A student nutrition and health assessment method and system based on facial expression and body feature analysis

Abstract

The present application relates to the technical field of student nutrition and health assessment, and particularly relates to a student nutrition and health assessment method and system based on facial expression and body posture feature analysis.The present application collects color images and depth point cloud data of the student to be measured, generates a high-precision three-dimensional human body model through registration and fusion, uses anatomic bone landmarks as absolute reference, respectively adopts bone landmark layering relative offset quantification method, epidermis pigment distribution statistics method, differential geometry curve curvature change rate analysis method and other exclusive algorithms, calculates five core indexes, converts the five core indexes into specific values, performs multi-dimensional analysis based on the natural development law of nutrition imbalance from shallow to deep, and generates a graded personalized nutrition intervention scheme, realizes non-invasive, rapid and accurate student nutrition and health full-dimension assessment, can accurately determine the type, development stage and severity of nutrition imbalance, and can be widely applied to large-scale normal student nutrition and health screening and precise intervention in schools.

Description

Technical Field

[0001] This invention relates to the field of student nutrition and health assessment technology, specifically to a student nutrition and health assessment method and system based on facial expression and body posture analysis. Background Technology

[0002] In recent years, the nutritional health status of students in my country has improved significantly. However, the structural problem of overnutrition and malnutrition coexisting remains prominent. Nutritional imbalance not only directly affects students' growth and development, immunity and learning ability, but also significantly increases the risk of developing chronic diseases such as cardiovascular disease and diabetes in adulthood. It has become an important public health problem that restricts the healthy development of Chinese youth. Currently, student nutrition and health assessments mainly rely on three traditional methods: questionnaires, physical measurements, and biochemical tests. Among these, questionnaires heavily depend on the subjective recollections of students and parents, and are prone to underreporting and misreporting, making it difficult to guarantee the accuracy of the assessment. Physical measurements can only obtain basic macroscopic indicators such as height, weight, and BMI, and cannot reflect deeper nutritional information such as body composition distribution, muscle development quality, and bone health status, making it difficult to detect early hidden nutritional imbalances. Although biochemical tests have high accuracy, they are invasive, costly, and time-consuming, and require professional medical personnel to operate, which completely fails to meet the actual needs of schools for large-scale, routine screening of all students. With the rapid development of computer vision and 3D reconstruction technologies, image-based non-contact nutritional health assessment methods have gradually become a research hotspot. However, existing technologies still have significant shortcomings: First, the measurement accuracy is low and the assessment dimensions are singular and fragmented. Most are based on two-dimensional image acquisition, which is significantly affected by lighting, shooting angle, and posture changes. Moreover, they only analyze facial surface features or independently detect postural abnormalities such as scoliosis, failing to establish a correlation between early facial nutritional signals and mid-to-late-stage organic changes in posture. They cannot cover the entire development cycle of nutritional imbalance from functional abnormalities to structural damage, making it difficult to achieve true early warning. Second, there is a lack of scientifically quantifiable assessment logic and unified judgment criteria, making it impossible to accurately determine the development stage of nutritional imbalance or establish a scientific mechanism for distinguishing between primary and secondary nutritional problems. When multiple indicators are abnormal at the same time, the dominant nutritional problem cannot be identified, resulting in intervention plans that lack specificity and clear priorities, seriously affecting the effectiveness of nutritional improvement.

[0003] To address the aforementioned shortcomings, a technical solution is provided. Summary of the Invention

[0004] To address the aforementioned shortcomings of existing technologies, this invention provides a student nutrition and health assessment method and system based on facial expression and body posture analysis. This method effectively solves the problems of insufficient accuracy of traditional assessment methods, which cannot meet the needs of large-scale routine screening, and the single dimension of existing non-contact assessment methods, which cannot cover the entire development cycle of nutritional imbalance and lack scientific quantitative assessment logic.

[0005] To achieve the above objectives, the present invention can be implemented through the following technical solutions: This invention provides a method for assessing student nutrition and health based on facial expression and body posture analysis, comprising the following steps: Multiple cameras were deployed to simultaneously collect color images and depth point cloud data of the students being tested. After registration, fusion and preprocessing, a three-dimensional human body model was generated. A three-dimensional human face model is extracted from a three-dimensional human body model, and a circular or arbitrary-shaped periorbital region is delineated as the facial test area with the geometric center of the bony orbit as the origin and a preset length as the radius. The relative offset quantification method based on bony benchmarks combined with the statistical method of epidermal pigment distribution was used to quantify the degree of fat bulging and skin color change in the facial test area, and the fat bulging index and skin color change index were obtained. A three-dimensional body model was extracted from a three-dimensional human body model. Differential geometric curve curvature change rate analysis method, spatial cubic spline curve fitting and curvature integral method, and spatial plane fitting and normal vector angle calculation method were used to calculate the upper arm muscle tightness index, the spine coronal plane curvature index, and the pelvic anterior and posterior tilt angle index. The five indices mentioned above are converted into specific values. Nutritional health status is analyzed based on these specific values, and corresponding nutritional intervention plans are generated.

[0006] Preferably, the specific steps for quantifying the degree of fat protrusion in the facial test area using the bony benchmark layered relative offset quantification method are as follows: Using the bony orbital contour as an anatomical benchmark, the bony orbital benchmark region is delineated, and the total three-dimensional surface area of ​​the bony orbital benchmark region is calculated. The area where the orbital fat pad contour extends beyond the bony orbital reference area and protrudes outward is defined as the bulging feature area. The three-dimensional surface area of ​​the bulging feature area is calculated. Starting from any point on the bony orbital reference boundary, the distance from that point to the boundary of the bulging feature area is measured along the outer normal direction. The distance from all boundary points to the boundary of the bulging area along the outer normal direction is taken, and the maximum value is recorded as the maximum radial outward expansion distance. Then, by introducing the maximum theoretical extension threshold of the orbital bony structure, the fat bulging index is obtained by fusing the two-dimensional parameters of the bulging area ratio and the maximum outward expansion distance.

[0007] Preferably, the specific steps for quantifying the degree of skin color change in the test area of ​​the face using the epidermal pigment distribution statistical method are as follows: The facial area to be tested is divided into several uniform sampling units. For each sampling unit, the spatial coordinates of the epidermis and the spatial coordinates of the bony orbit are extracted, and the vertical distance between the two points in the vertical axis direction is calculated as the epidermal-bone relative offset of the sampling unit. Then, an individual normalization factor is introduced to normalize the relative epidermal-bone offset of each sampling unit; Simultaneously, epidermal color information of each sampling unit is collected, color components are extracted in the standard color space, and the average skin color of the non-lesion area of ​​the face of the same individual is used as the individual's baseline skin color to calculate the color difference value of each sampling unit. The normalized relative offset between the epidermis and bone was used as the spatial morphological weight, and weighted and fused with the color difference value to calculate the skin color change index.

[0008] Preferably, the specific steps for calculating the upper arm muscle firmness index using the differential geometric curve curvature change rate analysis method are as follows: Based on the acromion and elbow points among the bony landmarks, the line connecting the acromion and elbow points is determined as the upper arm midline. The midpoint of this line is taken, and a plane perpendicular to the upper arm midline is drawn through the midpoint. The intersection of this plane and the three-dimensional body model is the upper arm cross-sectional contour curve. Obtain the geometric centroid of the upper arm cross-sectional profile curve, and convert the upper arm cross-sectional profile curve into polar coordinate parameter form with the geometric centroid as the pole. Based on the obtained upper arm cross-sectional profile curve in polar coordinate form, the curvature of the upper arm cross-sectional profile curve at each polar angle is determined. The first derivative of curvature with respect to the polar angle is then calculated using the central difference method, and the first derivative is integrated over the entire circumference to obtain the upper arm muscle firmness index.

