching model; the bra styles include: full cup, three-quarter cup, half cup, triangle cup, vest, and tube top; S7: Obtain chest measurement data and clothing preference information of the user to be tested; determine the chest shape of the user to be tested through a chest type discrimination model; clothing preference information includes: loose without pressure and tight with support; output recommended results of bra style and bra size through a bra matching model, combined with the chest shape, clothing preference information and chest measurement data of the user to be tested.

[0006] Optionally, step S1 includes: The female body data includes: three-dimensional body size scanning data and three-dimensional body size measurement data; S11: Use the Human solution 3D body scanner to perform a 3D scan on the female to obtain 3D body size scan data; S12: According to ergonomics, the characteristic points and characteristic lines of women are manually calibrated to perform three-dimensional body size measurement to obtain three-dimensional body size measurement data; S13: constructing a data set according to the three-dimensional human body size scanning data and the three-dimensional human body size measurement data.

[0007] Optionally, step S2 includes: S21: Use the box plot in SPSS software to remove outliers and missing values ​​in the data set; S22: Perform a normality test on the distribution type of samples in the data set.

[0008] Optionally, step S3 includes: S31: Correlation analysis was performed on the chest measurement items in the data set to obtain 11 measurement items for chest shape classification; the 11 measurement items include: chest circumference, chest width, lower chest circumference, chest convex angle, chest thickness, chest point distance, front neck point to chest point distance, lower breast cup arc length, outer breast cup arc length, breast base circumference, and breast transverse width; S32: KMO test and Bartlett test were used to verify the sample data of 11 measurement items; S33: Factor analysis was performed on the sample data of 11 measurement items to determine the main characteristic indicators that ultimately affect the shape of female breasts; the main characteristic indicators include: chest circumference, lower chest circumference, breast width, chest convex angle, lower breast cup arc length, breast base circumference, and distance from the front neck point to the chest point.

[0009] Optionally, step S4 includes: S41: Use the K-Means method in cluster analysis to quickly cluster the data corresponding to the main characteristic indicators in the data set to obtain the cluster analysis results; S42: Based on the cluster analysis results, a side profile description is performed on the sample at the center of the breast shape cluster; the side profile description includes: breast fullness and breast firmness; S43: Obtain a breast shape recognition result based on the side profile description result.

[0010] Optionally, step S5 includes: S51: construct a breast shape classification dataset through the dataset and breast shape recognition results; S52: Use the mini-batch optimization algorithm combined with the breast type classification dataset to train the breast type discrimination model.

[0011] Optionally, step S6 includes: S61: measuring the pressure values ​​of the subjects with different chest shapes wearing different bra styles; using an airbag contact pressure tester, performing pressure tests on points a, b, c, d, e, and f on the chest of the subjects with different chest shapes to obtain pressure test results of each point; S62: Based on the stress test results, determine the optimal bra style ranking table for each breast type and build a bra matching model.

[0012] A female bra recommendation system based on LSTM neural network, the system comprising: An information input module is used to input the chest measurement data and clothing preference information of the user to be tested; The chest measurement data include: lower chest circumference, chest circumference, lower breast cup arc length, breast base circumference, distance from front neck point to chest point, and breast width; Information on clothing preferences includes: loose without oppression and tight with support; Breast shape recognition module, used to build a breast shape discrimination model; The chest shape recognition module is also used to read chest measurement data through a chest shape discrimination model and output the chest shape of the user to be measured; The optimal style recommendation module for breast shape is used to build a bra matching model; The optimal breast style recommendation module is also used to output the recommended results of bra style and bra size through the bra matching model, combining the breast shape, clothing preference information and chest measurement data of the user to be tested.

[0013] A computer-readable storage medium stores instructions. When the instructions are executed, a female bra recommendation method based on LSTM neural network is executed.

