Intelligent body measurement method based on dynamic capture and motion modeling
By using multi-camera dynamic posture capture and 3D coordinate transformation technology, combined with spatial error compensation algorithms, the problem of inaccurate size measurement in existing intelligent body measurement during dynamic motion has been solved, enabling more precise clothing customization. This technology is suitable for large-scale production and online platforms, improving user experience and security.
Patent Information
- Application Number
- CN202511317241.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-09-16
- Publication Date
- 2025-12-16
AI Technical Summary
Existing intelligent body measurement technology cannot accurately capture the key dimensions of the human body during dynamic movement, resulting in ill-fitting problems when custom-made clothing is worn dynamically. In particular, it ignores the dynamic changes in core areas such as the waist and hips, and causes perspective distortion errors when the distance to the camera changes.
At least two RGB-D cameras are used to capture the stepping motion in place from different angles. Key points are captured through dynamic posture. Combined with three-dimensional coordinate transformation and spatial error compensation algorithms, multi-dimensional dimensional data covering human dynamic movement is generated to eliminate distance errors and optimize the calculation of key dimensions such as waist circumference and shoulder width.
It enables the acquisition of more comprehensive size data in dynamic motion, improves the accuracy and reliability of body measurement, reduces human error, is suitable for large-scale garment production, and enhances user experience and production efficiency, especially for the safety and comfort of body measurement for children, the elderly and people with special body types.
Smart Images
Figure BDA0005597452380000051
Abstract
Description
Technical Field
[0001] This invention relates to the field of clothing measurement, specifically to an intelligent measurement method based on dynamic capture and motion modeling. Background Technology
[0002] With the increasing demand for personalized clothing customization, intelligent body measurement technology has become a key link connecting consumers and customized production. Traditional manual body measurement relies on professionals holding soft tape measures inch by inch, which is not only cumbersome and inefficient, but also prone to subjective errors due to differences in measurement techniques and pressure, making it difficult to adapt to large-scale customization scenarios. Even if existing static intelligent body measurement systems capture the static posture of the human body through monocular or binocular cameras, they still do not take into account the dynamic extension of the limbs during human movement (such as shoulder swing and waist twisting when walking). This leads to ill-fitting problems such as shoulder tightness and waist constriction when customized clothing is worn dynamically, failing to meet consumers' high requirements for clothing comfort.
[0003] Existing technologies for dynamic body measurement still have many shortcomings: some methods only capture simple human movements (such as standing and turning), failing to select key movements closely related to clothing fit (such as limb circulation movements during stationary marching), resulting in incomplete key point motion trajectory data; key point extraction often focuses on the extremities, ignoring dynamic changes in core areas such as the abdomen and hips, making it difficult to accurately calculate key dimensions such as waist and hip circumference; furthermore, most dynamic body measurement methods do not compensate for errors due to distance differences between the camera and the human body, and when the measurement distance changes, perspective distortion can easily lead to deviations in dimensional measurements, for example, when the distance to the camera is far, the shoulder width measurement value is smaller, affecting the design of the upper garment pattern; at the same time, existing methods often directly use static projection distance to calculate waist circumference without incorporating the lateral change adjustment coefficient of the abdominal key point motion trajectory, failing to reflect the stretching of the waist circumference during human movement, resulting in problems such as crotch or waist laxity in customized trousers when walking. These defects all lead to insufficient accuracy and reliability of intelligent body measurement, seriously restricting the personalized experience and production efficiency of clothing customization. Summary of the Invention
[0004] This invention aims to provide an intelligent body measurement method based on dynamic capture and motion modeling, with the specific solution as follows:
[0005] A smart body measurement method based on dynamic capture and motion modeling, the specific steps of which are as follows:
[0006] Step 1. Dynamic pose capture and key point extraction:
[0007] The user's stepping motion is captured by at least two RGB-D cameras at different angles, and the data of key points is captured every four frames from the dynamic posture of the stepping motion video.