[0009] Preferably, the specific steps for calculating the coronal curvature index of the spine using spatial cubic spline curve fitting and curvature integral methods are as follows: Key points of 14 vertebral bodies of the spine were extracted from the 7th cervical vertebra, the 12th thoracic vertebra, and the 5th lumbar vertebra to form a spatial point set; Then project the spatial point set onto the XY plane of the body coordinate system to obtain the projected point set; The natural cubic spline interpolation algorithm is used to fit the projection point set to obtain the coronal plane curve of the spine. The natural cubic spline interpolation algorithm refers to constructing a cubic polynomial between each adjacent projection point so that the fitted curve passes through all projection points accurately, and the first and second derivatives are continuous at the internal nodes, while the second derivative is zero at the two endpoints. Calculate the curvature at any point on the coronal plane curve of the spine, and integrate the curvature over the entire curve length to obtain the coronal plane curvature index of the spine.

[0010] Preferably, the specific steps for calculating the pelvic tilt angle index using the spatial plane fitting and normal vector angle calculation method are as follows: Three bony landmarks—the left anterior superior iliac spine, the right anterior superior iliac spine, and the upper edge of the pubic symphysis—were selected as feature points of the pelvic plane. The spatial coordinates of these feature points were extracted from the three-dimensional body model, and the pelvic plane was fitted using the least squares method. The angle between the pelvic plane normal vector and the vertical reference vector was calculated and used as the pelvic anterior-posterior tilt angle index.

[0011] Preferably, the specific steps for converting the five indices into specific values ​​include: A health sample reference database covering students aged 6 to 18 was established. The database contains 15,000 healthy student samples with no chronic diseases, no skeletal development abnormalities, and whose nutritional indicators meet the standards. The mean and standard deviation of five indices were statistically analyzed by age and gender as standardized benchmark parameters. For the five indices of the tested students, the standardized baseline parameters of the corresponding groups are retrieved in combination with their actual age and gender, and the standardization transformation is completed. That is, the original index value is subtracted from the mean of the corresponding group and then divided by the standard deviation of the corresponding group to obtain the specific values ​​of the five indices.

[0012] Preferably, the specific process for nutritional health status analysis based on specific values ​​is as follows: Based on the absolute value of the specific value, the degree of abnormality is divided into three levels, including mild abnormality, moderate abnormality and severe abnormality. According to the abnormal combination pattern of the five indices, and combined with the law of nutritional imbalance from superficial to deep and from short-term symptoms to long-term organic changes, nutritional imbalance is divided into four stages: latent period, early stage, middle stage and late stage, and a corresponding warning level is matched for each stage. The dominant index priority method is used to determine the type of nutritional problem. Specifically, the index with the shortest response period is selected from all abnormal indices as the first dominant index. If two or more indices have the same response period, the index with the higher degree of abnormality is selected as the first dominant index. The nutrient corresponding to the first dominant index is the primary nutritional problem, and the nutrients corresponding to the other abnormal indices are the secondary nutritional problems. Based on the absolute values ​​of the five index specific values ​​and combined with the corresponding time-series weights, a comprehensive nutrition and health score is calculated, and the nutrition and health status is classified according to the comprehensive nutrition and health score. Based on the type, stage of development, and severity of nutritional problems, quantitative intervention priorities are generated and sorted from high to low priority, with the highest priority nutrients being the primary intervention targets.

[0013] Preferably, the determination criteria for the four stages are as follows: Incubation period: Only one of the skin color change index and fat bulging index was slightly abnormal, while the other four indices were normal; Early stage: Meeting any of the following conditions: ① One of the skin color change index and fat bulging index is moderately abnormal; ② The skin color change index or fat bulging index is mildly abnormal, and the upper arm muscle tightness index is also mildly abnormal. Mid-term: Meets any of the following conditions: ① Upper arm muscle tightness index is moderately abnormal; ② Upper arm muscle tightness index is mildly abnormal, and at the same time, pelvic anteroposterior tilt angle index is mildly abnormal. Late stage: Meeting any of the following conditions: ① The coronal curvature index of the spine is abnormal to any degree; ② Both the anteroposterior tilt angle index of the pelvis and the coronal curvature index of the spine are abnormal.

[0014] Preferably, a student nutrition and health assessment system based on facial expression and body posture analysis includes: The image acquisition and 3D reconstruction module simultaneously acquires color images and depth point cloud data of the students being tested by deploying multiple cameras, and generates a 3D human body model after registration, fusion and preprocessing. The facial nutrition feature quantitative analysis module extracts a three-dimensional face model from a three-dimensional human body model, and delineates a circular or arbitrary-shaped periorbital region as the facial test area with the geometric center of the bony orbit as the origin and a preset length as the radius. The relative offset quantification method based on bony benchmarks combined with the statistical method of epidermal pigment distribution was used to quantify the degree of fat bulging and skin color change in the facial test area, and the fat bulging index and skin color change index were obtained. The body nutrition characteristics quantitative analysis module extracts a three-dimensional body model from a three-dimensional human body model and calculates the upper arm muscle tightness index, the spine coronal plane curvature index, and the pelvic anterior and posterior tilt angle index by using differential geometric curve curvature change rate analysis method, spatial cubic spline curve fitting and curvature integral method, and spatial plane fitting and normal vector angle calculation method, respectively. The personalized intervention plan generation module converts the above five indices into specific values, analyzes nutritional health status based on the specific values, and generates corresponding nutritional intervention plans.

[0015] The technical solution provided by this invention has the following advantages compared with the known prior art: 1. This invention uses multiple cameras to simultaneously and non-contactly acquire color images and depth point cloud data. The entire acquisition process takes only 3-5 seconds, and there is no need for fasting, blood drawing, or complex sample processing, making it highly acceptable to students. No professional medical personnel are required to operate it, and ordinary teachers can complete the entire process after simple training, which can meet the school's need for large-scale, routine screening of all students once a semester.

[0016] 2. This invention generates a high-precision three-dimensional human body model by registering and fusing color images and depth point cloud data. All feature extraction and quantification calculations are based on anatomical bony landmarks as absolute benchmarks. Facial fat protrusion is based on the bony orbital contour, upper arm muscle tightness is based on the acromion and elbow points, and pelvic tilt angle is based on the anterior superior iliac spine and pubic symphysis. This fundamentally eliminates the interference of external factors such as individual body shape differences, shooting angle, and lighting. At the same time, it proposes a number of exclusive quantification algorithms, such as the relative offset quantification method of bony benchmark layering and the differential geometric curve curvature change rate analysis, to achieve accurate digital measurement of various features, reduce measurement errors, and make the evaluation results objective and reliable.

[0017] 3. This invention constructs five core indices: skin color change, fat bulging, upper arm muscle firmness, spinal coronal curvature, and pelvic tilt angle. Among them, skin color change and fat bulging are sensitive to short-term nutritional imbalances and can detect abnormalities within 1-3 months after the onset of nutritional imbalance; upper arm muscle firmness reflects mid-term nutritional status, with a response period of approximately 3-6 months; spinal coronal curvature and pelvic tilt angle represent organic changes caused by long-term severe nutritional imbalances, with a response period exceeding 1 year. This assessment index comprehensively covers the entire development cycle of nutritional imbalance from latent functional abnormalities to late-stage organic damage, enabling precise determination of the specific developmental stage of nutritional imbalance in the tested students and providing clear time points for subsequent graded intervention.