[0014] The beneficial effects of the technical solution provided by this application are: By acquiring female body data, a data set is established and data preprocessing is performed; according to the results of correlation analysis and factor analysis, seven characteristic indicators that affect the female chest shape are determined, and six chest shapes are obtained according to the results of K-Means clustering and factor extraction; based on the LSTM algorithm, the sample is identified as one of the six chest types according to the measurement data, and a chest type discrimination model is established; according to the pressure test results, the optimal bra style for each chest type is determined; a bra matching model is constructed; the user's chest size and clothing preference information is obtained and input into the recommendation system, and the recommendation results with recommended size, recommended style, and reference pictures are output; the underwear recommendation results are automatically calculated and generated by the system after the user inputs the size information and clothing preference, and can be viewed by clicking "Get". The user can enter the corresponding entry into the online shopping software according to the recommended style and recommended size of the recommendation result, and then accurately find the corresponding bra style based on the reference picture, realizing simple, convenient and high-precision recommendation, meeting the urgent needs of women for wearing fitted underwear. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] The present application will be further described below with reference to the accompanying drawings and embodiments, in which: Figure 1 is a step diagram in an embodiment of the present application; Figure 2 It is a schematic diagram of the discriminant model construction structure in the embodiment of the present application; Figure 3 is an operation flow chart of the female underwear recommendation system in an embodiment of the present application; Figure 4 is a schematic diagram of the operation interface in an embodiment of the present application; Figure 5 is a schematic diagram of cross-sectional shapes corresponding to the six types of chest shapes obtained by classification in the embodiment of the present application; Figure 6 It is a schematic diagram of the underwear sizes and underwear styles that are most suitable for various breast shapes in the embodiments of the present application; Figure 7 It is a schematic diagram of pressure test points on various styles of various breast shapes in the embodiments of the present application. DETAILED DESCRIPTION

[0016] In order to have a clearer understanding of the technical features, purposes and effects of the present application, the specific implementation methods of the present application are now described in detail with reference to the accompanying drawings.

[0017] An embodiment of the present application provides a method for recommending women's bras based on an LSTM neural network.

[0018] Please refer to Figure 1 , Figure 1 This is a step diagram of a female bra recommendation method based on an LSTM neural network in an embodiment of the present application, including: S1: Obtain female human body data; construct a data set based on the female human body data; S2: preprocess the data set; S3: Conduct correlation analysis and factor analysis on the preprocessed data set to determine the main characteristic indicators that affect the female breast shape; S4: Perform cluster analysis on the data corresponding to the main characteristic indicators in the data set to obtain the breast shape recognition results; S5: Based on the LSTM algorithm, a breast type discrimination model is established; the breast type discrimination model is trained through breast shape recognition results and data sets; S6: Measure the pressure values ​​of subjects with different breast shapes wearing different bra styles and build a bra matching model; the bra styles include: full cup, three-quarter cup, half cup, triangle cup, vest, and tube top; S7: Obtain chest measurement data and clothing preference information of the user to be tested; determine the chest shape of the user to be tested through a chest type discrimination model; clothing preference information includes: loose without pressure and tight with support; output recommended results of bra style and bra size through a bra matching model, combined with the chest shape, clothing preference information and chest measurement data of the user to be tested.

[0019] Specifically, the final underwear recommendation result is automatically calculated by the system after the user inputs the size information and clothing preferences, and can be viewed by clicking "Get"; users can enter the corresponding entries into the online shopping software based on the recommended styles and sizes of the underwear recommendation results, and then accurately find the corresponding bra style based on the reference pictures, achieving simple, convenient and high-precision recommendations, meeting women's urgent needs for wearing fitted underwear. LSTM (Long Short-Term Memory) is a special recurrent neural network.

