Real-time 3D motion modeling method

Abstract

The invention discloses a real-time 3D motion modeling method, and belongs to the technical field of image analysis. According to the method, firstly, an image quality state and detection reliability model is constructed by evaluating image definition, brightness, key point detection number and inter-frame stability, and then imaging quality confidence is generated in combination with a camera distance factor. And based on the confidence, a dynamic weight model is used to optimize a multi-camera 3D coordinate fusion process, and the key point positioning precision is improved. Then, the reconstructed dynamic human body model is compared with a standard action library, and the deviation of the joint angle, the action amplitude and the rhythm is obtained; meanwhile, the basic states such as the age of the person and the limb proportion and the external factors such as the environment temperature and the oxygen content are synthesized, an action completion degree model is built, and comprehensive and self-adaptive evaluation on the action quality is achieved. According to the invention, the accuracy and robustness of 3D motion modeling in a complex actual scene are effectively improved.

Description

Technical Field

[0001] This invention belongs to the field of image analysis technology, and in particular relates to a real-time 3D motion modeling method. Background Technology

[0002] Real-time 3D motion modeling technology is an important research direction in the field of image analysis, and it is widely used in virtual reality, sports training, human-computer interaction, and medical rehabilitation. This technology captures human movements using multiple cameras and reconstructs 3D human models in real time to achieve accurate analysis and feedback of movements. With the diversification of application scenarios, higher demands are being placed on the real-time performance, accuracy, and adaptability of the modeling.

[0003] In existing technologies, 3D motion modeling mainly relies on keypoint detection and 3D coordinate fusion from multi-view images. Common methods include deep learning-based human pose estimation and traditional multi-view geometric reconstruction. However, these methods often focus only on optimizing keypoint detection algorithms, neglecting the impact of image quality, environmental factors, and individual differences on the modeling results in real-world applications. For example, image blurring, lighting variations, and camera distance fluctuations can all lead to unstable keypoint detection and distorted 3D models.

[0004] Existing technologies have the following drawbacks: the lack of systematic evaluation of image quality (such as sharpness and brightness) and detection reliability (such as the number of key points and inter-frame stability) leads to unreasonable weight allocation during 3D coordinate fusion. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a real-time 3D motion modeling method that solves the aforementioned problems.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a real-time 3D motion modeling method, further comprising: An image quality status model is constructed based on image sharpness and image brightness, and the image quality status coefficient is output. A detection reliability model is constructed based on the inter-frame change rate of the distance between adjacent key points and the actual number of key points detected, and the detection reliability is output. An imaging quality confidence level is constructed based on the image quality state coefficient and detection reliability under the distance between the camera and the detection target. A dynamic weight model is constructed based on the image quality confidence level, and the weight of each camera's 3D coordinates in the preset 3D coordinate fusion model is output.

[0007] Based on the above technical solutions, the present invention also provides the following optional technical solutions: Further technical solutions include the following steps: A dynamic human body model is generated based on the 3D coordinates of key human body points (such as joints) generated by the 3D coordinate fusion model (existing technology). The dynamic human body model is compared with a 3D model database of preset standard movements to obtain the joint angle deviation (the maximum value among the absolute deviations of each joint angle from the standard value), the movement amplitude deviation (the maximum value among the absolute deviations of each joint range of motion from the standard range), and the movement rhythm deviation (the maximum value among the angle deviations of each movement node from the standard time node). A basic state model of personnel is constructed based on their age and the ratio of their sitting height to their standing height, and the basic state coefficients of personnel are output. An environmental state model is constructed based on ambient temperature and ambient oxygen content, and the environmental state coefficient is output. Based on the personnel's basic state coefficient and the environmental state coefficient, a motion completion model is constructed (the maximum value among the absolute deviations of each joint angle from the standard value), motion amplitude deviation (the maximum value among the absolute deviations of each joint's range of motion from the standard range), and motion rhythm deviation (the maximum value among the angle deviations of each motion node from the standard time node), and the motion completion rate is output.

[0008] Further technical solution: The 3D coordinate fusion model is represented as follows: in, Indicates the fusion of the first Three-dimensional coordinate vectors of each joint point Indicates the total number of cameras. Indicates the first The estimated number of cameras Three-dimensional coordinate vectors of each joint point Indicates the first Dynamic weights of each camera and , Indicates the first Confidence level of image quality for each camera.

[0009] A further technical solution: The dynamic weight model is expressed as follows: in, Indicates the first Dynamic weights of each camera This represents the lumpedity parameter (which controls the degree of concentration of the weight distribution; the larger the value, the greater the weight of the high-confidence camera). Indicates the first Confidence level of image quality for each camera.

[0010] Further technical solution: The steps for constructing imaging quality confidence based on the image quality state coefficient and detection reliability under the distance between the camera and the detection target are as follows: The distance between the camera and the target is processed by maximum-min normalization to obtain the distance index; An imaging quality confidence model is constructed based on the image quality state coefficient under the distance index and the detection reliability, and the imaging quality confidence score is output. The imaging quality confidence model is expressed as follows: in, Indicates confidence level in image quality. Represents the image quality status coefficient. Indicates the reliability of the test. Indicates the distance attenuation coefficient. Represents the distance index, the Furthermore, the larger the value, the higher the confidence level of the image quality.

[0011] Further technical solution: The steps for constructing a detection reliability model based on the inter-frame change rate of the distance between adjacent keypoints and the actual number of detected keypoints, and outputting the detection reliability, are as follows: The keypoint quantity index is obtained by processing the ratio of the actual number of keypoints detected to the total number of keypoints. The inter-frame change rate of the distance between adjacent key points is subjected to maximum-minimum normalization to obtain the inter-frame change rate exponent. A detection reliability model is constructed based on the keypoint quantity index and the inter-frame change rate index, and the detection reliability is output. The detection reliability model is expressed as follows: in, Indicates the reliability of the test. Indicates the number index of key points. This represents the inter-frame change rate exponent. The coefficient representing the rate of change sensitivity is... The higher the value, the more reliable the detection.