[0018] 4. This invention, by establishing a four-stage judgment standard and a three-level abnormality classification system based on the natural development law of nutritional imbalance, can accurately determine the development stage and severity of nutritional imbalance and match the corresponding warning level. Simultaneously, it proposes a dominant index priority method, distinguishing primary and secondary nutritional problems based on the response cycle and abnormality degree of different indices: the index with the shortest response cycle is selected as the first dominant index, corresponding to the primary nutritional problem; if the response cycles are the same, the index with a higher degree of abnormality is selected. This effectively solves the problem of judgment confusion when multiple indicators are abnormal, accurately locates the core nutritional problem, provides a scientific basis for generating targeted and clearly prioritized personalized intervention plans, and significantly improves the effectiveness of nutritional intervention. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 This is a flowchart illustrating the overall process of the present invention.

[0021] Figure 2 This is an overall module block diagram of the present invention. Detailed Implementation

[0022] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of 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, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0023] like Figure 1 As shown, a student nutrition and health assessment method based on facial expression and body posture analysis includes the following steps: Step 1: Have the student being tested enter the data collection chamber (the chamber measures 2.0m × 1.5m × 2.2m and has a diffuse white coating inside to avoid strong light reflection). The student should stand in the center of the marked position on the ground. Before the data collection, the student should wear a tight-fitting school uniform, remove glasses, jewelry, and other obstructions, and maintain a standard standing posture (such as feet together, arms hanging naturally, and eyes looking forward).

[0024] Four industrial-grade binocular depth cameras, each with a resolution of 2048×1536, were used. They were positioned directly in front of, to the left of, to the right of, and directly behind the student being tested. Each camera was 1.5 meters away from the student, and the center of the lens was 1.2 meters above the student. All cameras were triggered synchronously to simultaneously acquire color images and depth point cloud data.

[0025] The intrinsic parameter matrix of each camera is obtained in advance using Zhang Zhengyou's calibration method. and extrinsic parameter matrix , pixel coordinates Convert to three-dimensional coordinates in a unified world coordinate system The conversion formula is: ,in, For the camera intrinsic parameter matrix, and for and Directional focal length, and Principal point coordinates It is a 3×3 rotation matrix. It is a 3×1 translation vector. This is the depth value from the measured point to the camera plane.

[0026] Four initial point clouds obtained from four cameras (front, left, right, and rear) are used for pairwise registration and overall fusion using the Iterative Closest Point (ICP) algorithm to eliminate redundant data in overlapping areas and generate a complete 3D human point cloud model. The point cloud density of the fused model is no less than 500 points / cm² to ensure the accuracy of surface details.

[0027] The fused point cloud is denoised, outliers are removed, and holes are repaired. A triangular mesh model is generated using the BallPivoting algorithm. The mesh model is then smoothed and textured (based on synchronously acquired color images). Finally, a 3D human body model with geometric and color information is output.

[0028] It should be noted that Zhang Zhengyou's calibration method is a camera parameter calibration method based on a planar checkerboard target. This method extracts the corner pixel coordinates by photographing the checkerboard calibration board from multiple different angles, and uses the known physical dimensions of the checkerboard to establish a homography mapping relationship between image points and spatial points. Through a combination of linear solution and nonlinear optimization, it calculates the camera's intrinsic parameter matrix (focal length, principal point, distortion coefficients, etc.) and extrinsic parameter matrix (rotation matrix, translation vector). This method combines high calibration accuracy with ease of operation and is widely used in the joint calibration of binocular depth cameras. The iterative nearest point algorithm is a classic algorithm for accurate registration between two point clouds. Its core idea is: in each iteration, establish a correspondence between each point in one point cloud and the point in the other point cloud with the nearest Euclidean distance, and then calculate a minimum distance by minimizing the sum of squared average distances between corresponding points. The optimal rigid body transformation (rotation matrix and translation vector) is obtained, and the point cloud position is updated. The above process is repeated until the error is less than the set threshold or the upper limit of the number of iterations. In the human body acquisition system, ICP is used to register the point clouds acquired by four cameras at different angles to a unified coordinate system to form a seamlessly integrated 3D model. The rolling sphere algorithm is a reconstruction method that directly generates triangular mesh surfaces from point cloud data. Its basic principle is: a sphere with a given radius is "rolled" on the surface of the point cloud. When the sphere contacts three points in the point cloud at the same time and the three points are not occluded by existing triangular patches, the three points are connected to form a triangle. By continuously rolling the sphere, generating triangular patches and expanding the boundary, a complete triangular mesh model covering the surface of the point cloud is finally obtained. This method is suitable for point cloud data with high density, continuous surface and no obvious holes, and can better preserve the detailed morphology of the human body surface.

[0029] Step 2: Extract the 3D face model from the 3D human body model. Based on this, delineate a circular or arbitrarily shaped periorbital region as the facial test area with the geometric center of the bony orbit as the origin and a preset length as the radius. By using the relative offset quantification method of bony reference layering, supplemented by the epidermal pigment distribution statistical method, the degree of fat bulging and skin color change in the facial test area are quantified and calculated to obtain the fat bulging index and skin color change index.

[0030] The specific steps for quantifying the degree of fat protrusion in the facial test area using the bony benchmark layered relative offset quantification method include: Using the bony orbital contour as an anatomical reference, the bony orbital reference region is delineated, and the total three-dimensional surface area of ​​this bony orbital reference region is calculated and denoted as . .

[0031] The region where the orbital fat pad protrudes outward beyond the bony orbital reference area is defined as the bulging feature region. The three-dimensional surface area of ​​the bulging feature region is calculated and denoted as . Starting from any point on the baseline boundary of the bony orbit, measure the distance from that point along the outer normal direction (or along the ray direction originating from the geometric center of the orbit) to the boundary of the bulging feature area. Take the distances from all boundary points to the boundary of the bulging area along their outer normal direction, and record the maximum value as the maximum radial outward expansion distance. .

[0032] To eliminate quantitative bias caused by differences in orbital size among individuals, a maximum theoretical extrapolation threshold for orbital bony structures is introduced. Complete data normalization processing and further configure area weighting coefficients. Radial distance weighting coefficient The two coefficients satisfy the normalization constraint. By integrating the proportion of bulging area and the maximum outward expansion distance, a composite quantitative index that can comprehensively characterize the range and amplitude of fat protrusion is constructed, thus obtaining the fat bulging index. The specific calculation formula is as follows: .

[0033] The specific steps for quantifying the degree of skin color change in the test area of ​​the face using epidermal pigment distribution statistical methods include: The facial area to be tested is divided into several uniform sampling units, and the sampling unit number is denoted as follows: .

[0034] For each sampling unit, the spatial coordinates of the epidermis and the spatial coordinates of the bony orbit are extracted, and the vertical distance between the two points along the vertical axis (Z-axis) is calculated as the epidermal-bone relative offset of that sampling unit. The calculation formula is as follows: In the formula, The vertical component of the skin space coordinates. For the vertical axis component of the spatial coordinates of the bony orbit, For the first The relative offset of the epidermis to the bony structure of each sampling unit.