[0020] For details, see Figure 4 Schematic diagram of the operating interface of the underwear recommendation system. It should be noted that, in this embodiment, the underwear recommendation results are automatically calculated and generated by the system after the user inputs the size information and clothing preferences, and can be viewed by clicking "Get". The user can enter the corresponding entries into the online shopping software based on the recommended styles and recommended sizes of the recommendation results, and then accurately find the corresponding bra styles based on the reference pictures, thereby realizing simple, convenient and high-precision recommendations, meeting women's urgent needs for wearing well-fitting underwear.

[0021] Step S1 includes: The female body data includes: three-dimensional body size scanning data and three-dimensional body size measurement data; S11: Use the Human solution 3D body scanner to perform a 3D scan on the female to obtain 3D body size scan data; S12: According to ergonomics, the characteristic points and characteristic lines of women are manually calibrated to perform three-dimensional body size measurement to obtain three-dimensional body size measurement data; S13: constructing a data set according to the three-dimensional human body size scanning data and the three-dimensional human body size measurement data.

[0022] Specifically, in order to ensure the consistency and standardization of the data source, the female human body data of the experiment was collected by Humansolution 3D human body scanner and manual measurement. In addition, a data set was established based on the collected human body data. Since this topic is mainly based on the classification of female chest morphology, when determining the final sample size, the allowable error and standard deviation of female chest circumference were used as the basis for calculating the sample size. With reference to relevant research, the minimum sample size was determined to be 52. Considering the elimination of outliers, as well as subsequent evaluation and verification, 126 samples were finally collected to meet the experimental requirements.

[0023] Step S2 includes: S21: Use the box plot in SPSS software to remove outliers and missing values ​​in the data set; Specifically, SPSS software, whose Chinese name is "Statistical Products and Services Solutions", is a statistical analysis software widely used in statistical analysis operations, data mining, predictive analysis and decision support tasks.

[0024] S22: Perform a normality test on the distribution type of samples in the data set; S23: Using the descriptive analysis function of SPSS software, analyze the impact of different measurement items on the results of female breast shape identification, and determine the 11 measurement items with the largest impact values, which are the breast measurement items required for subsequent analysis. [Love Buried in the Coral Sea 1] Specifically, using the descriptive analysis function of SPSS to analyze the overall characteristics of the data can simply, directly and effectively reflect some data characteristics. In this study, the average chest circumference of women is 84.0cm, the average lower chest circumference is 74.6cm, and the average chest difference is 9.4cm. The average value is very important for determining the middle value of the body shape. It can reflect the central tendency of the data. Therefore, it can be judged that the central tendency of the chest circumference of the measured sample is small, which helps to judge the accuracy of the chest type classification results. The coefficient of variation of the chest difference is the largest, followed by the arc length of the lower breast cup, indicating that the discrete degree of the two parts is large, and the discrete degree of the other parts is small. It can be inferred that these two measurement items have a great influence on the results of breast type classification.

[0025] In this application, in order to ensure that the research data is true and accurate, and to exclude systematic errors and accidental errors, the box plot in SPSS software is used to pre-process the data to determine whether there are outliers or missing values. After the outliers are eliminated, the distribution type of the sample is tested for normality to determine whether its distribution is consistent with the normal distribution of random sampling. The scattered distribution of the data is close to a straight line, indicating that the data obtained after excluding outliers are all normally distributed, indicating that the sample distribution obeys the law of random sampling data.

[0026] Step S3 includes: S31: Correlation analysis was performed on the chest measurement items in the data set to obtain 11 measurement items for chest shape classification; the 11 measurement items include: chest circumference, chest width, lower chest circumference, chest convex angle, chest thickness, chest point distance, front neck point to chest point distance, lower breast cup arc length, outer breast cup arc length, breast base circumference, and breast transverse width; S32: KMO test and Bartlett test were used to verify the sample data of 11 measurement items; S33: Factor analysis was performed on the sample data of 11 measurement items to determine the main characteristic indicators that ultimately affect the shape of female breasts; the main characteristic indicators include: chest circumference, lower chest circumference, breast width, chest convex angle, lower breast cup arc length, breast base circumference, and distance from the front neck point to the chest point.