[0012] Further technical solution: The steps for constructing an image quality status model based on image sharpness and image brightness and outputting image quality status coefficients are as follows: The image sharpness and image brightness are processed by maximum-minimum normalization to obtain the image sharpness index and image brightness index; An image quality state model is constructed based on the image sharpness index and the image brightness index, and the image quality state coefficient is output. The image quality state model is expressed as follows: in, Represents the image quality status coefficient. Indicates the image sharpness index. Indicates the image brightness index. This represents the image sharpness sensitivity coefficient. The image brightness sensitivity coefficient is represented by the following: The larger the value, the better the image quality.

[0013] Further technical solution: Based on the personnel's basic state coefficient and environmental state coefficient, a motion completion model is constructed (the maximum absolute deviation of each joint angle from the standard value), a range of motion deviation (the maximum absolute deviation of each joint's range of motion from the standard range), and a rhythmic motion deviation (the maximum angular deviation of each motion node from the standard time node). The steps for outputting the motion completion score are as follows: The joint angle deviation, the range of motion deviation, and the rhythm of motion deviation are each compared with their respective maximum permissible values ​​to obtain the joint angle deviation index, the range of motion deviation index, and the rhythm of motion deviation index. A motion completion model is constructed based on the joint angle deviation index, motion amplitude deviation index, and motion rhythm deviation index under the personnel's basic state coefficient and environmental state coefficient, and the motion completion score is output. The motion completion model is expressed as follows: in, Indicates the degree of completion of the action. Represents the basic status coefficient of personnel. Represents the environmental state coefficient. Indicates the joint angle deviation index. Indicates the deviation index of movement amplitude. This indicates the deviation index of movement rhythm. This represents the sensitivity coefficient to joint angle deviation. This represents the sensitivity coefficient to deviations in the range of motion. The sensitivity coefficient for movement rhythm deviation is represented by the following. The higher the value, the higher the completion rate of the action.

[0014] Further technical solution: The steps for constructing an environmental state model based on ambient temperature and ambient oxygen content, and outputting environmental state coefficients, are as follows: The temperature deviation index is obtained by comparing the absolute difference between the ambient temperature and the optimal temperature with the allowable deviation from the optimal temperature. The oxygen content deviation index is obtained by comparing the absolute difference between the ambient oxygen content and the optimum oxygen content with the allowable deviation from the optimum oxygen content. An environmental state model is constructed based on the temperature deviation index and the oxygen content deviation index, and the environmental state coefficient is output. The environmental state model is expressed as follows: in, Represents the environmental state coefficient. This indicates the temperature deviation index. This indicates the oxygen deviation index. Indicates the temperature sensitivity coefficient. The oxygen sensitivity coefficient is represented by the following. The higher the value, the better the environmental condition.

[0015] Further technical solution: The steps for constructing a basic personnel state model based on personnel age and the ratio of sitting height to standing height, and outputting the basic personnel state coefficients are as follows: The age of the personnel and the ratio of sitting height to standing height were subjected to maximum-min normalization to obtain the age index and limb proportion index. A basic personnel status model is constructed based on age index and limb proportion index, and the basic personnel status coefficient is output. The basic personnel status model is expressed as follows: in, Represents the basic status coefficient of personnel. Indicates age index. Indicates the body proportion index. Represents the weight coefficient and The Furthermore, the higher the value, the better the basic condition of the personnel.

[0016] This invention provides a real-time 3D motion modeling method, which has the following advantages compared with the prior art: 1. This invention improves the accuracy and stability of key point detection by introducing an image quality status model, a detection reliability model, and an imaging quality confidence calculation. It also enhances the accuracy of 3D coordinate reconstruction by using a dynamic weight model to optimize multi-camera data fusion. 2. This invention combines the basic state of personnel and the state of the environment to achieve a multi-dimensional evaluation of the degree of action completion, making the modeling results more in line with the actual application scenario and significantly improving the robustness and applicability of real-time 3D motion modeling. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the process of the present invention. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0019] The specific implementation of the present invention will be described in detail below with reference to specific embodiments.

[0020] Please see Figure 1 The present invention provides a real-time 3D motion modeling method, comprising the following steps: An image quality status model is constructed based on image sharpness and image brightness, and the image quality status coefficient is output. A detection reliability model is constructed based on the inter-frame change rate of the distance between adjacent key points and the actual number of key points detected, and the detection reliability is output. An imaging quality confidence level is constructed based on the image quality state coefficient and detection reliability under the distance between the camera and the detection target. A dynamic weight model is constructed based on the image quality confidence level, and the weight of each camera's 3D coordinates in the preset 3D coordinate fusion model is output.

[0021] Through the above technical solution, this application solves the problem of unreasonable weight allocation caused by image quality fluctuations, detection instability, and changes in camera distance. For example, in virtual reality interaction scenarios, when a user's rapid movement causes blurring of some camera images, the system automatically reduces the weight of that camera and prioritizes using clear and stable camera data for coordinate fusion, thereby ensuring the continuity and accuracy of the 3D motion model. Furthermore, the dynamic weight allocation mechanism reduces the need for manual parameter tuning and improves the system's adaptability in complex lighting, occlusion, and distance-changing environments.

[0022] Preferably, the method further includes the following steps: A dynamic human body model is generated based on the 3D coordinates of key human body points (such as joints) generated by the 3D coordinate fusion model (existing technology). The dynamic human body model is compared with a 3D model database of preset standard movements to obtain the joint angle deviation (the maximum value among the absolute deviations of each joint angle from the standard value), the movement amplitude deviation (the maximum value among the absolute deviations of each joint range of motion from the standard range), and the movement rhythm deviation (the maximum value among the angle deviations of each movement node from the standard time node). A basic state model of personnel is constructed based on their age and the ratio of their sitting height to their standing height, and the basic state coefficients of personnel are output. An environmental state model is constructed based on ambient temperature and ambient oxygen content, and the environmental state coefficient is output. Based on the personnel's basic state coefficient and the environmental state coefficient, a motion completion model is constructed (the maximum value among the absolute deviations of each joint angle from the standard value), motion amplitude deviation (the maximum value among the absolute deviations of each joint's range of motion from the standard range), and motion rhythm deviation (the maximum value among the angle deviations of each motion node from the standard time node), and the motion completion rate is output.