[0035] It should be noted that the epidermal spatial coordinates refer to the three-dimensional spatial coordinates of each sampling unit corresponding to the epidermal position, while the bony orbital spatial coordinates refer to the spatial coordinates of each sampling unit on the bony orbit.

[0036] To eliminate the influence of differences in the depth dimensions of orbital bony structures among individuals, an individual normalization factor was introduced. (For example, the theoretical maximum depth or reference depth of the orbital bony structure on this side in the vertical direction), the relative offset of the epidermis to the bony structure of each sampling unit is normalized: .

[0037] Simultaneously, the skin color information of each sampling unit is collected, and color components are extracted in a standard color space (such as the CIELAB color space). The average skin color of the non-lesion areas of the face (such as the center of the forehead or the flat area of ​​the cheek) of the same individual was used as the individual's baseline skin color. Calculate the color difference value of each sampling unit. The calculation formula is as follows: .

[0038] Normalized epidermal-bony relative offset As a spatial form weight, and related to color difference value Perform weighted fusion calculation of skin color variation index The calculation formula is as follows: In the formula, This represents the total number of sampling units.

[0039] Step 3: Extract the three-dimensional body model from the three-dimensional human body model. Based on the differential geometric curve curvature change rate analysis method, the spatial cubic spline curve fitting and curvature integral method, and the spatial plane fitting and normal vector angle calculation method, calculate the upper arm muscle tightness index, the spinal coronal plane curvature index, and the pelvic anterior and posterior tilt angle index, respectively.

[0040] in: The specific steps for calculating the upper arm muscle firmness index based on the differential geometric curve curvature change rate analysis method are as follows: Based on the acromion and elbow apex points among the bony landmarks, the line connecting the acromion and elbow apex points is defined as the midline of the upper arm. The midpoint of this line is then taken, and a plane perpendicular to the upper arm midline is drawn through this midpoint. The intersection of this plane and the three-dimensional torso model is the cross-sectional contour curve of the upper arm. . To eliminate the influence of coordinate rotation on the calculation results, the cross-sectional profile curve of the upper arm is obtained. The geometric centroid is used as the pole, and the cross-sectional profile curve of the upper arm is converted into polar coordinate parametric form according to the following formula: ,in, Polar angle, The polar diameter (i.e., the cross-sectional profile curve of the upper arm at the polar angle) (Radial distance in the direction).

[0041] Based on the obtained upper arm cross-sectional profile curve in polar coordinate form, calculate the curvature of the upper arm cross-sectional profile curve at each polar angle. For a plane curve, the formula for calculating curvature is: ,in, and These are the first and second derivatives of the polar radius with respect to the polar angle, respectively.

[0042] Furthermore, the curvature is calculated using the central difference method. polar angle first derivative The formula for calculating the central difference is: ,in, For the first The polar angle corresponding to each sampling point and These are the polar angles of their adjacent sampling points before and after.

[0043] The upper arm muscle firmness index is obtained by integrating over the entire circumference. The calculation formula is as follows: The upper arm muscle firmness index characterizes the degree of drastic change in the curvature of the upper arm cross-sectional contour curve along the circumference. The gentler the curvature change (i.e., the smaller the upper arm muscle firmness index), the closer the upper arm cross-sectional contour is to a circle, and the fuller and smoother the muscle contour, reflecting higher muscle firmness. Conversely, the more drastic the curvature change (i.e., the larger the upper arm muscle firmness index), the more uneven or flat areas there are in the muscle contour, reflecting lower muscle firmness.

[0044] The specific steps for calculating the coronal curvature index of the spine based on spatial cubic spline curve fitting and curvature integral method are as follows: Key points of 14 vertebral bodies were extracted from the 7th cervical vertebra, the 12th thoracic vertebra, and the 5th lumbar vertebra to form a spatial point set. .

[0045] The spatial point set is then projected onto the XY plane (coronal plane) of the normalized body coordinate system to obtain the projected point set. .

[0046] The natural cubic spline interpolation algorithm is used to fit the projection point set to obtain the coronal curve of the spine. ,in To follow the curve from arrive The arc length parameter, , Let be the total length of the curve. The natural cubic spline interpolation algorithm refers to constructing a cubic polynomial between each adjacent projection point, such that the fitted curve passes through all projection points precisely, and the first and second derivatives are continuous at the internal nodes, while the second derivative is zero at the two endpoints.

[0047] Calculate the curvature at any point on the coronal plane curve of the spine using the plane curve curvature formula. Its curvature is calculated using the following formula: ,in, and The first derivative, and It is the second derivative.

[0048] Integrating the curvature over the entire curve length yields the coronal curvature index of the spine. The calculation formula is as follows: The coronal curvature index of the spine characterizes the overall curvature of the spine in the coronal plane. The larger the coronal curvature index, the greater the curvature of the spine in the coronal plane (e.g., the more severe the scoliosis); the smaller the coronal curvature index, the more the spine in the coronal plane tends to be straight.

[0049] The specific steps for calculating the pelvic tilt angle index based on spatial plane fitting and normal vector angle calculation are as follows: Three bony landmarks—the left anterior superior iliac spine, the right anterior superior iliac spine, and the upper border of the pubic symphysis—were selected as feature points of the pelvic plane. The spatial coordinates of these feature points were extracted from the 3D body model, and the pelvic plane was fitted using the least squares method. The angle between the pelvic plane normal vector and the vertical reference vector was calculated and used as the pelvic anteroposterior tilt angle index. The calculation formula is as follows: ,in, The normal vector of the pelvic plane. The vertical reference vector is The dot product of two vectors. The anteroposterior tilt index is the magnitude of the vector. It represents the degree of pelvic tilt in the sagittal plane (anteroposterior direction). The larger the anteroposterior tilt index, the more severe the anteroposterior tilt of the pelvis. The smaller the anteroposterior tilt index, the closer the pelvis is to a horizontal position.

[0050] Step 4: Convert the fat bulge index, skin color change index, upper arm muscle firmness index, spinal coronal curvature index, and pelvic anteroposterior tilt angle index into their corresponding specific values, and classify the abnormalities according to the absolute values ​​of the specific values. The specific analysis steps are as follows: A health sample reference database covering students aged 6 to 18 was pre-established. The database contains 15,000 healthy student samples with no chronic diseases, no skeletal development abnormalities, and whose nutritional indicators meet the standards. The mean and standard deviation of five indices were statistically analyzed by age and gender as standardized benchmark parameters. For the five indices of the tested students, the standardized baseline parameters of the corresponding groups were retrieved in combination with their actual age and gender. The standardization transformation was completed (the original index value was subtracted from the mean of the corresponding group and then divided by the standard deviation of the corresponding group) to obtain the specific values ​​of the five indices. This eliminated the assessment bias caused by differences in age, gender and individual development. The calculated specific values ​​were truncated to limit the range of specific values ​​to [-4,4] to avoid the impact of extreme outliers on subsequent assessments. Abnormalities are classified into three levels based on the absolute value of their specificity values: Normal: The absolute value of the specific value is <1; Mild abnormality: 1 ≤ absolute value of specificity < 2; Moderate abnormality: 2 ≤ absolute value of specificity < 3; Severe abnormality: The absolute value of the specific value is ≥3.

