A bmi estimation method based on multi-dimensional feature extraction
By employing a multi-dimensional feature extraction method, combining three-dimensional and two-dimensional anthropometric features with deep learning, the problem of BMI measurement difficulties in bedridden patients and outdoor environments was solved, achieving highly accurate BMI estimation.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-25
- Publication Date
- 2026-03-17
AI Technical Summary
Existing technologies face difficulties in measuring BMI in bedridden patients and in outdoor environments. Acquiring two-dimensional facial images raises privacy concerns, while acquiring three-dimensional human body data is sensitive. Furthermore, two-dimensional human body measurement features lack longitudinal information, affecting the accuracy of BMI estimation.
A multi-dimensional feature extraction method is adopted, which combines three-dimensional anthropometric features, deep learning and two-dimensional anthropometric features. Deep features are extracted through a deep convolutional neural network model, and BMI is estimated using methods such as support vector regression and kernel ridge regression.
It achieves high-accuracy BMI estimation in a single image, overcoming the measurement difficulties and privacy issues of traditional methods, and improving the accuracy and wide applicability of BMI estimation.
Smart Images
Figure CN119477832B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of BMI estimation technology, specifically a BMI estimation method based on multi-dimensional feature extraction. Background Technology
[0002] Body Mass Index (BMI) is a commonly used indicator to measure the degree of obesity in the human body, classifying individuals as underweight, normal weight, overweight, or obese. BMI is calculated by dividing weight (kg) by the square of height (m). Generally, individuals with a normal body type are considered more attractive and look healthier than those who are abnormally obese.
[0003] In recent years, research on predicting BMI has received increasing attention. Traditionally, BMI is usually obtained by measuring an individual's height and weight. However, this method presents certain difficulties for bedridden patients, as it requires their cooperation and often necessitates an upright standing posture. Therefore, to address this issue, visual-based BMI measurement methods have emerged.
[0004] Vision-based BMI measurement is a contactless method that not only makes the measurement process more comfortable but also effectively avoids the infection risks associated with touch sensors. Compared to traditional contact-based measurement methods, this vision-based method offers greater convenience and safety. Furthermore, vision-based BMI estimation methods support telemedicine applications, enabling measurements without face-to-face contact. This makes large-scale BMI measurements possible, allowing simultaneous estimation of BMI values for multiple individuals, significantly improving efficiency and applicability.
[0005] In previous studies, some researchers proposed predicting BMI from facial images by analyzing the relationship between facial features and BMI values. However, compared to two-dimensional human images, clear, frontal facial images are generally more difficult to obtain due to privacy concerns. Furthermore, BMI prediction based on facial images also faces certain accuracy challenges.
[0006] To overcome these challenges, some researchers have used 3D human body data or RGB-D images to calculate BMI values. While these methods can estimate BMI values relatively accurately under certain conditions, RGB-D image acquisition is highly sensitive to sunlight and measurement distance, making it extremely difficult to obtain high-quality 3D human body data in outdoor environments.
[0007] With the continuous development of computer technology and image processing technology, some scholars have used three-dimensional human body reconstruction methods to extract three-dimensional anthropometric features such as waist circumference and hip circumference from two-dimensional images, and then used regression methods to estimate BMI values. These three-dimensional anthropometric features have been shown to be closely related to BMI, and can also be predicted using parametric models. However, learning from these three-dimensional anthropometric features still cannot strongly represent human body shape information.
[0008] In recent years, many health science studies have shown that some two-dimensional anthropometric indicators are closely related to BMI. However, most two-dimensional anthropometric features, such as waist-to-shoulder ratio, waist-to-hip ratio, hip-to-head ratio, and waist-to-thigh ratio, are only horizontal ratios of the human body and do not include longitudinal information.
[0009] With the development of computing power and big data, deep learning, especially convolutional neural networks, has made significant progress in image processing. Deep network models extracting deep features have been applied to BMI estimation tasks. One scholar proposed a three-branch BMI estimation method that fuses manually defined 3D anthropometric features with features extracted by a CNN model to improve the accuracy of BMI estimation. This demonstrates that deep features can effectively represent the semantic features of images and improve the accuracy of BMI estimation.
[0010] To address this issue, those skilled in the art have proposed a BMI estimation method based on multi-dimensional feature extraction. Summary of the Invention
[0011] To address the shortcomings of existing technologies, this invention provides a BMI estimation method based on multi-dimensional feature extraction. It starts from two-dimensional and three-dimensional perspectives and extracts the depth features of images through deep learning methods, aiming to construct a BMI estimation method based on multi-dimensional feature extraction.