[0027] Specifically, in order to exclude the measurement items with extremely low or extremely high correlation with chest morphology, so that the chest measurement items can reach a positive definite matrix and thus meet the requirements of factor analysis, the correlation analysis of the existing measurement items is carried out. 1) The height has a high correlation coefficient only with chest circumference height, and the correlation coefficients with other chest-related measurement items are small, and chest circumference height cannot directly reflect chest characteristics, indicating that the height has a low correlation with chest size; 2) The chest circumference has a high correlation coefficient with lower chest circumference, chest thickness, chest convex angle, chest point distance, outer breast cup arc length, breast width and back width, indicating that there is a large correlation between chest circumference and breast characteristic size; 3) The chest difference has a strong correlation only with lower chest circumference, chest width, and chest point distance, and has a low correlation with other measurement data; 4) Back width has a high correlation only with chest circumference and chest thickness, and has a low correlation with other measurement items, and has limited description of the human chest. 5) The correlation between breast root circumference and breast depth is consistent with other measurement items, so you can choose one.

[0028] Based on the above analysis, 11 measurement items (chest circumference, chest width, lower chest circumference, chest convex angle, chest thickness, chest point distance, front neck point to chest point distance, lower breast cup arc length, outer breast cup arc length, breast base circumference, breast transverse width) were finally obtained for factor analysis.

[0029] In order to reduce the dimension of the data set of anthropometric items and retain the items that contribute the most to the variance in the data set, the 11 measurement items previously retained were factor analyzed to explore the main factors affecting the characteristics of female breasts. Finally, the main characteristic indicators affecting the shape of female breasts were determined to be the lower chest circumference, breast width, chest convex angle, lower breast cup arc length, breast root circumference, front neck point to chest point distance, and chest circumference. These indicators can reflect the circumference, arc, angle, width, and degree of uprightness of the female chest shape, and can describe the three-dimensional shape of the chest.

[0030] Step S4 includes: S41: Use the K-Means method in cluster analysis to quickly cluster the data corresponding to the main characteristic indicators in the data set to obtain the cluster analysis results; S42: Based on the cluster analysis results, a side profile description is performed on the sample at the center of the breast shape cluster; the side profile description includes: breast fullness and breast firmness; S43: Obtain a breast shape recognition result based on the side profile description result.

[0031] Specifically, the breast shape recognition results are as follows: Figure 6As shown, including: H type, Q type, C type, S type, P type, O type; H type breasts are less full and more upright, named micro breast type. Q type breasts are generally full, but less upright, named disc type. C type breasts are full and prominent, but generally upright, named hemispherical type. S type breasts are full, with uneven breasts, but generally upright, named papaya type. P type breasts are more natural overall, but prominent in circumference, named water drop type. O type breasts are relatively small overall, but more upright, named cone type. Parameters with the same letter can represent different meanings. H type: The side curve of the chest is H type, the lower chest circumference and chest circumference are both at an average level, and the chest difference is the smallest among the 6 types. The arc length of the lower breast cup, the circumference of the breast root, and the chest convex angle are all close to the maximum value, indicating that the breasts are less full and more upright. The second type of chest is Q-shaped on the side, with the lower chest circumference, chest circumference and chest difference all at the same level, and the chest convex angle and the distance from the front neck point to the chest point are at the lowest level, indicating that the breast shape of this type is generally full, but the degree of uprightness is poor. The third type of chest side curve is C-shaped, with the chest circumference and lower chest circumference at the maximum value, and the chest difference is close to the maximum value. Except for the chest convex angle at a relatively small level, the other items are the largest among the six categories. It can be seen that this type of chest is full and prominent, but the degree of uprightness is generally average. The fourth type of chest side curve is S-shaped, with the smallest lower chest circumference and a small chest circumference, but the chest difference is the maximum among the six categories. All other items are at a medium to lower level, indicating that the chest is full and the breasts are well-proportioned, but the degree of uprightness is generally average. The fifth type of chest side curve is P-shaped, with 7 characteristic indicators at a general level, and the chest difference is at a relatively large level, indicating that the overall shape of the chest is relatively natural, but the circumference is prominent. The sixth type of chest side curve is similar to an O-shaped, with the chest convex angle being the maximum value among the six categories, the breast base circumference being close to the minimum value among the six categories, and the rest of the items are all minimum values. It means that the overall shape of the chest is relatively small, but it is quite upright.