[0051] Based on the abnormal combination patterns of the five indices, and considering the progression of nutritional imbalance from superficial to deep and from short-term symptoms to long-term organic changes, nutritional imbalance is divided into four stages: latent, early, middle, and late, with corresponding warning levels. If a group of abnormal indices simultaneously meets the conditions for multiple stages, the stage with the higher warning level will be used. Specifically: Incubation period (1-4 weeks of nutritional imbalance), blue alert, criteria for judgment: only one of the skin color change index and fat bulging index is slightly abnormal, and the other four indices are normal; during this stage, nutritional imbalance only affects the body surface and subcutaneous soft tissue, without affecting muscles and bones, which is the best window of intervention.

[0052] Early stage (4-12 weeks of nutritional imbalance), yellow warning, judgment criteria: meeting any of the following conditions: ① One of the skin color change index and fat bulging index is moderately abnormal; ② The skin color change index or fat bulging index is mildly abnormal, and the upper arm muscle firmness index is mildly abnormal; at this stage, nutritional imbalance begins to affect skeletal muscle development.

[0053] Mid-stage (3-6 months of nutritional imbalance), orange alert, judgment criteria: meeting any of the following conditions: ① Upper arm muscle tightness index is moderately abnormal; ② Upper arm muscle tightness index is mildly abnormal, and at the same time, the pelvic anterior-posterior tilt angle index is mildly abnormal; at this stage, nutritional imbalance causes a decrease in muscle strength, which gradually affects pelvic posture.

[0054] Late stage (nutritional imbalance for more than 6 months), red alert, judgment criteria: meeting any of the following conditions: ① The coronal curvature index of the spine is abnormal to any degree; ② Both the pelvic anteroposterior tilt angle index and the coronal curvature index of the spine are abnormal or above; at this stage, nutritional imbalance has caused structural changes in the bones, and there is an irreversible risk of development.

[0055] The dominant index priority method is used to determine the type of nutritional problem. Specifically, from all abnormal indices, the index with the shortest response period is selected as the first dominant index. If two or more indices have the same response period, the index with the higher degree of abnormality is selected as the first dominant index. The nutrient corresponding to the first dominant index is the primary nutritional problem, and the nutrients corresponding to the other abnormal indices are the secondary nutritional problems.

[0056] It should be noted that nutrients may include iron, vitamin B12, zinc, protein, vitamin D, and calcium; The relationship between various indices and nutritional problems is shown in the table below:

[0057] The comprehensive nutrition and health score is calculated based on the absolute values ​​of the five index specific values ​​and their corresponding time-series weights. The specific calculation formula is as follows: ,in, , , , and These are the specific values ​​for the skin color change index, fat bulge index, upper arm muscle firmness index, pelvic anteroposterior tilt angle index, and spinal coronal curvature index.

[0058] At the same time, nutritional health status is classified according to the comprehensive nutritional health score: excellent: ; good: ; Mild abnormality: ; Moderate abnormality: ; Key anomalies: .

[0059] Based on the type, stage of development, and severity of nutritional problems, a quantitative intervention priority is generated, calculated using the following formula: ,in, For the first Intervention priorities for nutrient-like substances For basic nutrient weighting, possible values ​​are: protein 1.2, calcium 1.1, iron 1.0, vitamin D 1.0, vitamin B12 0.9, zinc 0.9, carbohydrates 0.8, and fat 0.8. For the first The absolute value of the specific value of the dominant index corresponding to each nutrient class. The value is the stage coefficient, with values ​​of 1.0 for the incubation period, 1.2 for the early stage, 1.5 for the middle stage, and 2.0 for the late stage.

[0060] The nutrients are ranked from highest to lowest priority for intervention, with the highest priority nutrients being the primary intervention targets.

[0061] Nutritional intervention plans are formulated based on comprehensive nutritional health scores, nutritional health status levels, stages of nutritional imbalance, nutrients requiring intervention, and the resulting problems.

[0062] In a specific implementation case: The subject was an 11-year-old male student. The standardized conversion results of the five indices were as follows: skin color change index Zy1=0.7 (normal), fat protrusion index Zy2=0.5 (normal), upper arm muscle tightness index Zy3=-1.8 (mildly abnormal), pelvic tilt angle index Zy4=1.6 (mildly abnormal), and spinal coronal curvature index Zy5=0.4 (normal). The abnormal indices in this study are the upper arm muscle firmness index and the pelvic anterior-posterior tilt angle index, which meet the criteria for mid-term assessment and correspond to an orange alert, indicating that the nutritional imbalance has lasted for 3 to 6 months. Of the two abnormal indices, the upper arm muscle firmness index was determined to be the primary dominant index; the primary nutritional problem was identified as protein deficiency, and the secondary nutritional problem as calcium deficiency. Substituting the absolute values ​​of each index specific value into the comprehensive nutrition and health score formula: S=100-(0.3×0.7+0.25×0.5+0.2×1.8+0.15×1.6+0.1×0.4)×10=76.2, the comprehensive nutrition and health score is 76.2 points, corresponding to a mild abnormality level; With a stage coefficient of 1.5, the outlier for protein index |Zy3| = 1.8, and the outlier for calcium index |Zy4| = 1.6. Protein priority: 1.2 × 1.8 × 1.5 = 3.24; Calcium priority: 1.1 × 1.6 × 1.5 = 2.64 The priority of all other nutrients was below 1.0, and the final intervention priority was: protein > calcium; Therefore, the student's overall nutritional health score was 76.2, indicating a slightly abnormal overall condition; the nutritional imbalance was in the middle stage and had lasted for 3 to 6 months, triggering an orange alert. The core problem was insufficient protein intake, which led to a decrease in core muscle strength and a slight anterior tilt of the pelvis. If not intervened in time, it will gradually affect bone development in the long term. Intervention recommendations: Prioritize supplementing with high-quality protein, adding 1 egg, 75g of lean meat, and 100g of soy products daily; combine with core strength training, including plank and glute bridge exercises, 20 minutes per session, 4 times per week; supplement with appropriate amounts of calcium, regularly retest body mass index, and track the effects of nutritional improvement.

[0063] like Figure 2 As shown, a student nutrition and health assessment system based on facial expression and body posture feature analysis includes: an image acquisition and three-dimensional reconstruction module, a facial nutrition feature quantitative analysis module, a body posture nutrition feature quantitative analysis module, and a personalized intervention plan generation module. The image acquisition and 3D reconstruction module is used to acquire color images and depth point cloud data of the students being tested. After registration, fusion, and preprocessing, a 3D human body model is generated. The specific implementation process is as follows: Four industrial-grade binocular depth cameras, each with a resolution of 2048×1536, were used. They were positioned directly in front of, to the left of, to the right of, and directly behind the student being tested. Each camera was 1.5 meters away from the student, and the center of the lens was 1.2 meters above the student. All cameras were triggered synchronously to simultaneously acquire color images and depth point cloud data. The intrinsic parameter matrix of each camera is obtained in advance using Zhang Zhengyou's calibration method. and extrinsic parameter matrix , pixel coordinates Convert to three-dimensional coordinates in a unified world coordinate system The conversion formula is: ,in, For the camera intrinsic parameter matrix, and for and Directional focal length, and Principal point coordinates It is a 3×3 rotation matrix. It is a 3×1 translation vector. This is the depth value from the measured point to the camera plane.

[0064] Four initial point clouds obtained from four cameras (front, left, right, and rear) are used for pairwise registration and overall fusion using the Iterative Closest Point (ICP) algorithm to eliminate redundant data in overlapping areas and generate a complete 3D human point cloud model. The point cloud density of the fused model is no less than 500 points / cm² to ensure the accuracy of surface details.

[0065] The fused point cloud is denoised, outliers are removed, and holes are repaired. A triangular mesh model is generated using the BallPivoting algorithm. The mesh model is then smoothed and textured (based on synchronously acquired color images). Finally, a 3D human body model with geometric and color information is output.

[0066] The facial nutritional feature quantification analysis module is used to extract a 3D face model from a 3D human body model. Based on this, a circular or arbitrarily shaped periorbital region is delineated with the geometric center of the bony orbit as the origin and a preset length as the radius as the facial test area. The bony benchmark layered relative offset quantification method, supplemented by the epidermal pigment distribution statistical method, is used to quantify the degree of fat bulging and skin color change in the facial test area, obtaining the fat bulging index and skin color change index. The specific implementation process is as follows: Using the bony orbital contour as an anatomical reference, the bony orbital reference region is delineated, and the total three-dimensional surface area of ​​this bony orbital reference region is calculated and denoted as . .

[0067] The region where the orbital fat pad protrudes outward beyond the bony orbital reference area is defined as the bulging feature region. The three-dimensional surface area of ​​the bulging feature region is calculated and denoted as . Starting from any point on the baseline boundary of the bony orbit, measure the distance from that point along the outer normal direction (or along the ray direction originating from the geometric center of the orbit) to the boundary of the bulging feature area. Take the distances from all boundary points to the boundary of the bulging area along their outer normal direction, and record the maximum value as the maximum radial outward expansion distance. .

[0068] To eliminate quantitative bias caused by differences in orbital size among individuals, a maximum theoretical extrapolation threshold for orbital bony structures is introduced. Complete data normalization processing and further configure area weighting coefficients. Radial distance weighting coefficient The two coefficients satisfy the normalization constraint. By integrating the proportion of bulging area and the maximum outward expansion distance, a composite quantitative index that can comprehensively characterize the range and amplitude of fat protrusion is constructed, thus obtaining the fat bulging index. The specific calculation formula is as follows: .

[0069] The facial area to be tested is divided into several uniform sampling units, and the sampling unit number is denoted as follows: .

[0070] For each sampling unit, the spatial coordinates of the epidermis and the spatial coordinates of the bony orbit are extracted, and the vertical distance between the two points along the vertical axis (Z-axis) is calculated as the epidermal-bone relative offset of that sampling unit. The calculation formula is as follows: In the formula, The vertical component of the skin space coordinates. For the vertical axis component of the spatial coordinates of the bony orbit, For the first The relative offset of the epidermis to the bony structure of each sampling unit.

[0071] It should be noted that the epidermal spatial coordinates refer to the three-dimensional spatial coordinates of each sampling unit corresponding to the epidermal position, while the bony orbital spatial coordinates refer to the spatial coordinates of each sampling unit on the bony orbit.

[0072] To eliminate the influence of differences in the depth dimensions of orbital bony structures among individuals, an individual normalization factor was introduced. (For example, the theoretical maximum depth or reference depth of the orbital bony structure on this side in the vertical direction), the relative offset of the epidermis to the bony structure of each sampling unit is normalized: .

[0073] Simultaneously, the skin color information of each sampling unit is collected, and color components are extracted in a standard color space (such as the CIELAB color space). The average skin color of the non-lesion areas of the face (such as the center of the forehead or the flat area of ​​the cheek) of the same individual was used as the individual's baseline skin color. Calculate the color difference value of each sampling unit. The calculation formula is as follows: .

[0074] Normalized epidermal-bony relative offset As a spatial form weight, and related to color difference value Perform weighted fusion calculation of skin color variation index The calculation formula is as follows: In the formula, This represents the total number of sampling units.

[0075] The body composition nutritional characteristics quantitative analysis module is used to extract a three-dimensional body model from a three-dimensional human body model. Based on the differential geometric curve curvature change rate analysis method, spatial cubic spline curve fitting and curvature integral method, and spatial plane fitting and normal vector angle calculation method, it calculates the upper arm muscle tightness index, the spinal coronal plane curvature index, and the pelvic anterior and posterior tilt angle index, respectively. The specific implementation process is as follows: Based on the acromion and elbow apex points among the bony landmarks, the line connecting the acromion and elbow apex points is defined as the midline of the upper arm. The midpoint of this line is then taken, and a plane perpendicular to the upper arm midline is drawn through this midpoint. The intersection of this plane and the three-dimensional torso model is the cross-sectional contour curve of the upper arm. . To eliminate the influence of coordinate rotation on the calculation results, the cross-sectional profile curve of the upper arm is obtained. The geometric centroid is used as the pole, and the cross-sectional profile curve of the upper arm is converted into polar coordinate parametric form according to the following formula: ,in, Polar angle, The polar diameter (i.e., the cross-sectional profile curve of the upper arm at the polar angle) (Radial distance in the direction).

[0076] Based on the obtained upper arm cross-sectional profile curve in polar coordinate form, calculate the curvature of the upper arm cross-sectional profile curve at each polar angle. For a plane curve, the formula for calculating curvature is: ,in, and These are the first and second derivatives of the polar radius with respect to the polar angle, respectively.

[0077] Furthermore, the curvature is calculated using the central difference method. polar angle first derivative The formula for calculating the central difference is: ,in, For the first The polar angle corresponding to each sampling point and These are the polar angles of their adjacent sampling points before and after.

[0078] The upper arm muscle firmness index is obtained by integrating over the entire circumference. The calculation formula is as follows: .

[0079] Key points of 14 vertebral bodies were extracted from the 7th cervical vertebra, the 12th thoracic vertebra, and the 5th lumbar vertebra to form a spatial point set. .

[0080] The spatial point set is then projected onto the XY plane (coronal plane) of the normalized body coordinate system to obtain the projected point set. .

[0081] The natural cubic spline interpolation algorithm is used to fit the projection point set to obtain the coronal curve of the spine. ,in To follow the curve from arrive The arc length parameter, , Let be the total length of the curve. The natural cubic spline interpolation algorithm refers to constructing a cubic polynomial between each adjacent projection point, such that the fitted curve passes through all projection points precisely, and the first and second derivatives are continuous at the internal nodes, while the second derivative is zero at the two endpoints.

[0082] Calculate the curvature at any point on the coronal plane curve of the spine using the plane curve curvature formula. Its curvature is calculated using the following formula: ,in, and The first derivative, and It is the second derivative.

[0083] Integrating the curvature over the entire curve length yields the coronal curvature index of the spine. The calculation formula is as follows: .

[0084] Three bony landmarks—the left anterior superior iliac spine, the right anterior superior iliac spine, and the upper border of the pubic symphysis—were selected as feature points of the pelvic plane. The spatial coordinates of these feature points were extracted from the 3D body model, and the pelvic plane was fitted using the least squares method. The angle between the pelvic plane normal vector and the vertical reference vector was calculated and used as the pelvic anteroposterior tilt angle index. The calculation formula is as follows: ,in, The normal vector of the pelvic plane. The vertical reference vector is The dot product of two vectors. Let be the magnitude of the vector.

[0085] The personalized intervention plan generation module converts the fat bulking index, skin color change index, upper arm muscle firmness index, spinal coronal curvature index, and pelvic anteroposterior tilt angle index into corresponding specific values. Based on these specific values, nutritional health status analysis is performed, and corresponding nutritional intervention plans are generated. The specific implementation process is as follows: A health sample reference database covering students aged 6 to 18 was pre-established. The database contains 15,000 healthy student samples with no chronic diseases, no skeletal development abnormalities, and whose nutritional indicators meet the standards. The mean and standard deviation of five indices were statistically analyzed by age and gender as standardized benchmark parameters. For the five indices of the tested students, the standardized baseline parameters of the corresponding groups were retrieved in combination with their actual age and gender. The standardization transformation was completed (the original index value was subtracted from the mean of the corresponding group and then divided by the standard deviation of the corresponding group) to obtain the specific values ​​of the five indices. This eliminated the assessment bias caused by differences in age, gender and individual development. The calculated specific values ​​were truncated to limit the range of specific values ​​to [-4,4] to avoid the impact of extreme outliers on subsequent assessments. Abnormalities are classified into three levels based on the absolute value of their specificity values: Normal: The absolute value of the specific value is <1; Mild abnormality: 1 ≤ absolute value of specificity < 2; Moderate abnormality: 2 ≤ absolute value of specificity < 3; Severe abnormality: The absolute value of the specific value is ≥3.

[0086] Based on the abnormal combination patterns of the five indices, and considering the progression of nutritional imbalance from superficial to deep and from short-term symptoms to long-term organic changes, nutritional imbalance is divided into four stages: latent, early, middle, and late, with corresponding warning levels. If a group of abnormal indices simultaneously meets the conditions for multiple stages, the stage with the higher warning level will be used. Specifically: Incubation period (1-4 weeks of nutritional imbalance), blue alert, criteria for judgment: only one of the skin color change index and fat bulging index is slightly abnormal, and the other four indices are normal; during this stage, nutritional imbalance only affects the body surface and subcutaneous soft tissue, without affecting muscles and bones, which is the best window of intervention.

[0087] Early stage (4-12 weeks of nutritional imbalance), yellow warning, judgment criteria: meeting any of the following conditions: ① One of the skin color change index and fat bulging index is moderately abnormal; ② The skin color change index or fat bulging index is mildly abnormal, and the upper arm muscle firmness index is mildly abnormal; at this stage, nutritional imbalance begins to affect skeletal muscle development.

[0088] Mid-stage (3-6 months of nutritional imbalance), orange alert, judgment criteria: meeting any of the following conditions: ① Upper arm muscle tightness index is moderately abnormal; ② Upper arm muscle tightness index is mildly abnormal, and at the same time, the pelvic anterior-posterior tilt angle index is mildly abnormal; at this stage, nutritional imbalance causes a decrease in muscle strength, which gradually affects pelvic posture.

[0089] Late stage (nutritional imbalance for more than 6 months), red alert, judgment criteria: meeting any of the following conditions: ① The coronal curvature index of the spine is abnormal to any degree; ② Both the pelvic anteroposterior tilt angle index and the coronal curvature index of the spine are abnormal or above; at this stage, nutritional imbalance has caused structural changes in the bones, and there is an irreversible risk of development.

[0090] The dominant index priority method is used to determine the type of nutritional problem. Specifically, from all abnormal indices, the index with the shortest response period is selected as the first dominant index. If two or more indices have the same response period, the index with the higher degree of abnormality is selected as the first dominant index. The nutrient corresponding to the first dominant index is the primary nutritional problem, and the nutrients corresponding to the other abnormal indices are the secondary nutritional problems.

[0091] The comprehensive nutrition and health score is calculated based on the absolute values ​​of the five index specific values ​​and their corresponding time-series weights. The specific calculation formula is as follows: ,in, , , , and These are the specific values ​​for the skin color change index, fat bulge index, upper arm muscle firmness index, pelvic anteroposterior tilt angle index, and spinal coronal curvature index.

[0092] At the same time, nutritional health status is classified according to the comprehensive nutritional health score: excellent: ; good: ; Mild abnormality: ; Moderate abnormality: ; Key anomalies: .

[0093] Based on the type, stage of development, and severity of nutritional problems, a quantitative intervention priority is generated, calculated using the following formula: ,in, For the first Intervention priorities for nutrient-like substances For basic nutrient weighting, possible values ​​are: protein 1.2, calcium 1.1, iron 1.0, vitamin D 1.0, vitamin B12 0.9, zinc 0.9, carbohydrates 0.8, and fat 0.8. For the first The absolute value of the specific value of the dominant index corresponding to each nutrient class. The value is the stage coefficient, with values ​​of 1.0 for the incubation period, 1.2 for the early stage, 1.5 for the middle stage, and 2.0 for the late stage.

[0094] The nutrients are ranked from highest to lowest priority for intervention, with the highest priority nutrients being the primary intervention targets.

[0095] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions will not cause the essence of the corresponding technical solutions to deviate from the protection scope of the technical solutions of the embodiments of the present invention.

Claims

1. A student nutrition and health assessment method based on facial expression and body posture analysis, characterized in that, Includes the following steps: Multiple cameras were deployed to simultaneously collect color images and depth point cloud data of the students being tested. After registration, fusion and preprocessing, a three-dimensional human body model was generated. A three-dimensional human face model is extracted from a three-dimensional human body model, and a circular or arbitrary-shaped periorbital region is delineated as the facial test area with the geometric center of the bony orbit as the origin and a preset length as the radius. The relative offset quantification method based on bony benchmarks combined with the statistical method of epidermal pigment distribution was used to quantify the degree of fat bulging and skin color change in the facial test area, and the fat bulging index and skin color change index were obtained. A three-dimensional body model was extracted from a three-dimensional human body model. Differential geometric curve curvature change rate analysis method, spatial cubic spline curve fitting and curvature integral method, and spatial plane fitting and normal vector angle calculation method were used to calculate the upper arm muscle tightness index, the spine coronal plane curvature index, and the pelvic anterior and posterior tilt angle index. The five indices mentioned above are converted into specific values. Nutritional health status is analyzed based on these specific values, and corresponding nutritional intervention plans are generated.

2. The student nutrition and health assessment method based on facial expression and body posture analysis according to claim 1, characterized in that, The specific steps for quantifying the degree of fat protrusion in the facial test area using the bony benchmark layered relative offset quantification method are as follows: Using the bony orbital contour as an anatomical benchmark, the bony orbital benchmark region is delineated, and the total three-dimensional surface area of ​​the bony orbital benchmark region is calculated. The area where the orbital fat pad contour extends beyond the bony orbital reference area and protrudes outward is defined as the bulging feature area. The three-dimensional surface area of ​​the bulging feature area is calculated. Starting from any point on the bony orbital reference boundary, the distance from that point to the boundary of the bulging feature area is measured along the outer normal direction. The distance from all boundary points to the boundary of the bulging area along the outer normal direction is taken, and the maximum value is recorded as the maximum radial outward expansion distance. Then, by introducing the maximum theoretical extension threshold of the orbital bony structure, the fat bulging index is obtained by fusing the two-dimensional parameters of the bulging area ratio and the maximum outward expansion distance.