[0008] Step 2. Three-dimensional coordinate transformation and dimension calculation:
[0009] Based on the camera focal length, principal point coordinates, and extrinsic parameters, the key points of the 2D image are converted into three-dimensional spatial coordinates;
[0010] Step 3. Spatial error compensation algorithm:
[0011] The three-dimensional spatial coordinates obtained in step 2 are compensated by spatial error compensation to eliminate the error caused by the different distances between the person and the camera;
[0012] Step 4. Generate curves:
[0013] In the 3D spatial coordinates after the spatial error compensation algorithm, all positions of some key points are aggregated to generate a curve;
[0014] Step 5. Extract the following data and organize it into a table.
[0015] The key points in step (1) are the left shoulder, right shoulder, left elbow, right elbow, left hip, right hip, left abdomen, right abdomen, left knee, right knee, left ankle, and right ankle, a total of 12 key points.
[0016] The left abdomen is 3 cm to the left of the navel, the right abdomen is 3 cm to the left and right of the navel, and the left and right hips are the greater trochanter of the femur.
[0017] In step (1), the camera's shooting frame rate is set to 30-60fps.
[0018] In step (1), the stepping motion in place does not involve strenuous activity or limb crossing.
[0019] The spatial error compensation formula in step (3) is as follows:
[0020] Compensated size = Original size × (1 + a·d) 2 +b·d); where d is the distance between the human body and the camera, a = 0.002, b = -0.01.
[0021] The measurement position of d is the distance from the midpoint between the left and right hips to the camera, and the range of d is 0.5-3 meters.
[0022] The key points in step (4) are the left shoulder, right shoulder, left elbow, right elbow, left abdomen, right abdomen, left knee, and right knee.
[0023] The data in step (5) includes: the Euclidean distance of the left and right shoulders in their initial state, the leftmost position of the left shoulder movement path, the rightmost position of the right shoulder movement path, waist circumference data, the vertical distance from the hip to the left ankle, the vertical distance from the hip to the right ankle, the vertical distance from the left knee to the left hip and the right hip respectively when the left knee and the right knee are raised to their highest positions, the initial distance of the left ankle, the right ankle, the left elbow and the right elbow.
[0024] The waist data is the horizontal projection distance between the left and right abdomens multiplied by a waist coefficient, which is 1.4-1.8. The waist coefficient is determined based on the lateral distance of the movement trajectories of the left and right abdomens.
[0025] This invention has the following advantages:
[0026] 1. Based on dynamic capture technology, this invention acquires extreme motion data such as the leftmost position of the left shoulder movement path, the rightmost position of the right shoulder movement path, and the highest positions of the left and right knees, as well as dynamic information such as the lateral distance of the movement trajectories of the left and right abdomens. Compared with traditional static body measurement, which can only obtain static dimensions under fixed postures, this invention covers multi-dimensional dimensional characteristics of the human body in both static and dynamic states. Not only is the data more comprehensive, but it can also truly reflect the limb extension limits of the human body in actual activities, providing more accurate dimensional support for the sports functional design of clothing. This is because the demand for sports functional clothing is increasing rapidly year by year. However, some sports functional clothing designs are inherently more fitted, so the acquisition of the above data is even more necessary.
[0027] 2. By employing multi-camera, multi-angle shooting and 3D coordinate transformation technology, combined with spatial error compensation algorithms, distance errors are effectively eliminated, improving the accuracy and reliability of body measurement data. Compared with traditional manual measurement, this reduces human error. In practice, we found that if the influence of distance is not eliminated, the impact on the results is significant. Furthermore, since size adjustments for clothing are calculated in centimeters, the data has no reference value if adjustments are not made.
[0028] 3. Automated data collection and processing reduces manual intervention, shortens measurement time, and improves measurement efficiency. It is suitable for large-scale garment production or online customization platforms, enhancing user experience and enterprise production efficiency.