[0023] Through the above technical solution, this application solves the problem of insufficient accuracy in movement assessment caused by ignoring individual differences and environmental factors. For example, in medical rehabilitation scenarios, it can accurately identify the limited range of motion caused by abnormal limb proportions, avoiding misjudgment as substandard movement execution; in high-altitude training scenarios, it can automatically correct the impact of the low-oxygen environment on the athlete's movement rhythm, obtaining movement completion assessment results that truly reflect the training effect.

[0024] Preferably, the 3D coordinate fusion model is represented as follows: in, Indicates the fusion of the first Three-dimensional coordinate vectors of each joint point Indicates the total number of cameras. Indicates the first The estimated number of cameras Three-dimensional coordinate vectors of each joint point Indicates the first Dynamic weights of each camera and , Indicates the first Confidence level of image quality for each camera.

[0025] Among them, dynamic weights This refers to weighting coefficients that are dynamically adjusted based on the confidence level of camera image quality. Specifically, this can be implemented using a normalized exponential function, such as the Softmax function, which maps confidence levels to weights, ensuring the sum of the weights is 1. Its purpose is to prioritize the allocation of weights to high-confidence cameras and suppress the impact of low-quality data on the fusion result. Image quality confidence This refers to a confidence index generated by combining the image quality status coefficient and detection reliability. Specifically, it can be achieved through an exponential decay model combined with a distance factor. Its function is to quantify the reliability of the current imaging conditions of the camera, providing a basis for weight allocation. Normalized denominator This refers to normalizing the weighted confidence scores, which can be achieved by summing and scaling the results proportionally. Its purpose is to eliminate the impact of absolute differences in confidence scores from different cameras on the weighting distribution, while preserving the relative confidence ratios. Specifically, this model optimizes the fusion process of multi-camera data through a dual adjustment mechanism of dynamic weights and image quality confidence. When calculating the 3D coordinates of each joint point, the dynamic weights of each camera are first used as the basis for the calculation. and image quality confidence The coordinates output by each camera Perform a weighted summation. Dynamic weights. The image quality confidence score is generated by a normalized exponential function, which gives higher weight to high-confidence cameras; This approach integrates factors such as image sharpness, brightness, keypoint detection stability, and camera distance to dynamically reflect the reliability of each camera. The normalized denominator further eliminates differences in confidence metrics between different cameras, ensuring that the fusion result reflects only the relative reliability of each camera's data. Through this mechanism, the contribution of data from low-quality or unreliable cameras is suppressed, and data from high-confidence cameras dominates the fusion result, thereby reducing model distortion. Compared to existing technologies, current methods typically employ fixed weights or rely solely on a single factor (such as camera resolution) to allocate weights. This results in the weight allocation failing to adapt adaptively to fluctuations in image quality, changes in the distance to the detected target, or environmental interference, making the fusion results susceptible to interference from low-quality data. This proposed solution, through the synergistic effect of dynamic weights and imaging quality confidence, assesses the reliability of camera data in real time and dynamically adjusts the weights based on this reliability. This makes the fusion process more closely reflect actual imaging conditions, significantly improving the rationality and robustness of weight allocation. Through the above technical solution, this application can effectively suppress 3D coordinate fusion errors caused by unstable camera imaging quality or decreased detection reliability. Especially in scenarios with changes in lighting, rapid target movement, or large differences in distance between multiple cameras, by dynamically adjusting weights, the negative impact of low-quality data on the final model is reduced, thereby improving the accuracy of 3D coordinate reconstruction of human body key points.

[0026] Preferably, the step of constructing an image quality status model based on image sharpness and image brightness and outputting image quality status coefficients is as follows: The image sharpness and image brightness are processed by maximum-minimum normalization to obtain the image sharpness index and image brightness index; An image quality state model is constructed based on the image sharpness index and the image brightness index, and the image quality state coefficient is output. The image quality state model is expressed as follows: in, Represents the image quality status coefficient. Indicates the image sharpness index. Indicates the image brightness index. This represents the image sharpness sensitivity coefficient. The image brightness sensitivity coefficient is represented by the following: The larger the value, the better the image quality.

[0027] Image sharpness refers to the sharpness of details in an image, which can be calculated using gradient energy functions or edge detection algorithms, reflecting the sharpness level by quantifying edge strength. Image brightness refers to the overall brightness of an image, which can be evaluated by calculating the average value of pixel grayscale values ​​or histogram distribution. Max-min normalization transforms the original data linearly to the [0,1] interval, eliminating the influence of different units on subsequent models. The image sharpness index and image brightness index are normalized values ​​used to unify the scale of data from different sources. The image sharpness sensitivity coefficient and image brightness sensitivity coefficient are parameters that adjust the weight of their contribution to quality evaluation. They can be determined, for example, through preset empirical values, experimental calibration, or optimization algorithms, and are used to balance the importance of sharpness and brightness in comprehensive quality assessment.

[0028] Specifically, the image sharpness and brightness are first normalized, converting the original measurements into exponents between 0 and 1. For example, if the current image sharpness measurement is 200, and the maximum value in its sample is 400 and the minimum is 50, the normalization result is (200-50) / (400-50) = 0.428. Similarly, the brightness index is calculated using the same method. Next, the two indices are input into the image quality state model, fusing the information through an exponential function. The sensitivity coefficient in the model controls the contribution of sharpness and brightness to the final quality coefficient; for example, when the sharpness sensitivity coefficient is large, the sharpness index has a more significant impact on the quality coefficient. The final output image quality state coefficient increases with increasing sharpness and brightness, dynamically reflecting the image quality level and providing a reliable weight allocation basis for subsequent 3D coordinate fusion.

[0029] Compared to existing technologies, traditional methods typically evaluate sharpness or luminance separately, lacking quantitative analysis of their combined impact and failing to consider the varying effects of sharpness and luminance on detection stability across different scenarios. Existing technologies often employ linear weighting to fuse quality metrics, which struggles to accurately describe the nonlinear coupling relationship between sharpness and luminance. This proposed solution fuses normalized sharpness and luminance indices using a nonlinear exponential model, more precisely characterizing the image quality degradation caused by their combined effect. Furthermore, dynamic weight adjustment via a sensitivity coefficient adapts to the quality assessment needs of different environments.