[0012] To achieve the above objectives, the present invention provides the following technical solution:
[0013] A BMI estimation method based on multi-dimensional feature extraction specifically includes the following steps:
[0014] Step 1, 3D human body measurement feature extraction: Extract 3D human body measurement features based on human pose estimation and 3D human body reconstruction methods;
[0015] Step 2, Two-dimensional human measurement feature extraction: Two-dimensional human measurement features are extracted based on human skeleton point detection and human contour detection methods;
[0016] Step 3, Deep Feature Extraction: A deep network model incorporating an attention mechanism is proposed for extracting deep features;
[0017] Step 4, Model Training and Selection: The training set is trained using SVR, KRR, DTR, and GRR regression methods; the test set images are input into the four established regression methods for BMI estimation, and the performance of the model is verified by evaluation indicators. The regression method with the best performance is selected; SVR is support vector regression; KRR is kernel ridge regression; DTR is decision tree regression; and GPR is Gaussian process regression.
[0018] As a further preferred embodiment of the BMI estimation method based on multi-dimensional feature extraction of the present invention, step 1 specifically includes the following steps:
[0019] Step 1.1: Data preprocessing is performed using the OpenPose method for human pose estimation to obtain human pose information;
[0020] Step 1.2: Use the SMPL-X human 3D model for human 3D reconstruction to obtain body shape, body offset and posture parameters;
[0021] Step 1.3: Use the Shapy method of 3D body regression to calculate the 3D anthropometric features of waist circumference, hip circumference, and the ratio of waist circumference to hip circumference.
[0022] As a further preferred embodiment of the BMI estimation method based on multi-dimensional feature extraction of the present invention, step 2 specifically includes the following steps:
[0023] Step 2.1: Obtain skeletal point information using the human skeletal point detection method;
[0024] Step 2.2: Perform human body contour detection;
[0025] Step 2.3: Calculate the two-dimensional anthropometric features: waist-to-thigh width ratio, waist-to-hip ratio, waist-to-head width ratio, hip-to-head width ratio, body area between waist and hips, Euclidean distance from nose to knee to waist width ratio, waist width to shoulder width ratio, and Euclidean distance from nose to knee to hip ratio.
[0026] As a further preferred embodiment of the BMI estimation method based on multi-dimensional feature extraction of the present invention, step 3 specifically includes the following steps:
[0027] A deep network model incorporating an attention mechanism is proposed for extracting deep features. This deep learning model is based on the deep convolutional neural network model VGG and adds a lightweight visual Transformer MobileViT attention mechanism, which helps the network to focus more on important regions in the image and improves the network's ability to extract key features.
[0028] As a further preferred embodiment of the BMI estimation method based on multi-dimensional feature extraction of the present invention, the specific steps of the three-dimensional anthropometric feature calculation method in step 1.3 are as follows:
[0029] Step 1.31, calculate the waist circumference WC, the formula is as follows:
[0030]
[0031] Where i and j are two points in the convex hull, and β is the shape parameter of SMPL-X; (x i y i , z i ) represents the coordinates of a landmark point on the waist, t i Index for the body triangle;
[0032] Step 1.32, calculate the hip circumference HC; given the coordinates of the landmark point of the hip (h i m i n i ) and the body triangle index t i The hip circumference can be calculated using formula (1).