[0029] 4. The non-contact measurement method avoids hygiene risks and psychological resistance caused by direct contact with the human body, making it especially suitable for children, the elderly and people with special body types, thus improving the safety and comfort of the measurement process.
[0030] 5. It has significant advantages, especially in optimizing waist circumference, shoulder width, and leg length (including the length and width of the thigh and calf). Detailed Implementation
[0031] A smart body measurement method based on dynamic capture and motion modeling, the specific steps of which are as follows:
[0032] Step 1. Dynamic pose capture and key point extraction:
[0033] The system captures the user's stepping motion from different angles using at least two RGB-D cameras, and captures key point data every four frames from the dynamic posture of the stepping video (this setting balances capturing dynamic changes with computational efficiency, and performs noise reduction and contrast enhancement processing on the sampled frames).
[0034] The key points are the left shoulder, right shoulder, left elbow, right elbow, left hip, right hip, left abdomen, right abdomen, left knee, right knee, left ankle, and right ankle, a total of 12 key points. Among them, the left abdomen is 3 cm to the left of the navel, the right abdomen is 3 cm to the left and right of the navel, and the left and right hips are the greater trochanter of the femur.
[0035] The camera's shooting frame rate is set to 30-60fps.
[0036] Keypoint identification can be performed using pose estimation algorithms (such as OpenPose or MediaPipe), which are existing technologies and will not be elaborated upon here.
[0037] Users are required to stand naturally with their feet shoulder-width apart and step in place at a comfortable cadence (1.5-2 steps / second), avoiding violent shaking or crossing of limbs.
[0038] Step 2. Three-dimensional coordinate transformation and dimension calculation:
[0039] Based on the camera focal length, principal point coordinates, and extrinsic parameters, the key points of the 2D image are converted into three-dimensional spatial coordinates (the above steps are existing technology).
[0040] The process is as follows:
[0041] 1. Camera parameter calibration
[0042] The camera's intrinsic parameters (focal length f) are obtained in advance using Zhang Zhengyou's calibration method. x f y Principal point coordinates c x c y The world coordinate system and the image coordinate system are established by using external parameters (rotation matrix R, translation vector T) and external parameters (rotation matrix R, translation vector T).
[0043] 2.2D to 3D coordinate transformation
[0044] Based on the perspective projection model, the coordinates (u,v) of key points on the image plane are transformed into three-dimensional coordinates (X,v) in the camera coordinate system. c ,Y c Z c ):
[0045]
[0046] Among them, X w These are three-dimensional coordinates in the world coordinate system.
[0047] Step 3. Spatial error compensation algorithm:
[0048] 1. Measurement of distance d
[0049] The distance d (in meters) from the human body's center of mass (midpoint between the left and right hips) to the camera is obtained directly using the camera's depth information. If the result is less than 0.5 meters or greater than 3 meters, the user should be prompted to adjust their position.
[0050] 2. Application of the error compensation formula
[0051] The three-dimensional coordinates X obtained in step 2 w Compensation is performed to correct the scaling error caused by changes in distance d:
[0052] Compensated size = xw = (1 + a·d) 2 +b·d); where d is the distance between the human body and the camera, a = 0.002, b = -0.01.
[0053] Step 4. Generate curves:
[0054] In the 3D spatial coordinates after the spatial error compensation algorithm, all positions of some key points are aggregated to generate curves; the key points are the left shoulder, right shoulder, left elbow, right elbow, left abdomen, right abdomen, left knee, and right knee, totaling 8.
[0055] Step 5. Extract the following data and organize it into a table.
[0056] The data that needs to be organized is as follows:
[0057] The initial Euclidean distance between the left and right shoulders;
[0058] The leftmost position of the left shoulder's movement path, and the rightmost position of the right shoulder's movement path;
[0059] Waist circumference data is calculated by multiplying the horizontal projection distance between the left and right abdomens by a waist coefficient, which ranges from 1.4 to 1.8. The waist coefficient is ultimately determined based on the lateral distance of the movement trajectories of the left and right abdomens and the body type (normally it is 1.5).