[0030] Through the above technical solution, this application can effectively suppress image quality fluctuations caused by sudden changes in illumination or lens defocusing, and improve the stability of key point detection. By dynamically generating image quality state coefficients, low-quality image frames can be accurately identified and their weight in 3D coordinate fusion can be reduced, thereby minimizing the impact of erroneous key point coordinates on overall modeling accuracy. This solution solves the problem of modeling error accumulation caused by ignoring dynamic changes in sharpness and brightness in traditional methods, providing a reliable quality assessment basis for real-time 3D motion modeling in multi-camera collaborative scenarios.

[0031] Preferably, the step of constructing a detection reliability model based on the inter-frame change rate of the distance between adjacent keypoints and the actual number of detected keypoints, and then outputting the detection reliability, is as follows: The keypoint quantity index is obtained by processing the ratio of the actual number of keypoints detected to the total number of keypoints. The inter-frame change rate of the distance between adjacent key points is subjected to maximum-minimum normalization to obtain the inter-frame change rate exponent. A detection reliability model is constructed based on the keypoint quantity index and the inter-frame change rate index, and the detection reliability is output. The detection reliability model is expressed as follows: in, Indicates the reliability of the test. Indicates the number index of key points. This represents the inter-frame change rate exponent. The coefficient representing the rate of change sensitivity is... The higher the value, the more reliable the detection.

[0032] The keypoint quantity index refers to the ratio of the actual number of detected keypoints to the total number of keypoints. Specifically, it can be calculated by comparing the number of keypoints output by the target detection algorithm with the preset total number of keypoints. This reflects the completeness of the detection and avoids subsequent coordinate fusion errors due to missed detections. The inter-frame change rate index is the normalized value of the distance change between keypoints in adjacent frames. Specifically, it can be calculated by comparing the absolute value of the coordinate difference of the same keypoint between adjacent frames with the preset maximum change threshold. This quantifies motion continuity and eliminates the impact of different motion amplitudes on stability assessment. The detection reliability model is a fractional relationship model based on the keypoint quantity index and the inter-frame change rate index. Specifically, it can use a nonlinear function to combine the quantity index as a positive factor and the change rate index as a negative factor. The sensitivity coefficient is used to adjust the degree of suppression of reliability by dynamic stability, thereby balancing detection coverage and motion continuity.

[0033] Specifically, the keypoint quantity index reflects the completeness of detection in the current frame. When the number of detected points approaches the theoretical total, the index approaches 1, indicating sufficient detection coverage. The inter-frame change rate index captures sudden changes in motion or jitter. When the distance between keypoints in adjacent frames changes beyond the normal range, this index increases, indicating decreased dynamic stability. The detection reliability model uses the quantity index as the numerator to enhance the contribution of complete detection, while using the change rate index as the denominator to suppress the impact of abnormal fluctuations. The sensitivity coefficient can adjust the strength of the change rate's suppression of reliability. For example, when the sensitivity coefficient is set to 0.5, for every unit increase in the change rate index, the detection reliability will decrease by approximately 33%. Thus, this model can dynamically quantify detection reliability, providing an objective basis for multi-camera weight allocation.

[0034] Compared to existing technologies, traditional methods typically rely solely on the number of keypoints or a single motion feature to assess detection quality, neglecting the impact of inter-frame dynamic stability on reliability. For example, existing technologies may directly use the proportion of detected objects as a reliability indicator, but fail to identify coordinate jumps caused by abrupt changes in motion. This solution introduces an inter-frame rate of change exponent, combines normalization processing to eliminate interference from differences in motion amplitude, and constructs a fractional model to dynamically balance detection integrity and motion continuity, effectively addressing the weight allocation bias problem caused by insufficient assessment of detection stability.

[0035] Through the above technical solution, this application can systematically evaluate the completeness and dynamic stability of key point detection, and output objective detection reliability indicators through a quantitative model. This indicator can accurately reflect the credibility of data from each perspective in multi-camera collaborative modeling, thereby optimizing the weight allocation strategy during 3D coordinate fusion, avoiding model distortion caused by missed detections or sudden changes in motion, and ultimately improving the stability and accuracy of motion modeling.

[0036] Preferably, the step of constructing the imaging quality confidence score based on the image quality state coefficient under the distance between the camera and the detection target and the detection reliability is as follows: The distance between the camera and the target is processed by maximum-min normalization to obtain the distance index; An imaging quality confidence model is constructed based on the image quality state coefficient under the distance index and the detection reliability, and the imaging quality confidence score is output. The imaging quality confidence model is expressed as follows: in, Indicates confidence level in image quality. Represents the image quality status coefficient. Indicates the reliability of the test. Indicates the distance attenuation coefficient. Represents the distance index, the Furthermore, the larger the value, the higher the confidence level of the image quality.

[0037] The distance index refers to the transformation of the physical distance between the camera and the target into a dimensionless standardized parameter through max-min normalization. Specifically, a linear scaling method can be used to map the original distance to the [0,1] interval to eliminate the influence of the difference in dimensions between different cameras on the calculation.

[0038] The image quality status coefficient is a comprehensive index that reflects the sharpness and brightness of an image. Specifically, it can be generated by fusing the sharpness index and brightness index after normalization through an exponential function, and is used to quantify the imaging quality of a single frame image.

[0039] Detection reliability refers to a parameter that reflects the stability of keypoint detection. It can be calculated from the number of keypoints and the inter-frame change rate and is used to evaluate the robustness of the detection algorithm.

[0040] The distance attenuation coefficient is an adjustment parameter that controls the impact of distance on the confidence level of image quality. Specifically, it can be a preset fixed value or dynamically adjusted according to the scene, and is used to adjust the weight ratio of distance factors in the confidence level calculation.

[0041] Imaging quality confidence refers to the quantitative evaluation result of three factors: image quality, detection stability, and spatial distance. Specifically, it is generated by combining a product form with an exponential decay function and is used to dynamically adjust the weights of multi-camera fusion.