[0032] Specifically, in order to show the clustering results more clearly, the corresponding side profile descriptions are performed for the samples close to the cluster centers of the six types of breast shapes. Figure 5 The cross-sectional shapes corresponding to the 6 types of chest shapes obtained by classification.

[0033] Step S5 includes: S51: construct a breast shape classification dataset through the dataset and breast shape recognition results; S52: Use the mini-batch optimization algorithm combined with the breast type classification dataset to train the breast type discrimination model.

[0034] Specifically, Matlab software was used to construct an LSTM (‌Long Short-Term Memory) neural network discrimination model. The breast type discrimination model was constructed as follows: Figure 2As shown in the figure. Since the neural network has a strong fitting ability, the sample data needs to be randomly shuffled after the data is imported to avoid the same combination from recurring. The model may "remember" the order of these samples, thus affecting the generalization ability. At the same time, the dropout parameter is used to randomly hide some nodes of the full connection, and the small batch optimization algorithm is used to train the model to avoid falling into the local optimal situation.

[0035] Step S6 includes: S61: measuring the pressure values ​​of the subjects with different chest shapes wearing different bra styles; using an airbag contact pressure tester, performing pressure tests on points a, b, c, d, e, and f on the chest of the subjects with different chest shapes to obtain pressure test results of each point; S62: Based on the stress test results, determine the optimal bra style ranking table for each breast type and build a bra matching model.

[0036] For details, see Figure 5 , Figure 7 According to the pressure test results, the best bra style for each breast type is determined. In the pressure test, an airbag contact pressure tester is used to record the pressure values ​​of each volunteer's chest when wearing each bra, so as to select a bra that is more suitable for them; the schematic diagram of the pressure test points on various styles of various breast types is shown below, Figure 7 As shown. Point A is five millimeters below the intersection of the three-quarter boundary line of the left chest and the vertical line of the BP point; point E is below the BP point and half of the BP vertical line segment; point B is a point on the BP horizontal line that can be connected with point E to form a curve and has the same bending trend as the bra; point F is the midpoint of the E and B curves; D is the corresponding point of point F with the BP vertical line as the axis; point C is a point on the BP horizontal line that is 10 mm closer to the side seam than point B. a, b, c, d, e, and f are the corresponding points of the left chest on the right chest. Points a, b, c, d, e, and f are the six key feature points that affect the shape of young women's chests. Based on the principle of ergonomics, these points are located in the key parts of the breasts and can fully reflect the pressure of the bra on the breasts; based on the structural characteristics of the breasts, these points are also the parts of the breasts that are easily compressed. Evaluate the pressure distribution of the bra in these parts to ensure that it does not cause excessive compression or discomfort to the breasts.

[0037] After the test begins, pressure values ​​are collected at each point every second. After the pressure test, each tester evaluates the comfort of the three bras they are wearing. H, Q, C, S, P, and O represent the first, second, third, fourth, fifth, and sixth types of breast shapes, respectively. From the scores, we can see that the most comfortable bra for the first type of breast shape is the triangle cup type; the most comfortable bra for the second type of breast shape is the tube top type; the most comfortable bra for the third type of breast shape is the triangle cup type; the most comfortable bra for the fourth type of breast shape is the three-quarter cup type; the most comfortable bra for the fifth type of breast shape is the half cup type; and the most comfortable bra for the sixth type of breast shape is the tube top type. The following diagrams show the most suitable bra sizes and bra styles for each type of breast shape, as shown in the figure below. Figure 5 shown.