3. The student nutrition and health assessment method based on facial expression and body posture analysis according to claim 1, characterized in that, The specific steps for quantifying the degree of skin color change in the test area of ​​the face using epidermal pigment distribution statistical methods are as follows: The facial area to be tested is divided into several uniform sampling units. For each sampling unit, the spatial coordinates of the epidermis and the spatial coordinates of the bony orbit are extracted, and the vertical distance between the two points in the vertical axis direction is calculated as the epidermal-bone relative offset of the sampling unit. Then, an individual normalization factor is introduced to normalize the relative epidermal-bone offset of each sampling unit; Simultaneously, epidermal color information of each sampling unit is collected, color components are extracted in the standard color space, and the average skin color of the non-lesion area of ​​the face of the same individual is used as the individual's baseline skin color to calculate the color difference value of each sampling unit. The normalized relative offset between the epidermis and bone was used as the spatial morphological weight, and weighted and fused with the color difference value to calculate the skin color change index.

4. The student nutrition and health assessment method based on facial expression and body posture analysis according to claim 1, characterized in that, The specific steps for calculating the upper arm muscle firmness index using the differential geometric curve curvature change rate analysis method are as follows: Based on the acromion and elbow points among the bony landmarks, the line connecting the acromion and elbow points is determined as the upper arm midline. The midpoint of this line is taken, and a plane perpendicular to the upper arm midline is drawn through the midpoint. The intersection of this plane and the three-dimensional body model is the upper arm cross-sectional contour curve. Obtain the geometric centroid of the upper arm cross-sectional profile curve, and convert the upper arm cross-sectional profile curve into polar coordinate parameter form with the geometric centroid as the pole. Based on the obtained upper arm cross-sectional profile curve in polar coordinate form, the curvature of the upper arm cross-sectional profile curve at each polar angle is determined. The first derivative of curvature with respect to the polar angle is then calculated using the central difference method, and the first derivative is integrated over the entire circumference to obtain the upper arm muscle firmness index.

5. The student nutrition and health assessment method based on facial expression and body posture analysis according to claim 1, characterized in that, The specific steps for calculating the coronal curvature index of the spine using spatial cubic spline curve fitting and curvature integral methods are as follows: Key points of 14 vertebral bodies of the spine were extracted from the 7th cervical vertebra, the 12th thoracic vertebra, and the 5th lumbar vertebra to form a spatial point set; Then project the set of spatial points onto the X-Y plane of the body coordinate system to obtain the projected set of points; The natural cubic spline interpolation algorithm is used to fit the projection point set to obtain the coronal plane curve of the spine. The natural cubic spline interpolation algorithm refers to constructing a cubic polynomial between each adjacent projection point so that the fitted curve passes through all projection points accurately, and the first and second derivatives are continuous at the internal nodes, while the second derivative is zero at the two endpoints. Calculate the curvature at any point on the coronal plane curve of the spine, and integrate the curvature over the entire curve length to obtain the coronal plane curvature index of the spine.

6. The student nutrition and health assessment method based on facial expression and body posture analysis according to claim 1, characterized in that, The specific steps for calculating the pelvic tilt angle index using spatial plane fitting and the angle between normal vectors are as follows: Three bony landmarks—the left anterior superior iliac spine, the right anterior superior iliac spine, and the upper edge of the pubic symphysis—were selected as feature points of the pelvic plane. The spatial coordinates of these feature points were extracted from the three-dimensional body model, and the pelvic plane was fitted using the least squares method. The angle between the pelvic plane normal vector and the vertical reference vector was calculated and used as the pelvic anterior-posterior tilt angle index.

7. The student nutrition and health assessment method based on facial expression and body posture analysis according to claim 1, characterized in that, The specific steps for converting the five indices into specific values ​​include: Establish a health sample reference database covering students aged 6 to 18, and statistically analyze the mean and standard deviation of five indices by age and gender as standardized benchmark parameters; For the five indices of the tested students, the standardized baseline parameters of the corresponding groups are retrieved in combination with their actual age and gender, and the standardization transformation is completed. That is, the original index value is subtracted from the mean of the corresponding group and then divided by the standard deviation of the corresponding group to obtain the specific values ​​of the five indices.

8. The student nutrition and health assessment method based on facial expression and body posture analysis according to claim 1, characterized in that, The specific process for nutritional health status analysis based on specific values ​​is as follows: Based on the absolute value of the specific value, the degree of abnormality is divided into three levels, including mild abnormality, moderate abnormality and severe abnormality. According to the abnormal combination pattern of the five indices, and combined with the law of nutritional imbalance from superficial to deep and from short-term symptoms to long-term organic changes, nutritional imbalance is divided into four stages: latent period, early stage, middle stage and late stage, and a corresponding warning level is matched for each stage. The dominant index priority method is used to determine the type of nutritional problem. Specifically, the index with the shortest response period is selected from all abnormal indices as the first dominant index. If two or more indices have the same response period, the index with the higher degree of abnormality is selected as the first dominant index. The nutrient corresponding to the first dominant index is the primary nutritional problem, and the nutrients corresponding to the other abnormal indices are the secondary nutritional problems. Based on the absolute values ​​of the five index specific values ​​and combined with the corresponding time-series weights, a comprehensive nutrition and health score is calculated, and the nutrition and health status is classified according to the comprehensive nutrition and health score. Based on the type, stage of development, and severity of nutritional problems, quantitative intervention priorities are generated and sorted from high to low priority, with the highest priority nutrients being the primary intervention targets.

9. A student nutrition and health assessment method based on facial expression and body posture analysis according to claim 8, characterized in that, The criteria for determining the four stages are as follows: Incubation period: Only one of the skin color change index and fat bulging index was slightly abnormal, while the other four indices were normal; Early stage: Meeting any of the following conditions: ① One of the skin color change index and fat bulging index is moderately abnormal; ② The skin color change index or fat bulging index is mildly abnormal, and the upper arm muscle tightness index is also mildly abnormal. Mid-term: Meets any of the following conditions: ① Upper arm muscle tightness index is moderately abnormal; ② Upper arm muscle tightness index is mildly abnormal, and at the same time, pelvic anteroposterior tilt angle index is mildly abnormal. Late stage: Meeting any of the following conditions: ① The coronal curvature index of the spine is abnormal to any degree; ② Both the anteroposterior tilt angle index of the pelvis and the coronal curvature index of the spine are abnormal.

10. A student nutrition and health assessment system based on facial expression and body posture analysis, applied to the student nutrition and health assessment method based on facial expression and body posture analysis as described in claim 1, characterized in that, include: The image acquisition and 3D reconstruction module simultaneously acquires color images and depth point cloud data of the students being tested by deploying multiple cameras, and generates a 3D human body model after registration, fusion and preprocessing. The facial nutrition feature quantitative analysis module extracts a three-dimensional face model from a three-dimensional human body model, and delineates a circular or arbitrary-shaped periorbital region as the facial test area with the geometric center of the bony orbit as the origin and a preset length as the radius. The relative offset quantification method based on bony benchmarks combined with the statistical method of epidermal pigment distribution was used to quantify the degree of fat bulging and skin color change in the facial test area, and the fat bulging index and skin color change index were obtained. The body nutrition characteristics quantitative analysis module extracts a three-dimensional body model from a three-dimensional human body model and calculates the upper arm muscle tightness index, the spine coronal plane curvature index, and the pelvic anterior and posterior tilt angle index by using differential geometric curve curvature change rate analysis method, spatial cubic spline curve fitting and curvature integral method, and spatial plane fitting and normal vector angle calculation method, respectively. The personalized intervention plan generation module converts the above five indices into specific values, analyzes nutritional health status based on the specific values, and generates corresponding nutritional intervention plans.

Images