[0042] Specifically, the distance between the camera and the target is first processed by max-min normalization to unify the distance data from different cameras to the range of [0,1], forming a distance index. Subsequently, the image quality status coefficient was... With detection reliability Multiply by, then multiply by an exponential function β controls the intensity of distance attenuation. As the distance exponent increases, the exponential function value decreases non-linearly, significantly reducing the confidence level of the image quality of long-distance cameras. This calculation method ensures that the confidence level reflects not only the sharpness and brightness of the image itself, but also the stability of the detection algorithm and the influence of the camera's spatial position, ultimately outputting an image quality confidence value in the range [0,1].

[0043] Compared to existing technologies, traditional methods do not incorporate camera distance as an independent variable into the image quality evaluation system, relying solely on image quality or detection results for weight allocation. This leads to distant cameras being assigned excessively high weights even when imaging is blurry but detection is stable. This solution introduces a distance exponent and an exponential decay function to couple spatial location parameters with image quality and detection reliability. This allows for the quantification and evaluation of issues such as decreased image resolution and perspective distortion caused by increased distance, thereby automatically reducing the weight of distant cameras during the coordinate fusion stage.

[0044] Through the above technical solution, this application solves the problem of image quality assessment deviation caused by changes in camera distance. By normalizing the distance parameter and using an exponential decay mechanism, the weight allocation of a multi-camera system can dynamically adapt to changes in the spatial position of the target object. For example, in a sports training scenario, when an athlete moves away from a certain camera, the confidence level of that camera automatically decreases, preventing its low-quality image data from interfering with 3D model reconstruction, thereby improving the accuracy of multi-view data fusion and the stability of motion modeling.

[0045] Preferably, the dynamic weight model is expressed as: in, Indicates the first Dynamic weights of each camera This represents the lumpedity parameter (which controls the degree of concentration of the weight distribution; the larger the value, the greater the weight of the high-confidence camera). Indicates the first Confidence level of image quality for each camera.

[0046] Dynamic weight refers to the contribution ratio of each camera in 3D coordinate fusion. Specifically, it can be achieved by combining exponential function and normalization. By exponential transformation, the weight difference of high-confidence cameras is amplified, and then normalization is used to ensure that the sum of weights is 1, thereby solving the problem of unreasonable weight allocation of multiple cameras.

[0047] The lumped parameter refers to an adjustable coefficient that controls the degree of concentration of the weight distribution. It can be achieved through preset or dynamic adjustment. For example, different values ​​can be set according to the needs of the application scenario. The larger the value, the higher the weight of high-confidence cameras, thereby strengthening the decision weight of high-quality data sources.

[0048] Among them, the image quality confidence score refers to a comprehensive evaluation index of the image quality of the camera. Specifically, it can be achieved by multiplying the image quality state coefficient, detection reliability, and distance attenuation factor. It is used to quantify the data credibility of different cameras in the current environment and provide a basis for weight allocation.

[0049] Specifically, this technical solution uses an exponential function to non-linearly transform the image quality confidence level, significantly increasing the weight of high-confidence cameras. Normalization ensures that the sum of all weights is 1, preventing bias in the fusion result due to uneven weight distribution. The introduction of a lumped parameter allows adjustment of the steepness of the weight distribution according to actual needs. For example, in scenarios requiring high-precision modeling, this parameter can be increased, making the system more reliant on data from a few high-quality cameras; in scenarios requiring balanced multi-view data, this parameter can be decreased, making the weight distribution more gradual. This dynamic adjustment mechanism enables the system to automatically adapt to different imaging conditions. For instance, when some cameras produce blurred images due to excessively distant targets, their confidence level decreases, and their weights are suppressed, thereby reducing the interference of low-quality data on the fusion result.

[0050] Compared to existing technologies, traditional methods typically employ fixed weights or linear weighting, which fail to effectively distinguish the differences in imaging quality between different cameras, leading to low-quality data impacting modeling accuracy. Our proposed solution, however, achieves non-linear weight allocation through an exponential function, combined with adjustable lumped parameters. This allows for more precise amplification of the contribution from high-quality data sources, while normalization maintains numerical stability, significantly improving the robustness and scene adaptability of multi-camera data fusion.

[0051] Through the above technical solution, this application can dynamically adjust the weights of each camera during the multi-camera 3D coordinate fusion process based on the real-time calculated image quality confidence level, effectively suppressing interference from low-quality data sources and improving the accuracy of the fused coordinates. For example, in sports training scenarios, when athletes move quickly, causing some cameras to blur their images, the system automatically reduces the weights of these cameras and prioritizes high-quality data from stable perspectives, thereby ensuring that the reconstructed 3D motion model is closer to the real posture.

[0052] Preferably, the steps for constructing a basic personnel state model based on personnel age and the ratio of sitting height to standing height, and outputting the basic personnel state coefficients, are as follows: The age of the personnel and the ratio of sitting height to standing height were subjected to maximum-min normalization to obtain the age index and limb proportion index. A basic personnel status model is constructed based on age index and limb proportion index, and the basic personnel status coefficient is output. The basic personnel status model is expressed as follows: in, Represents the basic status coefficient of personnel. Indicates age index. Indicates the body proportion index. Represents the weight coefficient and The Furthermore, the higher the value, the better the basic condition of the personnel.

[0053] Among these, age refers to the actual age of the subject being tested, which can be obtained using ID card information or user input data, reflecting the impact of aging on motor flexibility and strength. The sitting-to-standing height ratio is the ratio of sitting height to standing height, which can be collected using a 3D scanner or anthropometry equipment, representing the impact of limb proportions on movement range. Maximum-minimum normalization involves linearly mapping the raw data to the [0,1] interval according to the maximum and minimum values, eliminating the dimensional differences between age and limb proportions. The age index is the normalized age parameter, which can be standardized by setting an age range, such as 18-60 years, to quantify the negative impact of age on baseline performance. The limb proportion index is the normalized sitting-to-standing height ratio parameter, which can be standardized by setting a standard proportion range, such as 0.52-0.56, to represent the positive contribution of limb coordination to movement completion. The weighting coefficient refers to the weight of the influence of age and limb proportions on baseline performance, which can be expressed using empirical values, such as... and The allocation is used to regulate the importance of different physiological parameters.