[0038] A female bra recommendation system based on LSTM neural network, the system includes: An information input module is used to input the chest measurement data and clothing preference information of the user to be tested; The chest measurement data include: lower chest circumference, chest circumference, lower breast cup arc length, breast base circumference, distance from front neck point to chest point, and breast width; Information on clothing preferences includes: loose without oppression and tight with support; Breast shape recognition module, used to build a breast shape discrimination model; The chest shape recognition module is also used to read chest measurement data through a chest shape discrimination model and output the chest shape of the user to be measured; The optimal style recommendation module for breast shape is used to build a bra matching model; The optimal breast style recommendation module is also used to output the recommended results of bra style and bra size through the bra matching model, combining the breast shape, clothing preference information and chest measurement data of the user to be tested.

[0039] Specifically, by acquiring female human body data, a data set is established, and data preprocessing is performed based on the data set to ensure the authenticity and accuracy of the research data, eliminate systematic errors and accidental errors, simplify the subsequent analysis process, and facilitate self-measurement. According to the results of correlation analysis and factor analysis, 7 characteristic indicators that affect the female chest shape are determined. According to the K-Means clustering results and factor extraction results, the chest type recognition results are obtained. Based on the LSTM algorithm, according to the measurement data, the sample is identified as one of the six types of chest shapes, and a chest type discrimination model is established. According to the pressure test results, the optimal bra style for each chest type is determined, and a bra matching model is constructed. The user's chest size and clothing preference information is obtained, and the recommended results with recommended size, recommended style, and reference pictures are output. The underwear recommendation results are automatically calculated and generated by the system after the user inputs the size information and clothing preference, and can be viewed by clicking "Get". The user can enter the corresponding entry into the online shopping software according to the recommended style and recommended size of the recommendation result, and then accurately find the corresponding bra style based on the reference picture, so as to achieve simple, convenient and high-precision recommendation, and meet the urgent needs of women for wearing fitted underwear.

[0040] The present application also discloses a computer-readable storage medium storing a plurality of instructions, which are suitable for a processor to load to execute the above-mentioned method for recommending women's bras based on an LSTM neural network.

[0041] The above are only exemplary embodiments of the present disclosure and cannot be used to limit the scope of the present disclosure. That is, any equivalent changes and modifications made according to the teachings of the present disclosure are still within the scope of the present disclosure. After considering the disclosure of the specification and the truth of practice, those skilled in the art will easily think of other embodiments of the present disclosure.

[0042] This application is intended to cover any variation, use or adaptation of the present disclosure, which follows the general principles of the present disclosure and includes common knowledge or customary technical means in the art not described in the present disclosure. The description and examples are to be regarded as exemplary only, and the scope and spirit of the present disclosure are defined by the claims.

Claims

1. A female bra recommendation method based on LSTM neural network, characterized in that: The method comprises the following steps: S1: Obtain female human body data; construct a data set based on the female human body data; S2: preprocess the data set; S3: Conduct correlation analysis and factor analysis on the preprocessed data set to determine the main characteristic indicators that affect the female breast shape; S4: Perform cluster analysis on the data corresponding to the main characteristic indicators in the data set to obtain the breast shape recognition results; S5: Based on the LSTM algorithm, a breast type discrimination model is established; the breast type discrimination model is trained through breast shape recognition results and data sets; S6: Measure the pressure values ​​of subjects with different breast shapes wearing different bra styles and build a bra matching model; the bra styles include: full cup, three-quarter cup, half cup, triangle cup, vest and tube top; S7: Obtain chest measurement data and clothing preference information of the user to be tested; determine the chest shape of the user to be tested through a chest type discrimination model; clothing preference information includes: loose without pressure and tight with support; output recommended results of bra style and bra size through a bra matching model, combined with the chest shape, clothing preference information and chest measurement data of the user to be tested.