[0060] Vertical distance from hip to left ankle, and from hip to right ankle;
[0061] The vertical distance between the left and right knees at their highest positions and the left and right hips, respectively;
[0062] The initial distance between the left ankle, right ankle, left elbow, and right elbow.
[0063] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.
Claims
1. An intelligent body measurement method based on dynamic capture and motion modeling, characterized in that, The specific steps are as follows: Step 1. Dynamic pose capture and key point extraction: The user's stepping motion is captured by at least two RGB-D cameras at different angles, and the data of key points is captured every four frames from the dynamic posture of the stepping motion video. Step 2. Three-dimensional coordinate transformation and dimension calculation: Based on the camera focal length, principal point coordinates, and extrinsic parameters, the key points of the 2D image are converted into three-dimensional spatial coordinates; Step 3. Spatial error compensation algorithm: The three-dimensional spatial coordinates obtained in step 2 are compensated by spatial error compensation to eliminate the error caused by the different distances between the person and the camera; Step 4. Generate curves: In the 3D spatial coordinates after the spatial error compensation algorithm, all positions of some key points are aggregated to generate a curve; Step 5. Extract the following data and organize it into a table.
2. The intelligent body measurement method based on dynamic capture and motion modeling as described in claim 1, characterized in that: The key points in step (1) are the left shoulder, right shoulder, left elbow, right elbow, left hip, right hip, left abdomen, right abdomen, left knee, right knee, left ankle and right ankle, a total of 12 key points.
3. The intelligent body measurement method based on dynamic capture and motion modeling as described in claim 2, characterized in that: The left abdomen is 3 cm to the left of the navel, the right abdomen is 3 cm to the left and right of the navel, and the left and right hips are the greater trochanter of the femur.
4. The intelligent body measurement method based on dynamic capture and motion modeling as described in claim 1, characterized in that: In step (1), the camera's shooting frame rate is set to 30-60fps.
5. The intelligent body measurement method based on dynamic capture and motion modeling as described in claim 1, characterized in that: In step (1), the stepping motion in place does not involve strenuous activity or limb crossing.
6. The intelligent body measurement method based on dynamic capture and motion modeling as described in claim 1, characterized in that: The spatial error compensation formula in step (3) is as follows: Where d is the distance between the human body and the camera, a=0.002, b=-0.
01.
7. The intelligent body measurement method based on dynamic capture and motion modeling as described in claim 6, characterized in that: The measurement position of d is the distance from the midpoint between the left and right hips to the camera, and the range of d is 0.5-3 meters.
8. The intelligent body measurement method based on dynamic capture and motion modeling as described in claim 1, characterized in that: The key points in step (4) are the left shoulder, right shoulder, left elbow, right elbow, left abdomen, right abdomen, left knee, and right knee.
9. The intelligent body measurement method based on dynamic capture and motion modeling as described in claim 1, characterized in that: The data in step (5) includes: the Euclidean distance of the left and right shoulders in their initial state, the leftmost position of the left shoulder movement path, the rightmost position of the right shoulder movement path, waist circumference data, the vertical distance from the hip to the left ankle, the vertical distance from the hip to the right ankle, the vertical distance from the left knee to the left hip and the right hip respectively when the left knee and the right knee are raised to their highest position, the initial distance of the left ankle, the right ankle, the left elbow and the right elbow.
10. The intelligent body measurement method based on dynamic capture and motion modeling as described in claim 9, characterized in that, The waist data is the horizontal projection distance between the left and right abdomens multiplied by a waist coefficient, which is 1.4-1.
8. The waist coefficient is determined based on the lateral distance of the movement trajectories of the left and right abdomens.