[0033] Step 1.33, calculate the ratio of waist circumference to hip circumference, WCHCR; based on the circumferences of the waist and hips, it can be calculated using the following formula:
[0034]
[0035] As a further preferred embodiment of the BMI estimation method based on multi-dimensional feature extraction of the present invention, the specific steps of the two-dimensional anthropometric feature calculation method in step 2.3 are as follows:
[0036] Step 2.31, calculate the waist-to-shoulder width ratio (WSR); Given similar BMIs, some people have wider shoulders or shorter waists, and WSR represents this difference; the calculation formula based on waist and shoulder width is as follows:
[0037]
[0038] Among them, D w For waist width, D s The shoulder width is given; the vertical coordinate of the waist is provided, and the formula for calculating the waist width is as follows:
[0039]
[0040] Where st represents the constraint condition, l represents the linear function, and b represents the intercept; x1 and x2 have numerical ranges under the constraint conditions, meaning they are two variables; x ew The x-coordinate of the waist center point can be approximated by the x-coordinate of the center shoulder.cs The calculation formula is as follows:
[0041]
[0042] Where, x rs Let x be the x-coordinate of the right shoulder. ls The x-coordinate of the left shoulder; y in formula (4) w Let be the vertical coordinate of the waist, and its calculation formula is as follows:
[0043]
[0044] Among them, y cs The y-coordinate is the center shoulder; h Let be the ordinate of the buttocks, calculated using the following formula:
[0045]
[0046] Among them, y lh Let y be the ordinate of the left hip. rh The y-coordinate is the right hip. cs The ordinate of the center shoulder is calculated using the following formula:
[0047]
[0048] Among them, y rs Let y be the ordinate of the right shoulder. ls D is the ordinate of the left shoulder; s The calculation formula is as follows:
[0049]
[0050] Step 2.32, calculate the ratio of waist circumference to thigh width WTR; add the distance from the left thigh boundary to the center of the left thigh to the distance from the right thigh boundary to the center of the right thigh as the thigh width, and refer to formula (5) to obtain the thigh width D. t WTR can be calculated using the following formula, based on the width of the waist and thigh:
[0051]
[0052] Step 2.33, calculate the waist-to-hip ratio (WHpR); given the coordinates of the left and right hip joint skeletons, and hip circumference D. hp It can be calculated in the same way as formula (4); WHpR can be calculated by the following formula:
[0053]
[0054] Step 2.34, calculate the waist-to-head width ratio (WHdR); given the coordinates of the skeletal joints of the left and right ears (x...le ,y le ) and (x re ,y re Head width D hd The distance between two points can be approximated using the Euclidean distance, calculated as follows:
[0055]
[0056] Step 2.35, calculate the ratio of hip width to head width, HPHdR. Based on the widths of the hips and head, HPHdR can be calculated using the following formula:
[0057]
[0058] Step 2.36, calculate the body area between the waist and hips (Area). Since the area between the waist and hip circumferences can serve as an indicator of leanness and obesity, this information can be represented using a characteristic area when scale information is unknown. The Area can be calculated using the following formula:
[0059]
[0060]
[0061] Among them, pixcel num The number of pixels in this region, pix x，y This represents the number of pixels at the (x, y) coordinate point.
[0062] Step 2.37, the Euclidean distance from nose to knee and the waist-to-width ratio HWR; HWR can be calculated using the following formula:
[0063]
[0064] Among them, D nk The Euclidean distance from the nose to the knee is calculated using the following formula:
[0065]
[0066] Wherein, the knee coordinate point is (x k ,y k The calculation formula is as follows:
[0067]
[0068] y k =y lk =y rk (twenty one)
[0069] The coordinates of the left knee are (x lk y lkThe coordinates of the right knee are (x rk ,y rk );
[0070] Step 2.38 calculates the Euclidean distance from nose to knee and the hip ratio H2H. Two-dimensional anthropometric features such as waist-to-thigh width ratio, waist-to-hip ratio, waist-to-head width ratio, hip-to-head width ratio, body area between waist and hips, and waist-to-shoulder width ratio only include horizontal aspect ratios and do not include vertical information. While the Euclidean distance from nose to knee and waist-to-width ratio includes vertical information, it is not effective for identifying individuals with similar waist widths but large differences in hip circumference. Therefore, H2H is proposed to represent the aspect ratio; its calculation formula is as follows:
[0071]
[0072] Compared with the prior art, the technical advantages of the present invention are as follows:
[0073] 1) A multi-dimensional feature extraction BMI estimation framework is proposed, which can estimate BMI values from a single two-dimensional image.
[0074] 2) Three-dimensional anthropometric features of waist circumference, hip circumference and the ratio of waist circumference to hip circumference are proposed, as well as two-dimensional anthropometric features and depth features of the Euclidean distance from nose to knee and hip circumference ratio.