2. A female bra recommendation method based on LSTM neural network as claimed in claim 1, characterized in that: Step S1 includes: The female body data includes: three-dimensional body size scanning data and three-dimensional body size measurement data; S11: Use the Human solution 3D body scanner to perform a 3D scan on the female to obtain 3D body size scan data; S12: According to ergonomics, the characteristic points and characteristic lines of women are manually calibrated to perform three-dimensional body size measurement to obtain three-dimensional body size measurement data; S13: constructing a data set according to the three-dimensional human body size scanning data and the three-dimensional human body size measurement data.

3. A female bra recommendation method based on LSTM neural network as claimed in claim 1, characterized in that: Step S2 includes: S21: Use the box plot in SPSS software to remove outliers and missing values ​​in the data set; S22: Perform a normality test on the distribution type of samples in the data set.

4. A female bra recommendation method based on LSTM neural network as claimed in claim 1, characterized in that: Step S3 includes: S31: Correlation analysis was performed on the chest measurement items in the data set to obtain 11 measurement items for chest shape classification; the 11 measurement items include: chest circumference, chest width, lower chest circumference, chest convex angle, chest thickness, chest point distance, front neck point to chest point distance, lower breast cup arc length, outer breast cup arc length, breast base circumference, and breast transverse width; S32: KMO test and Bartlett test were used to verify the sample data of 11 measurement items; S33: Factor analysis was performed on the sample data of 11 measurement items to determine the main characteristic indicators that ultimately affect the shape of female breasts; the main characteristic indicators include: chest circumference, lower chest circumference, breast width, chest convex angle, lower breast cup arc length, breast base circumference, and distance from the front neck point to the chest point.

5. A female bra recommendation method based on LSTM neural network as claimed in claim 1, characterized in that: Step S4 includes: S41: Use the K-Means method in cluster analysis to quickly cluster the data corresponding to the main characteristic indicators to obtain the cluster analysis results; S42: Based on the cluster analysis results, a side profile description is performed on the sample at the center of the breast shape cluster; the side profile description includes: breast fullness and breast firmness; S43: Obtain a breast shape recognition result based on the side profile description result.

6. A method for recommending women's bras based on LSTM neural network as claimed in claim 1, characterized in that: Step S5 includes: S51: construct a breast shape classification dataset through the dataset and breast shape recognition results; S52: Use the mini-batch optimization algorithm combined with the breast type classification dataset to train the breast type discrimination model.

7. A female bra recommendation method based on LSTM neural network as claimed in claim 1, characterized in that: Step S6 includes: S61: measuring the pressure values ​​of the subjects with different chest shapes wearing different bra styles; using an airbag contact pressure tester, performing pressure tests on points a, b, c, d, e, and f on the chest of the subjects with different chest shapes to obtain pressure test results of each point; S62: Based on the stress test results, determine the optimal bra style ranking table for each breast type and build a bra matching model.

8. A female bra recommendation system based on LSTM neural network, characterized in that: The system comprises: The information input module is used to input the chest measurement data and clothing preference information of the user to be tested; the chest measurement data include: lower chest circumference, chest circumference, lower breast cup arc length, breast base circumference, distance from front neck point to chest point, breast width; clothing preference information includes: loose without pressure and tight with support; Breast shape recognition module, used to build a breast shape discrimination model; The chest shape recognition module is also used to read chest measurement data through a chest shape discrimination model and output the chest shape of the user to be measured; The optimal style recommendation module for breast shape is used to build a bra matching model; The optimal breast style recommendation module is also used to output the recommended results of bra style and bra size through the bra matching model, combining the breast shape, clothing preference information and chest measurement data of the user to be tested.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores instructions, and when the instructions are executed by a computer, the method according to any one of claims 1 to 8 is executed.

Images