[0054] Specifically, the process begins by acquiring data on the individual's age and sitting-to-standing height ratio through measuring devices or user input. The age data is mapped to a preset age range; for example, if the minimum age is set to 18 and the maximum to 60, a 30-year-old subject will receive an age index of 0.285. The sitting-to-standing height ratio is then normalized according to a standard human proportion range; for example, if the ratio is 0.54 and the standard range is 0.52-0.56, the limb proportion index is 0.5. Subsequently, the age index is converted... The format is designed to reflect the negative impact of aging; for example, the converted value for the 30-year-old subject mentioned above is 0.715. Finally, the converted age index is combined with the limb proportion index using a weighted summation formula. For example, when the weighting coefficients are 0.6 and 0.4 respectively, the subject's baseline state coefficient is 0.6 × 0.715 + 0.4 × 0.5 = 0.629. This coefficient is input into the movement completion model to correct for the evaluation results of parameters such as joint angle deviation, thereby reflecting the differences in the impact of different individuals' physiological characteristics on movement standardization.

[0055] Compared to existing technologies, current human motion modeling methods typically assume that the subject is in a standard physiological state, failing to consider deviations in movement amplitude caused by age-related decreases in joint mobility or abnormal limb proportions. For example, elderly subjects may exhibit slower movement rhythms due to muscle weakness, but existing technologies cannot provide personalized corrections for such deviations in the model. This approach constructs a quantified baseline state coefficient, transforming age and limb proportion parameters into calculable correction factors, enabling dynamic adjustments to motion completion assessments based on the subject's physiological characteristics.

[0056] Through the above technical solution, this application can effectively eliminate the interference of individual physiological differences on motion modeling. For example, in sports training scenarios, for athletes with special limb proportions, the system can automatically increase the tolerance threshold for their motion amplitude deviation; in medical rehabilitation scenarios, for elderly patients with limited joint mobility, the system can dynamically adjust the evaluation weight of motion rhythm deviation. This personalized correction mechanism makes the motion analysis results more consistent with the actual physiological conditions of the test subjects, improving the objectivity and adaptability of motion assessment.

[0057] Preferably, the steps for constructing an environmental state model based on ambient temperature and ambient oxygen content, and outputting environmental state coefficients, are as follows: The temperature deviation index is obtained by comparing the absolute difference between the ambient temperature and the optimal temperature with the allowable deviation from the optimal temperature. The oxygen content deviation index is obtained by comparing the absolute difference between the ambient oxygen content and the optimum oxygen content with the allowable deviation from the optimum oxygen content. An environmental state model is constructed based on the temperature deviation index and the oxygen content deviation index, and the environmental state coefficient is output. The environmental state model is expressed as follows: in, Represents the environmental state coefficient. This indicates the temperature deviation index. This indicates the oxygen deviation index. Indicates the temperature sensitivity coefficient. The oxygen sensitivity coefficient is represented by the following. The higher the value, the better the environmental condition.

[0058] The temperature deviation index refers to the degree of deviation between the ambient temperature and the preset optimal temperature. Specifically, it can be calculated by dividing the absolute difference between the measured ambient temperature and the optimal temperature by the maximum allowable temperature deviation threshold. This index quantifies the interference of temperature on motion modeling; a larger deviation indicates a worse environmental condition. The oxygen deviation index refers to the degree of deviation between the ambient oxygen content and the preset optimal oxygen content. Specifically, it can be calculated by dividing the absolute difference between the measured oxygen content and the optimal value by the maximum allowable oxygen content deviation threshold. This index measures the impact of abnormal oxygen content on motion completion; a larger value indicates more unfavorable environmental conditions. The environmental state model is a mathematical expression for calculating the environmental state coefficient through a weighted combination of the temperature and oxygen deviation indices. Specifically, it can be implemented using a reciprocal function, introducing temperature sensitivity coefficients and oxygen sensitivity coefficients into the denominator to adjust the contribution of different environmental factors, so that the environmental state coefficient decreases as the degree of deviation increases. The temperature sensitivity coefficient and oxygen sensitivity coefficient can be set empirically.

[0059] Specifically, during motion modeling, fluctuations in ambient temperature and oxygen levels can alter human motion characteristics. For example, low temperatures may induce muscle stiffness, while low-oxygen environments may reduce exercise endurance. By comparing measured temperature and oxygen levels with their optimal values ​​and calculating the ratio of deviation to the allowable deviation range, differences in physical dimensions can be eliminated. The temperature deviation index and oxygen deviation index are multiplied by their corresponding sensitivity coefficients, summed, and then mapped to the 0-1 interval using a reciprocal function to form the environmental state coefficient. When the temperature or oxygen deviation increases, the denominator increases, leading to a decrease in the environmental state coefficient. This automatically reduces the scoring weight of unfavorable environmental conditions in subsequent motion completion assessments, avoiding misjudgments due to environmental interference.

[0060] In some specific implementations, the optimal temperature can be set to 25℃, with an allowable deviation threshold of ±5℃. In this case, the temperature deviation index is calculated as |T-25| / 5. The optimal oxygen content can be set to 20.9%, with an allowable deviation threshold of ±2%, and the oxygen deviation index is |O-20.9| / 2. The temperature sensitivity coefficient b1 and oxygen sensitivity coefficient b2 can be calibrated experimentally, for example, by taking values ​​of 0.8 and 1.2 respectively in high-temperature and high-altitude environments to enhance the influence of changes in oxygen content.

[0061] Compared to existing technologies, traditional methods do not incorporate environmental parameters into the motion modeling and evaluation system, relying solely on visual data for analysis. This solution, however, establishes a quantitative environmental state model, transforming temperature and oxygen content changes into calculable correction coefficients for the first time, thus resolving the problem of motion feature distortion caused by environmental interference. Through normalization and sensitivity coefficient adjustment, it ensures both equivalent processing capability for different environmental parameters and retains adaptive adjustment space for specific scenarios.