[0075] 3) A comprehensive comparison with state-of-the-art methods and extensive ablation studies were conducted to verify the effectiveness of the proposed method. Attached Figure Description
[0076] Figure 1 This is a data processing flowchart of Embodiment 1 of the present invention;
[0077] Figure 2 This is the extracted depth feature map of Embodiment 1 of the present invention;
[0078] Figure 3 This is a performance comparison chart of different deep learning models in Embodiment 1 of the present invention;
[0079] Figure 4 This is a performance comparison chart of different attention modules in Embodiment 1 of the present invention;
[0080] Figure 5 This is a performance comparison chart of different depth feature extraction numbers in Embodiment 1 of the present invention. Detailed Implementation
[0081] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0082] A BMI estimation method based on multi-dimensional feature extraction includes:
[0083] Step 1, 3D human body measurement feature extraction: Extract 3D human body measurement features based on human pose estimation and 3D human body reconstruction methods;
[0084] Step 2, Two-dimensional human measurement feature extraction: Two-dimensional human measurement features are extracted based on human skeleton point detection and human contour detection methods;
[0085] Step 3, Deep Feature Extraction: An Attention-Enhanced VGG (deep network with attention mechanism) model is proposed for extracting deep features;
[0086] Step 4, Model Training and Selection: The training set is trained using SVR, KRR, DTR, and GRR regression methods; the test set images are input into the four established regression methods for BMI estimation, and the performance of the model is verified by evaluation indicators. The regression method with the best performance is selected; SVR is support vector regression; KRR is kernel ridge regression; DTR is decision tree regression; and GPR is Gaussian process regression.
[0087] Step (1) Specific steps:
[0088] First, data preprocessing is performed using the OpenPose method for human pose estimation to obtain human pose information. Then, the SMPL-X 3D human model is used for 3D reconstruction to obtain body shape, body offset, and pose parameters. Finally, the Shapy method is used to calculate 3D anthropometric features, including waist circumference, hip circumference, and the ratio of waist circumference to hip circumference, which are then preprocessed using the OpenPose method to obtain human pose information.
[0089] Step (1) Calculation of 3D human body measurement features:
[0090] 1. Calculate the waist circumference WC using the following formula:
[0091]
[0092] Where i and j are two points in the convex hull, and β is the shape parameter of SMPL-X. (x i yi , z i ) represents the coordinates of a landmark point on the waist, t i Index the body triangle.
[0093] 2. Calculate the hip circumference HC. Given the coordinates of a landmark point on the hip (h... i h i h i ) and the body triangle index t i The hip circumference can be calculated using formula (2).
[0094] 3. Calculate the ratio of waist circumference to hip circumference, WCHCR. Based on the circumferences of the waist and hips, this can be calculated using the following formula:
[0095]
[0096] Step (2) Feature extraction specific steps:
[0097] The human skeleton point detection method is used to obtain skeleton point information, then human contour detection is performed, and finally the two-dimensional human measurement features such as waist to thigh width ratio, waist to hip ratio, waist to head width ratio, hip width to head width ratio, body area between waist and hip, Euclidean distance from nose to knee to waist width ratio, waist width to shoulder width ratio, and Euclidean distance from nose to knee to hip ratio are calculated.
[0098] Step (2) Calculation of two-dimensional human body measurement features:
[0099] 1. Calculate the waist-to-shoulder width ratio (WSR). Even with similar BMIs, some people have wider shoulders or shorter waists; therefore, this article uses WSR to represent this difference. The calculation formula based on waist and shoulder width is as follows:
[0100]
[0101] Among them, D w For waist width, D s The width is the shoulder width. The vertical coordinate of the waist is given, and the formula for calculating waist width is as follows:
[0102]
[0103] Where st represents the constraint, l represents the linear function, and b represents the intercept; x1 and x2 have numerical ranges under the constraint, meaning they are two variables; x cw The x-coordinate of the waist center point can be approximated by the x-coordinate of the center shoulder. cs The calculation formula is as follows:
[0104]
[0105] Where, x rs Let x be the x-coordinate of the right shoulder. ls Let y be the x-coordinate of the left shoulder. In formula (4), y... w Let be the vertical coordinate of the waist, and its calculation formula is as follows:
[0106]
[0107] Among them, y cs The y-coordinate is the center shoulder. h Let be the ordinate of the buttocks, calculated using the following formula:
[0108]
[0109] Among them, y lh Let y be the ordinate of the left hip. rh Let y be the ordinate of the right hip. cs The ordinate of the center shoulder is calculated using the following formula:
[0110]
[0111] Among them, y rs Let y be the ordinate of the right shoulder. ls D represents the ordinate of the left shoulder. s The calculation formula is as follows:
[0112]
[0113] 2. Calculate the ratio of waist circumference to thigh width (WTR). This paper uses the distance from the left thigh boundary to the center of the left thigh plus the distance from the right thigh boundary to the center of the right thigh as the thigh width. Referring to formula (5), the thigh width D can be obtained. t WTR can be calculated using the following formula, based on the width of the waist and thigh:
[0114]
[0115] 3. Calculate the waist-to-hip ratio (WHpR). Given the coordinates of the left and right hip joints and the hip circumference D. hp It can be calculated in the same way as formula (4). Therefore, WHpR can be calculated by the following formula:
[0116]
[0117] 4. Calculate the waist-to-head width ratio (WHdR). Given the coordinates of the skeletal joints of the left and right ears (x... le ,y le ) and (x re ,y re Head width D hdThe distance between two points can be approximated using the Euclidean distance, calculated as follows:
[0118]
[0119] 5. Calculate the hip width to head width ratio (HpHdR). Based on the widths of the hips and head, HPHdR can be calculated using the following formula:
[0120]
[0121] 6. Calculate the body area between the waist and hips (Area). Since the area between the waist and hip circumferences can serve as an indicator of leanness and obesity, it can be used to represent this information when dimensional information is unknown. Area can be calculated using the following formula:
[0122]
[0123] Among them, pixcel num The number of pixels in this region, pix x，y This represents the number of pixels at the (x, y) coordinate point.
[0124] 7. The European distance from nose to knee and the waist-to-width ratio (HWR). HWR can be calculated using the following formula:
[0125]
[0126] Among them, D nk The Euclidean distance from the nose to the knee is calculated using the following formula:
[0127]
[0128] Wherein, the knee coordinate point is (x k ,y k The calculation formula is as follows:
[0129]
[0130] y k =y lk =y rk (twenty one)
[0131] The coordinates of the left knee are (x lk y lk The coordinates of the right knee are (x rk ,y rk ).
[0132] 8. Calculate the Euclidean distance from nose to knee and the hip ratio (H2H). Two-dimensional anthropometric features such as waist-to-thigh width ratio, waist-to-hip ratio, waist-to-head width ratio, hip-to-head width ratio, body area between waist and hips, and waist-to-shoulder width ratio only include horizontal aspect ratios and do not include vertical information. While the Euclidean distance from nose to knee and waist-to-width ratio includes vertical information, it is not effective for identifying individuals with similar waist widths but large differences in hip circumference. Therefore, H2H is proposed to represent the aspect ratio. Its calculation formula is as follows:
[0133]
[0134] Example 1:
[0135] 1. Data Source: The proposed method was evaluated using the publicly available 2D image-to-BMI dataset. This dataset contains 4189 images, including 1477 males and 2712 females. The dataset was randomly split into training and testing sets in a 7:3 ratio. The training set contains images of 1032 males and 1903 females, and is further randomly split into training and validation sets in a 4:1 ratio. There is no overlap between the training and validation sets. The testing set contains images of 445 males and 809 females.
[0136] 2. 3D Human Measurement Feature Extraction: Extract 3D human measurement features based on human pose estimation and 3D human reconstruction methods.
[0137] 3. Two-dimensional anthropometric feature extraction: Two-dimensional anthropometric features are extracted based on human skeleton point detection and human contour detection methods.
[0138] 4. Deep Feature Extraction: An Attention-Enhanced VGG model is proposed for extracting deep features, such as... Figure 2 As shown.
[0139] 5. Model Training and Selection: Four regression methods—SVR, KRR, DTR, and GRR—were used to train the model on the training set. Next, the test set images were input into the four established models for BMI estimation. The model performance was verified using evaluation metrics, and the model with the best performance was selected to determine the final BMI assessment method. The comparison results of the four regression methods are shown in Table 1. As can be seen from Table 1, in terms of mean absolute error (MAE), the KRR regression method performed best in the three cases of males, females, and the overall population, followed by the SVR regression method. KRR outperformed the DTR and GRR regression methods in both MAE and MRR. Therefore, KRR was selected as the regression method.
[0140] Table 1
[0141]
[0142] 4. Performance Comparison of Different Deep Learning Models: Performance comparisons of deep learning models were performed on the same dataset. The results are as follows: Figure 3 As shown. From Figure 3 As can be seen, compared with other deep network models, the proposed Attention-Enhanced VGG model exhibits the best performance in both MAE and MAPE metrics. This result demonstrates that by introducing an attention mechanism, the proposed model can more effectively capture the intrinsic features of the data, thereby achieving higher accuracy.