[0062] Through the above technical solution, this application can effectively eliminate the negative impact of abnormal ambient temperature and insufficient oxygen content on motion modeling. In high-temperature operations or high-altitude training scenarios, the system can automatically reduce the environmental state coefficient, avoiding misjudging motion deformation caused by the environment as insufficient motion completion, thereby improving the objectivity and accuracy of the evaluation results. At the same time, the configurability of the sensitivity coefficient allows the model to adapt to the application needs of different geographical environments, such as strengthening the weight of oxygen content monitoring in high-altitude areas.

[0063] Preferably, the steps for constructing a motion completion model based on the personnel's basic state coefficient and environmental state coefficient (the maximum absolute deviation of each joint angle from the standard value), the range of motion deviation (the maximum absolute deviation of each joint's range of motion from the standard range), and the rhythm of motion deviation (the maximum angular deviation of each motion node from the standard time node), and outputting the motion completion score are as follows: The joint angle deviation, the range of motion deviation, and the rhythm of motion deviation are each compared with their respective maximum permissible values ​​to obtain the joint angle deviation index, the range of motion deviation index, and the rhythm of motion deviation index. A motion completion model is constructed based on the joint angle deviation index, motion amplitude deviation index, and motion rhythm deviation index under the personnel's basic state coefficient and environmental state coefficient, and the motion completion score is output. The motion completion model is expressed as follows: in, Indicates the degree of completion of the action. Represents the basic status coefficient of personnel. Represents the environmental state coefficient. Indicates the joint angle deviation index. Indicates the deviation index of movement amplitude. This indicates the deviation index of movement rhythm. This represents the sensitivity coefficient to joint angle deviation. This represents the sensitivity coefficient to deviations in the range of motion. The sensitivity coefficient for movement rhythm deviation is represented by the following. The higher the value, the higher the completion rate of the action.

[0064] The joint angle deviation index is the ratio of the maximum absolute deviation of each joint angle from the standard value to the preset maximum allowable value. It can be achieved by dividing the actual angle deviation by a preset threshold. This index quantifies the degree to which a joint angle deviates from the standard value, and can be calculated using the Euler angle difference of joint rotation angles in a three-dimensional coordinate system. It is used to quantify the spatial difference between joint posture and standard movement. The movement amplitude deviation index is the ratio of the maximum absolute deviation of each joint's range of motion from the standard range to the preset maximum allowable value. It can be achieved by calculating the ratio of the actual amplitude deviation to the upper limit of the allowable deviation. This index reflects the proportion of movement amplitude exceeding the standard range, and can be achieved by comparing the displacement of the joint movement trajectory in three-dimensional space. The movement rhythm deviation index is the ratio of the maximum angle deviation of each movement node from the standard time node to the preset maximum allowable value. It can be achieved by dividing the absolute value of the time deviation by the allowable time error threshold. This index characterizes the matching degree between the movement rhythm and the standard timing sequence, and can be achieved using the timing alignment method of keyframe angle changes in time series analysis. The basic physical condition coefficient refers to a physiological state indicator calculated using an individual's age and the ratio of sitting height to standing height. Specifically, it can be generated by weighted summation of normalized age and limb proportions. This coefficient reflects the potential impact of an individual's physiological conditions on motor performance. The environmental condition coefficient refers to an environmental impact indicator calculated using ambient temperature and oxygen content. Specifically, it can be generated by weighted combination of temperature deviation index and oxygen deviation index. This coefficient quantifies the degree of interference from the external environment on motor performance.

[0065] Specifically, firstly, ratio processing is used to convert joint angle deviation, movement amplitude deviation, and movement rhythm deviation into dimensionless standardized indices, eliminating the influence of different physical dimensions on the evaluation results. Then, the human baseline state coefficient and the environmental state coefficient are multiplied as the basic factor for movement completion, reflecting the combined influence of individual physiological conditions and environmental factors on movement performance. Further, an exponential function is used to combine the three standardized deviation indices with a sensitivity coefficient to form a deviation penalty term, utilizing the exponential decay characteristic to amplify the negative impact of deviation on completion. Finally, the movement completion model dynamically links human, environmental factors, and the deviation penalty term through a multiplicative relationship, ensuring that the evaluation results include both objective movement data and subjective physiological and environmental parameters.

[0066] Compared to existing technologies, current motion assessment methods rely solely on static threshold judgments based on absolute deviations in joint angles, amplitudes, and rhythms, failing to consider changes in motor ability due to age differences or the influence of environmental factors on motion execution. This proposed solution quantifies individual physiological differences using a baseline state coefficient and characterizes external environmental interference using an environmental state coefficient. This allows for dynamic adjustments to the assessment criteria based on actual personnel conditions and environmental conditions, resolving the issue of traditional methods' inaccurate assessment results due to neglecting individual differences and environmental interference.

[0067] Through the above technical solutions, this application can effectively eliminate motion assessment errors caused by differences in age, limb proportions, and fluctuations in environmental temperature and oxygen content, thereby improving the objectivity and adaptability of motion completion assessment. For example, in high-temperature and low-oxygen environments, the completion score for the same motion deviation will be lowered accordingly due to the decrease in the environmental condition coefficient, more realistically reflecting the actual impact of harsh environments on motion execution. Simultaneously, for older individuals or those with unusual limb proportions, the scoring benchmark is dynamically adjusted through the individual's baseline condition coefficient, avoiding misjudgments caused by physiological limitations and making the assessment results more closely reflect actual performance capabilities.

[0068] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus.

[0069] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A real-time 3D motion modeling method, characterized in that, Includes the following steps: An image quality status model is constructed based on image sharpness and image brightness, and the image quality status coefficient is output. A detection reliability model is constructed based on the inter-frame change rate of the distance between adjacent key points and the actual number of key points detected, and the detection reliability is output. An imaging quality confidence level is constructed based on the image quality state coefficient and detection reliability under the distance between the camera and the detection target. A dynamic weight model is constructed based on the image quality confidence level, and the weight of each camera's 3D coordinates in the preset 3D coordinate fusion model is output.