[0143] 5. Performance Comparison of Different Attention Modules: Performance comparisons of different attention modules were performed on the same dataset. The results are as follows: Figure 4 As shown. From Figure 4 As can be seen, compared to other attention modules, the MobileViTAttention module in this paper performs best in terms of MAE and MAPE, demonstrating optimal performance. This result shows that the MobileViTAttention module is more effective in capturing data features, thereby achieving lower MAE and MAPE values and further improving the model's prediction accuracy.
[0144] 6. Performance Comparison of Feature Extraction Quantity at Different Depths: Performance comparisons were performed on the same dataset for feature extraction quantities at different depths. The results are as follows: Figure 5 As shown. From Figure 5 It can be seen that as the number of deep features increases, the overall trend of MAE and MAPE is a decrease from 5 to 10, an increase from 10 to 20, a decrease from 20 to 25, and then an increase again after 25. Simultaneously, the total number of network parameters also increases. Considering this trade-off, this paper sets the number of deep features to 10 as the final setting.
[0145] 7. Validation of Multi-Dimensional Feature Extraction: Validation experiments were conducted on multi-dimensional feature extraction, including 2D anthropometry features (2DFs), 3D anthropometry features (3DFs), and depth features (DFs), using the same dataset. The experimental results are shown in Table 2. As can be seen from Table 2, the MAE and MAPE values of the proposed multi-dimensional feature extraction are significantly higher than those of other feature extraction methods. This indicates that multi-dimensional feature extraction can effectively improve prediction accuracy, provide more comprehensive information, and ensure better model performance.
[0146] Table 2
[0147]
[0148] The above detailed description is a specific description of one of the feasible embodiments of the present invention. This embodiment is not intended to limit the patent scope of the present invention. All equivalent implementations or modifications that do not depart from the present invention should be included within the scope of the technical solution of the present invention.
Claims
1. A BMI estimation method based on multi-dimension feature extraction, characterized in that, Specifically comprising the following steps: Step 1, three-dimensional anthropometric feature extraction: based on human pose estimation and three-dimensional human body reconstruction method to extract three-dimensional anthropometric features; Step 2, two-dimensional anthropometric feature extraction: based on human skeleton point detection and human contour detection method to extract two-dimensional anthropometric features; Step 3, depth feature extraction: a deep network model with attention mechanism is proposed to extract depth features; Step 4, model training and selection: connect three-dimensional anthropometric features, two-dimensional anthropometric features and depth features, input regressor; adopt SVR, KRR, DTR and GRR regression method to train the model of training set; input test set image into the established four kinds of regression methods to estimate BMI, and verify the performance of the model through evaluation index, and select the regression method with the best performance; wherein, SVR is support vector regression; KRR is kernel ridge regression; DTR is decision tree regression; GPR is Gaussian process regression.
2. The method of BMI estimation based on multi-dimension feature extraction according to claim 1, wherein, The step 1 specifically comprises the following steps: Step 1.1, data preprocessing is carried out by human pose estimation OpenPose method, so as to obtain human pose information; Step 1.2, human three-dimensional model SMPL-X model is used for human three-dimensional reconstruction, so as to obtain body shape, body offset and posture parameters; Step 1.3, three-dimensional anthropometric features waist circumference, hip circumference and waist circumference to hip circumference ratio are calculated by using three-dimensional shape regression Shapy method.
3. The method of BMI estimation based on multi-dimension feature extraction as claimed in claim 1, wherein, The step 2 specifically comprises the following steps: Step 2.1, obtain skeleton point information by human skeleton point detection method; Step 2.2, human contour detection is carried out; Step 2.3, two-dimensional anthropometric features waist circumference to thigh width ratio, waist circumference to hip circumference ratio, waist circumference to head width ratio, hip width to head width ratio, body area between waist and hip, euclidean distance from nose to knee to waist width ratio, waist width to shoulder width ratio and euclidean distance from nose to knee to hip circumference ratio are calculated.
4. The BMI estimation method based on multi-dimensional feature extraction according to claim 1, wherein, The step 3 specifically comprises the following steps: A deep network model with attention mechanism is proposed to extract depth features; the deep learning model is based on deep convolutional neural network model VGG model, and a lightweight visual MobileViT attention mechanism is added, so as to help the network focus more on the important areas in the image, and improve the extraction ability of the network to key features.