2. The real-time 3D motion modeling method according to claim 1, characterized in that, It also includes the following steps: A dynamic human body model is generated based on the 3D coordinates of key human body points generated by the 3D coordinate fusion model. The dynamic human body model is then compared with a database of 3D models of preset standard movements to obtain joint angle deviations, movement amplitude deviations, and movement rhythm deviations. A basic state model of personnel is constructed based on their age and the ratio of their sitting height to their standing height, and the basic state coefficients of personnel are output. An environmental state model is constructed based on ambient temperature and ambient oxygen content, and the environmental state coefficient is output. A motion completion model is constructed based on the joint angle deviation, motion amplitude deviation, and motion rhythm deviation under the basic state coefficients of personnel and environmental state coefficients, and the motion completion degree is output.

3. The real-time 3D motion modeling method according to claim 1, characterized in that, The 3D coordinate fusion model is represented as follows: in, Indicates the fusion of the first Three-dimensional coordinate vectors of each joint point Indicates the total number of cameras. Indicates the first The estimated number of cameras Three-dimensional coordinate vectors of each joint point Indicates the first Dynamic weights of each camera and , Indicates the first Confidence level of image quality for each camera.

4. The real-time 3D motion modeling method according to claim 3, characterized in that, The dynamic weight model is expressed as follows: in, Indicates the first Dynamic weights of each camera Represents lumped parameters, Indicates the first Confidence level of image quality for each camera.

5. The real-time 3D motion modeling method according to claim 4, characterized in that, The steps for constructing imaging quality confidence based on the image quality state coefficient and detection reliability under the distance between the camera and the target are as follows: The distance between the camera and the target is processed by maximum-min normalization to obtain the distance index; An imaging quality confidence model is constructed based on the image quality state coefficient under the distance index and the detection reliability, and the imaging quality confidence score is output. The imaging quality confidence model is expressed as follows: in, Indicates confidence level in image quality. Represents the image quality status coefficient. Indicates the reliability of the test. Indicates the distance attenuation coefficient. Represents the distance index, the Furthermore, the larger the value, the higher the confidence level of the image quality.

6. The real-time 3D motion modeling method according to claim 5, characterized in that, The steps for constructing a detection reliability model based on the inter-frame change rate of the distance between adjacent keypoints and the actual number of keypoints detected are as follows: The keypoint quantity index is obtained by processing the ratio of the actual number of keypoints detected to the total number of keypoints. The inter-frame change rate of the distance between adjacent key points is subjected to maximum-minimum normalization to obtain the inter-frame change rate exponent. A detection reliability model is constructed based on the keypoint quantity index and the inter-frame change rate index, and the detection reliability is output. The detection reliability model is expressed as follows: in, Indicates the reliability of the test. Indicates the number index of key points. This represents the inter-frame change rate exponent. The coefficient representing the rate of change sensitivity is... The higher the value, the more reliable the detection.

7. The real-time 3D motion modeling method according to claim 5, characterized in that, The steps for constructing an image quality status model based on image sharpness and image brightness, and then outputting the image quality status coefficient, are as follows: The image sharpness and image brightness are processed by maximum-minimum normalization to obtain the image sharpness index and image brightness index; An image quality state model is constructed based on the image sharpness index and the image brightness index, and the image quality state coefficient is output. The image quality state model is expressed as follows: in, Represents the image quality status coefficient. Indicates the image sharpness index. Indicates the image brightness index. This represents the image sharpness sensitivity coefficient. The image brightness sensitivity coefficient is represented by the following: The larger the value, the better the image quality.

8. The real-time 3D motion modeling method according to claim 2, characterized in that, Based on the personnel's basic state coefficient and the environmental state coefficient, a motion completion model is constructed using joint angle deviation, motion amplitude deviation, and motion rhythm deviation. The steps for outputting the motion completion score are as follows: The joint angle deviation, the range of motion deviation, and the rhythm of motion deviation are each compared with their respective maximum permissible values ​​to obtain the joint angle deviation index, the range of motion deviation index, and the rhythm of motion deviation index. A motion completion model is constructed based on the joint angle deviation index, motion amplitude deviation index, and motion rhythm deviation index under the personnel's basic state coefficient and environmental state coefficient, and the motion completion score is output. The motion completion model is expressed as follows: in, Indicates the degree of completion of the action. Represents the basic status coefficient of personnel. Represents the environmental state coefficient. Indicates the joint angle deviation index. Indicates the deviation index of movement amplitude. This indicates the deviation index of movement rhythm. This represents the sensitivity coefficient to joint angle deviation. This represents the sensitivity coefficient to deviations in the range of motion. The sensitivity coefficient for movement rhythm deviation is represented by the following. The higher the value, the higher the completion rate of the action.

9. The real-time 3D motion modeling method according to claim 8, characterized in that, The steps for constructing an environmental state model based on ambient temperature and oxygen content, and outputting the environmental state coefficients, are as follows: The temperature deviation index is obtained by comparing the absolute difference between the ambient temperature and the optimal temperature with the allowable deviation from the optimal temperature. The oxygen content deviation index is obtained by comparing the absolute difference between the ambient oxygen content and the optimum oxygen content with the allowable deviation from the optimum oxygen content. An environmental state model is constructed based on the temperature deviation index and the oxygen content deviation index, and the environmental state coefficient is output. The environmental state model is expressed as follows: in, Represents the environmental state coefficient. This indicates the temperature deviation index. This indicates the oxygen deviation index. Indicates the temperature sensitivity coefficient. The oxygen sensitivity coefficient is represented by the following. The higher the value, the better the environmental condition.

10. The real-time 3D motion modeling method according to claim 8, characterized in that, The steps for constructing a basic state model of individuals based on their age and the ratio of their sitting height to their standing height, and then outputting the basic state coefficients, are as follows: The age of the personnel and the ratio of sitting height to standing height were subjected to maximum-min normalization to obtain the age index and limb proportion index. A basic personnel status model is constructed based on age index and limb proportion index, and the basic personnel status coefficient is output. The basic personnel status model is expressed as follows: in, Represents the basic status coefficient of personnel. Indicates age index. Indicates the body proportion index. Represents the weight coefficient and The Furthermore, the higher the value, the better the basic condition of the personnel.

Images