5. The method of claim 2, wherein the method is based on multi-dimensional feature extraction. In step 1.3, the specific steps of three-dimensional anthropometric feature calculation method are as follows: Step 1.31, waist circumference WC is calculated, and the calculation formula is as follows: where, where i and j are two points in the convex hull, β is the SMPL-X shape parameter; (x i , y i , z i ) are the waist landmark point coordinates, t i is the body triangle index; Step 1.32, calculate the hip circumference HC; given the coordinates (h i , m i , n i ) of the hip landmark points and the body triangle index t i ; calculate the hip circumference according to formula (1); Step 1.33, waist circumference to hip circumference ratio WCHCR is calculated; according to the circumference of waist and hip, WCHCR is calculated by the following formula:
6. The method of claim 3, wherein the method further comprises: In step 2.3, the specific steps of two-dimensional anthropometric feature calculation method are as follows: Step 2.31, waist width to shoulder width ratio WSR is calculated; in the case of similar BMI, some people have long shoulder width or short waist width, and WSR is used to represent the difference; according to the width of waist and shoulder, the calculation formula is as follows: where D w is the waist width, D s is the shoulder width; the longitudinal coordinate of the waist is given, and the waist width is calculated as follows: Where s.t. is a constraint condition, l is a linear function, b is an intercept; x1 and x2 are two variables within the numerical range of the constraint condition cw x is the horizontal coordinate of the center of the waist, and y is the vertical coordinate of the center of the waist cs x is the horizontal coordinate of the center of the waist, and y is the vertical coordinate of the center of the waist where x rs is the horizontal coordinate of the right shoulder, x ls is the horizontal coordinate of the left shoulder; y w in equation (4) is the vertical coordinate of the waist, which is calculated as follows: where y cs is the longitudinal coordinate of the center of the shoulder; y h is the longitudinal coordinate of the hip, calculated as follows: where y lh is the longitudinal coordinate of the left hip, y rh is the longitudinal coordinate of the right hip; and y cs is the longitudinal coordinate of the center shoulder, which is calculated as follows: where y rs is the ordinate of the right shoulder, y ls is the ordinate of the left shoulder; D s is calculated as follows: Step 2.32, calculating the waist-to-thigh ratio WTR; the distance from the left thigh boundary to the left thigh center plus the distance from the right thigh boundary to the right thigh center is taken as the thigh width, and the thigh width D is obtained by referring to formula (4) t ; according to the widths of the waist and the thigh, the WTR is calculated by the following formula: Step 2.33, calculating the waist-to-hip ratio WHpR; given the coordinates of the left and right hip joints, the hip circumference D hp WHpR is calculated in the same way as for equation (4): Step 2.34, calculate the waist-to-head ratio WHdR; given the left and right ear skeleton joint coordinates (x le ,y le ) and (x re ,y re ), the head width D hd is approximated by the Euclidean distance between the two points, calculated as follows: Step 2.35, hip width to head width ratio HpHdR is calculated; according to the width of hip and head, HpHdR is calculated by the following formula: Step 2.36, calculate the body area between the waist and the hips Area; since the area between the waist and the hips is an indicator of thin and obese people, in the case of unknown scale information, the characteristic area is used to represent the information; Area is calculated by the following formula: where pixcel num is the number of pixels in the region, pix x，y is the number of pixels at the (x, y) coordinate point; Step 2.37, the Euclidean distance from the nose to the knee and the waist width ratio HWR; HWR is calculated by the following formula: where D nk is the Euclidean distance from the nose to the knee, calculated as follows: Wherein, the knee coordinate point is (x k ,y k ), and the calculation formula is as follows: y k = y lk = y rk (21) Wherein, the left knee coordinate point is (x lk y lk ), and the right knee coordinate point is (x rk ,y rk ). Step 2.38, calculate the Euclidean distance from the nose to the knee and the hip ratio H2H; the waist circumference and the thigh width ratio, the waist circumference and the hip circumference ratio, the waist circumference and the head width ratio, the hip width and the head width ratio, the body area between the waist and the hips, and the waist width and shoulder width ratio These two-dimensional anthropometric features only contain human body transverse feature ratio and do not contain human body longitudinal information; the Euclidean distance from the nose to the knee and the waist width ratio two-dimensional anthropometric feature contains human body longitudinal information, but for people with similar waist width and large differences in hip circumference, the Euclidean distance from the nose to the knee and the waist width ratio two-dimensional anthropometric feature cannot be well distinguished; H2H is proposed to represent the aspect ratio; The calculation formula is as follows: