Volleyball training methods and systems based on AI posture capture

Abstract

This invention relates to the field of sports training technology, and more particularly to a volleyball training method and system based on AI posture capture. It includes the following steps: a hybrid posture capture system collects training data from athletes in real time during volleyball training; an AI processing unit analyzes the training data based on a pre-set database of volleyball movements from athletes with different skill levels and physical conditions to obtain analytical data; a personalized training plan is generated based on the analytical data, combined with the athlete's historical data and individual characteristics; the personalized training plan is fed back to the athlete in real time via a visual interactive method, and simultaneously, the athlete is given explanations and guidance via voice interaction based on the personalized training; the athlete conducts volleyball training based on the visual and voice interactions. This invention, through the acquired analytical data, makes the generated personalized training plan more scientific and effective, thereby improving the training effect for athletes during volleyball training.

Description

Technical Field

[0001] This invention relates to the field of sports training technology, and in particular to a volleyball training method and system based on AI posture capture. Background Technology

[0002] In modern volleyball training, traditional training methods and techniques face numerous challenges. Currently, most volleyball training data collection relies on single devices, such as using ordinary cameras to record athlete movements. This method results in data with limited dimensions, failing to comprehensively acquire three-dimensional posture data, movement trajectory data, and joint angle change data during training, making in-depth analysis of athlete movements difficult. Furthermore, the movement analysis phase lacks multi-dimensional comprehensive consideration, simply comparing movement forms without combining mechanical models to analyze the force generation and hitting point of the hitting action, and failing to consider the rationality of athlete positioning and running in actual tactical scenarios. This leads to an inability to accurately pinpoint the root cause of athlete movement problems, resulting in a lack of scientific rigor and effectiveness in training guidance.

[0003] In terms of training program development, current technologies often adopt uniform training standards and models, neglecting the individual differences among athletes in terms of physical fitness, technical level, learning ability, and other unique characteristics. This makes it difficult for training programs to fully meet the actual needs of each athlete, potentially leading to excessive training intensity that could harm the athlete's health, or insufficient training intensity that may fail to effectively improve the athlete's performance, thus making it difficult to maximize training results.

[0004] In summary, given the technical problem of low training effectiveness for athletes in volleyball training in the prior art, the actual technical problem solved by the present invention is how to improve the training effectiveness for athletes in volleyball training. Summary of the Invention

[0005] To overcome the technical shortcomings of existing technologies in terms of low training effectiveness for athletes during volleyball training, the present invention aims to provide a volleyball training method and system based on AI posture capture. By analyzing the force generation method and hitting point information of the hitting action based on a mechanical model, as well as the athlete's running and standing information in combination with tactical scenarios, the generated personalized training plan is more scientific and effective, thereby improving the training effect for athletes during volleyball training.

[0006] This invention discloses a volleyball training method based on AI posture capture, comprising the following steps:

[0007] A hybrid attitude capture system is constructed using a fisheye camera, a high-definition camera, and an inertial sensor. The system collects training data of athletes in real time during volleyball training. The training data includes at least three-dimensional attitude information, motion trajectory information, and joint angle change information.

[0008] The AI ​​processing unit analyzes the training data based on a pre-set database of volleyball movements of athletes with different skill levels and physical conditions to obtain analytical data. The analytical data is obtained by analyzing the force generation method and hitting point of the hitting action based on a mechanical model, as well as the athlete's running and standing information in combination with tactical scenarios.

[0009] Personalized training plans are generated based on the analyzed data and combined with the athlete's historical data and individual characteristics.

[0010] Personalized training plans are fed back to athletes in real time through visual interaction. At the same time, the athletes are given explanations and guidance through voice interaction based on the personalized training. Athletes conduct volleyball training based on visual and voice interaction.

[0011] Preferably, when the fisheye camera, high-definition camera, and inertial sensor are installed, a unified three-dimensional coordinate system is established through a spatial calibration algorithm, so that the three-dimensional posture information, motion trajectory information, and joint angle change information collected by the fisheye camera, high-definition camera, and inertial sensor are at the same spatial reference.

[0012] The spatial calibration algorithm is based on the Zhang calibration method combined with the calibration parameters of the inertial sensor for joint calibration.

[0013] Preferably, the fisheye camera has a horizontal viewing angle greater than or equal to 180° and a vertical viewing angle greater than or equal to 150°, and the fisheye camera is equipped with an image distortion correction algorithm to perform real-time distortion correction on the fisheye image when the fisheye camera acquires the fisheye image.

[0014] The high-definition camera has a resolution of 4K or higher and a frame rate of 60fps or higher.

[0015] The inertial sensor has an angular error of less than ±0.5° and an acceleration error of less than ±2.1 m / s². 2 .

[0016] Preferably, a fisheye camera is used to acquire panoramic data of the training venue, a high-definition camera is used to acquire motion data of the athletes, and an inertial sensor is used to acquire joint motion data of the athletes.

[0017] After preprocessing the panoramic data, motion data, and joint motion data, three-dimensional pose information, motion trajectory information, and joint angle change information are obtained based on the preprocessed panoramic data, and / or motion data, and / or joint motion data.

[0018] Preprocessing includes noise reduction and data compression.

[0019] Preferably, in the step of analyzing training data based on a pre-set volleyball action database containing athletes with different skill levels and physical conditions by an AI processing unit, the step further includes analyzing the angular velocity ω and angular acceleration α at the moment of impact when the athlete hits the volleyball, so as to evaluate the stability of the athlete's hitting action based on the angular velocity ω and angular acceleration α. ​​When the angular velocity ω is greater than the angular velocity threshold, and / or when the angular acceleration α is greater than the angular acceleration threshold, it is determined that the stability of the athlete's hitting action is insufficient.

[0020] The formula for calculating angular velocity ω is: The formula for calculating angular acceleration α is:

[0021] Where Δθ represents the change in joint angle and Δt represents the time interval.

[0022] Preferably, after the step of generating a personalized training plan based on the analyzed data and combined with the athlete's historical data and individual characteristics, the following steps are also included:

[0023] Based on personalized training plans, athletes' historical data, and individual characteristics, a training effect prediction report is generated. The analytic hierarchy process (AHP) is used to assign weights to various performance indicators during volleyball training to construct a comprehensive evaluation model. This model is then used to calculate a quantitative evaluation score for the athlete's training effectiveness. The scoring formula is as follows:

[0024]

[0025] Among them, w i Let x represent the weight of the i-th action indicator. i Let represent the score of the i-th action indicator, and n represent the total number of action indicators involved in the comprehensive evaluation calculation.

[0026] Preferably, the comprehensive evaluation model includes a comprehensive scoring threshold;

[0027] When the comprehensive evaluation model calculates the athlete's quantitative evaluation score, it compares the quantitative evaluation score with the comprehensive scoring threshold. If the quantitative evaluation score is less than the comprehensive scoring threshold, training data is reacquired to generate a new personalized training plan.

[0028] In view of this, a second objective of the present invention is to provide a volleyball training system based on AI posture capture, comprising a data acquisition module, an analysis module, a training scheme generation module, and a training module; wherein,

[0029] Acquisition module: Used to acquire training data of athletes in volleyball training in real time based on a hybrid attitude capture system consisting of a fisheye camera, a high-definition camera and an inertial sensor. The training data includes at least three-dimensional attitude information, motion trajectory information and joint angle change information.

[0030] Analysis module: Connected to the acquisition module, it is used by the AI ​​processing unit to analyze training data based on a pre-set database of volleyball movements of athletes with different skill levels and physical conditions to obtain analysis data. The analysis data is obtained by analyzing the force generation method and hitting point information of the hitting action based on the mechanical model, as well as the athlete's running and standing information combined with the tactical scenario.

[0031] Training program generation module: Connected to the analysis module, it is used to generate personalized training programs based on the analyzed data and combined with the athlete's historical data and individual characteristics.

[0032] Training Module: Connected to the training plan generation module, it is used to provide real-time feedback of personalized training plans to athletes in a visual and interactive manner. At the same time, it provides explanations and guidance to athletes through voice interaction based on personalized training. Athletes conduct volleyball training based on visual and voice interaction.

[0033] Preferably, it further includes a scoring calculation module; wherein,

[0034] Scoring Calculation Module: Connected to the training plan generation module, this module generates a training effect prediction report based on personalized training plans, athletes' historical data, and individual characteristics. It uses the analytic hierarchy process (AHP) to weight various performance indicators during volleyball training to construct a comprehensive evaluation model. Based on this model, it calculates a quantitative evaluation score for the athlete's training effectiveness. The scoring formula is as follows:

[0035]

[0036] Among them, w i Let x represent the weight of the i-th action indicator. i Let represent the score of the i-th action indicator, and n represent the total number of action indicators involved in the comprehensive evaluation calculation.

[0037] Preferably, the comprehensive evaluation model includes a comprehensive scoring threshold. The scoring calculation module is also used to compare the quantitative evaluation score with the comprehensive scoring threshold when the comprehensive evaluation model calculates the athlete's quantitative evaluation score. When the quantitative evaluation score is less than the comprehensive scoring threshold, training data is re-acquired to generate a new personalized training plan.

[0038] Compared with existing technologies, the beneficial effects of this invention, after adopting the above technical solution, are as follows: By analyzing the force generation method and hitting point information of the hitting action based on a mechanical model, and combining the athlete's running and standing information with tactical scenarios, the generated personalized training plan is more scientific and effective, thereby improving the training effect of athletes in volleyball training. Furthermore, by combining the historical data and personalized characteristics of individual athletes with the analyzed data to generate personalized training plans, the targeting of individual athletes is further enhanced, thus improving the training effect of individual athletes in volleyball training. In addition, this invention utilizes a hybrid attitude capture system composed of a fisheye camera, a high-definition camera, and an inertial sensor to deeply analyze the athlete's movements, thereby improving the accuracy of the analyzed data and indirectly improving the effectiveness of the personalized training plan, thus further enhancing the training effect. Attached Figure Description

[0039] Figure 1 This is a schematic diagram illustrating the steps of a volleyball training method and system based on AI posture capture according to the present invention. Detailed Implementation

[0040] The advantages of the present invention will be further illustrated below with reference to the accompanying drawings and specific embodiments.

[0041] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numerals in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this disclosure. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this disclosure as detailed in the appended claims.

[0042] The terminology used in this disclosure is for the purpose of describing particular embodiments only and is not intended to be limiting of the disclosure. The singular forms “a,” “the,” and “the” as used in this disclosure and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used herein refers to and includes any and all possible combinations of one or more of the associated listed items.

[0043] It should be understood that although the terms first, second, third, etc., may be used in this disclosure to describe various information, such information should not be limited to these terms. These terms are used only to distinguish information of the same type from one another. For example, without departing from the scope of this disclosure, first information may also be referred to as second information, and similarly, second information may also be referred to as first information. Depending on the context, the word "if" as used herein may be interpreted as "when," "when," or "in response to determination."

[0044] In the description of this invention, it should be understood that the terms "longitudinal", "lateral", "up", "down", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0045] In the description of this invention, unless otherwise specified and limited, it should be noted that the terms "installation", "connection" and "linking" should be interpreted broadly. For example, they can refer to mechanical or electrical connections, or internal connections between two components. They can be direct connections or indirect connections through an intermediate medium. Those skilled in the art can understand the specific meaning of the above terms according to the specific circumstances.

[0046] In the following description, suffixes such as "module," "part," or "unit" used to denote elements are used only for the convenience of the description of the invention and have no specific meaning in themselves. Therefore, "module" and "part" can be used interchangeably.

[0047] This embodiment provides a volleyball training method based on AI posture capture, including the following steps: A hybrid posture capture system is constructed using a fisheye camera, a high-definition camera, and an inertial sensor. This system is used to collect real-time training data from athletes during volleyball training. The training data includes at least three-dimensional posture information, motion trajectory information, and joint angle change information. An AI processing unit, pre-loaded with a database of volleyball movements from athletes of varying skill levels and physical conditions, analyzes the training data to obtain analytical data. This analytical data is obtained by analyzing the force application method and hitting point information of the hitting action based on a mechanical model, as well as the athlete's running and positioning information in conjunction with tactical scenarios. A personalized training plan is generated based on the analytical data, combined with the athlete's historical data and individual characteristics. The personalized training plan is then fed back to the athlete in real-time via a visual interactive method. Simultaneously, the personalized training is explained and guided to the athlete via voice interaction. The athlete then conducts volleyball training based on both the visual and voice interactions.

[0048] See Figure 1 As shown in the figure, this embodiment will describe in detail a volleyball training method based on AI posture capture, which specifically includes the following steps:

[0049] Step S100: A hybrid attitude capture system, consisting of a fisheye camera, a high-definition camera, and an inertial sensor, is used to collect training data from athletes in real time during volleyball training. Specifically, the fisheye camera has a horizontal viewing angle greater than or equal to 180° and a vertical viewing angle greater than or equal to 150°, and is deployed at the top of the training court to acquire panoramic information. The high-definition camera has a resolution greater than or equal to 4K and a frame rate greater than or equal to 60fps, and is positioned around the perimeter of the training court to focus on the details of the athletes' movements. The inertial sensor has a measurement accuracy with an angle error of less than ±0.5° and an acceleration error of less than ±2.1 m / s². 2 This includes, but is not limited to, wearing on key joints such as the shoulders, elbows, and wrists of athletes.

[0050] It should be noted that this training data includes at least three-dimensional posture information, motion trajectory information, and joint angle change information. The three-dimensional posture information is obtained by fusing triangulation data from fisheye and high-definition cameras with information acquired by inertial sensors to construct the three-dimensional coordinates of the athlete's joints, thus acquiring three-dimensional posture data. The motion trajectory information is obtained by tracking the athlete's movement trajectory in the training field in real time using a fisheye camera combined with a field coordinate system. The joint angle change information is obtained by calculating the change in joint angle based on the angular velocity ω and angular acceleration α collected in real time by inertial sensors. The calculation method for the joint angle change Δθ is Δθ=ω·Δt.

[0051] It should be noted that the court coordinate system is a reference frame used to determine the positions of athletes and volleyballs within the training court. It is established as follows:

[0052] Camera calibration: By setting up calibration objects at known locations in the training area, these calibration objects are photographed using a fisheye camera and a high-definition camera. Image processing algorithms are then used to calculate the internal parameters (such as focal length, principal point, etc.) and external parameters (such as rotation and translation relationships) of the fisheye and high-definition cameras, thereby determining their position and orientation within the field coordinate system. In this way, each pixel in the image captured by the camera can be correlated with its actual position in the field coordinate system through coordinate transformation.

[0053] Field Feature Point Extraction: Identifying fixed feature points in the training field, such as the four corners of the field and the intersection of the center line and sidelines. The positions of these feature points in the field coordinate system are known. By detecting the positions of these feature points in images captured by fisheye cameras and high-definition cameras, a mapping relationship between image coordinates and field coordinates can be established.

[0054] Inertial sensor-assisted positioning: Inertial sensors worn on key joints of the athlete can provide the angular velocity and angular acceleration of the joints. By integrating and processing the angular velocity and angular acceleration, the positional changes of the athlete's joints in the local coordinate system can be obtained. Then, combined with the previously obtained field coordinates, the position of the athlete's joints is transformed into the field coordinate system, enabling precise tracking and positioning of the athlete's movements.

[0055] Step S200: An AI processing unit, based on a pre-set database of volleyball movements containing athletes of varying skill levels and physical conditions, utilizes a deep learning algorithm and a mechanical model to analyze the force application and contact point of the hitting motion. For example, by applying Newton's laws of motion, combined with information on joint angle changes and body posture data, it calculates the magnitude and direction of the force applied at the moment of impact, thus determining the rationality of the force application method. By analyzing the position and trajectory of various body parts during the hit, it determines the accuracy of the contact point. Simultaneously, the AI ​​processing unit analyzes the athlete's running and positioning information in conjunction with the tactical scenario. It compares the athlete's actual running and positioning with the pre-set optimal tactical plan, taking into account the current match situation and the opponent's tactical arrangements, to evaluate the athlete's performance in tactical execution. Through comprehensive analysis of this multi-dimensional information, the AI ​​processing unit obtains analytical data. This analytical data can accurately pinpoint the problems athletes have in technical movements and tactical understanding and execution, providing a scientific basis for generating personalized training plans, making training more targeted, and effectively improving athletes' technical level and tactical literacy.

[0056] Step S300: In this step, the AI ​​processing unit generates a personalized training plan based on the analysis data obtained in step S200, combined with the athlete's historical data and personalized characteristics. Historical data records the athlete's past training performance, technical improvements, and injury recovery information. Analyzing this data allows for an understanding of the athlete's training progress and potential. Personalized characteristics include the athlete's physical attributes (such as height, weight, and fitness level), technical characteristics (such as preferred offensive methods and defensive weaknesses), and learning ability.

[0057] Reinforcement learning algorithms play a crucial role in this process. Based on analyzed data and individual athlete differences, they select the most suitable training program from a large number of combinations of training movements and intensities. For athletes with poor physical fitness but good technical foundations, physical fitness training might be prioritized to gradually improve their endurance and strength, while maintaining appropriate technical training intensity. For athletes with significant technical weaknesses, specific training movements would be designed to address those weaknesses. For example, for athletes with inconsistent serves, a dedicated plan would be developed to correct their serve posture and improve power generation. This personalized training program fully considers the uniqueness of each athlete, maximizing their strengths, compensating for weaknesses, and improving training effectiveness and efficiency.

[0058] Step S400: The generated personalized training plan is fed back to the athlete in real time via a visual interactive method, while explanation and guidance are provided to the athlete through voice interaction. The visual interaction includes, but is not limited to, presenting the training plan through displays, projection equipment, or smart wearable devices worn by the athlete in the training venue. It uses intuitive graphics and animations to show the standard posture of training movements, the progress of the training plan, and the differences between the athlete's current movements and the standard movements, allowing the athlete to quickly and clearly understand their training status and areas for improvement.

[0059] Voice interaction provides athletes with detailed explanations and real-time guidance. During training, the voice interaction module issues timely instructions and suggestions based on the athlete's performance, such as "The arm angle isn't high enough, raise it another 10 degrees" or "Pay attention to the rhythm of power at the moment of impact," helping athletes better understand and execute the training plan. Athletes train volleyball using both visual and voice interaction, receiving timely feedback and adjusting their movements, forming a closed-loop system of "training-feedback-improvement." This makes training more efficient, helps athletes quickly master correct techniques and tactical awareness, and continuously improves their training level and competitive ability.

[0060] Furthermore, during installation, the fisheye camera, HD camera, and inertial sensor establish a unified three-dimensional coordinate system through a spatial calibration algorithm, ensuring that the three-dimensional attitude information, motion trajectory information, and joint angle change information collected by the fisheye camera, HD camera, and inertial sensor are on the same spatial reference; the spatial calibration algorithm is based on Zhang's calibration method combined with the calibration parameters of the inertial sensor for joint calibration.

[0061] In this embodiment, the installation of the hybrid attitude capture system will be described in detail. For example, when installing it in a volleyball court, technicians use the Zhang's calibration method. By placing a calibration board inside the court, fisheye cameras and high-definition cameras capture images of the calibration board from different angles. Combined with the factory calibration parameters of the inertial sensor, joint calibration is performed to construct a unified three-dimensional coordinate system. This ensures that the images captured by the fisheye and high-definition cameras correspond to the joint angle change information collected by the inertial sensor, avoiding data errors caused by inconsistencies in the coordinate system, thereby ensuring the accuracy and consistency of data acquisition.

[0062] Furthermore, the fisheye camera has a horizontal viewing angle greater than or equal to 180° and a vertical viewing angle greater than or equal to 150°, and is equipped with an image distortion correction algorithm to perform real-time distortion correction on the fisheye image during acquisition; the high-definition camera has a resolution greater than or equal to 4K and a frame rate greater than or equal to 60fps; the inertial sensor has a measurement accuracy with an angle error of less than ±0.5° and an acceleration error of less than ±2.1m / s².2 .

[0063] In this embodiment, the fisheye camera, high-definition camera, and inertial sensor of the hybrid attitude capture system will be described in detail.

[0064] Fisheye cameras utilize ultra-wide-angle lenses with a horizontal field of view greater than or equal to 180° and a vertical field of view greater than or equal to 150°. For example, some models boast a horizontal field of view of 200° and a vertical field of view of 160°, capable of covering the entire volleyball training court, allowing the AI ​​processing unit to fully grasp the athletes' positions on the court. Furthermore, its built-in image distortion correction algorithm processes the image immediately after acquisition, restoring normal image distortion caused by wide-angle shooting.

[0065] The high-definition camera has a resolution of 4K (3840×2160) or higher and a frame rate of 60fps or higher. When athletes perform actions such as fast spikes and saves, it can clearly capture details such as the instantaneous force exertion of the athlete's muscles and the subtle movements of their fingers.

[0066] The high precision of inertial sensors ensures the reliability of the data. For example, when measuring the knee joint angle of an athlete, the error is less than ±0.5°, and when measuring acceleration, the error is less than ±0.1 m / s². 2 This enables the system to accurately acquire motion information of the athlete's joints.

[0067] Furthermore, a fisheye camera is used to acquire panoramic data of the training venue, a high-definition camera is used to acquire motion data of the athletes, and an inertial sensor is used to acquire joint motion data of the athletes. After preprocessing the panoramic data, motion data, and joint motion data, three-dimensional posture information, motion trajectory information, and joint angle change information are acquired based on the preprocessed panoramic data, and / or motion data, and / or joint motion data. The preprocessing includes noise reduction and data compression.

[0068] In this embodiment, the camera, high-definition camera, and inertial sensor in the hybrid attitude capture system are described in detail again. The camera, high-definition camera, and inertial sensor generate a large amount of information; therefore, the various acquired information needs to be preprocessed before transmission to improve transmission efficiency. Preprocessing includes, but is not limited to, noise reduction and data compression. For example, the fisheye camera acquires panoramic data of the training field, and Kalman filtering is used to remove noise interference from the panoramic data. The high-definition camera acquires motion data of the athletes, and Kalman filtering is also used to remove noise interference from the motion data. After the inertial sensor acquires the joint motion data of the athletes, the joint motion data is compressed using the H.265 video coding standard and lossless compression algorithm to reduce the pressure during transmission and improve transmission efficiency.

[0069] Furthermore, in the step of analyzing training data using an AI processing unit based on a pre-set database of volleyball movements from athletes with different skill levels and physical conditions, the analysis further includes analyzing the angular velocity ω and angular acceleration α at the moment of impact when the athlete hits the volleyball. The stability of the athlete's hitting motion is assessed based on the angular velocity ω and angular acceleration α. ​​If the angular velocity ω is greater than an angular velocity threshold, and / or if the angular acceleration α is greater than an angular acceleration threshold, the athlete's hitting motion is deemed to be unstable. The formula for calculating the angular velocity ω is: The formula for calculating angular acceleration α is: Where Δθ represents the change in joint angle and Δt represents the time interval.

[0070] In this embodiment, step S200 will be described in detail again. In volleyball training, the stability of the hitting action is crucial to the hitting effect. Simply calculating the force at the moment of impact and judging the deviation of the hitting point cannot comprehensively assess the stability of an athlete's hitting action. For example, two athletes may generate the same force at the moment of impact and have a small deviation in the hitting point, but due to differences in the stability of joint movement during the force exertion process, the quality of the hit may differ significantly. Therefore, introducing angular velocity ω and angular acceleration α to analyze the stability of the hitting action from the perspective of dynamic motion can more accurately identify problems in the athlete's action, thereby developing a more effective training plan. More specifically, when assessing the stability of an athlete's hitting action based on angular velocity ω and angular acceleration α, a judgment is made by setting angle thresholds and angular acceleration thresholds. When the angular velocity ω is greater than the angular velocity threshold, and / or when the angular acceleration α is greater than the angular acceleration threshold, the athlete's hitting action is judged to be unstable.

[0071] The formula for calculating angular velocity ω is: The formula for calculating angular acceleration α is: Where Δθ represents the change in joint angle and Δt represents the time interval.

[0072] Furthermore, after generating a personalized training plan based on the analyzed data and combined with the athlete's historical data and individual characteristics, the process also includes the following steps: generating a training effect prediction report based on the personalized training plan, the athlete's historical data, and individual characteristics; using the analytic hierarchy process (AHP) to assign weights to various action indicators of the athlete during volleyball training to construct a comprehensive evaluation model; and calculating a quantitative evaluation score for the athlete's training effect based on the comprehensive evaluation model. The scoring formula is as follows: Among them, w i Let x represent the weight of the i-th action indicator. i Let represent the score of the i-th action indicator, and n represent the total number of action indicators involved in the comprehensive evaluation calculation.

[0073] In this embodiment, the volleyball training method based on AI posture capture will be described in detail again. Following step S300, step S500 is included: A training effect prediction report is generated using the personalized training plan generated in step S300, along with the athlete's historical data and personalized characteristics. The analytic hierarchy process (AHP) is used to assign weights to various movement indicators of the athlete during volleyball training, thereby constructing a comprehensive prediction model. Finally, a quantitative evaluation score is calculated based on the comprehensive evaluation model to assess the athlete's training effect. The scoring formula is as follows: Among them, w i Let x represent the weight of the i-th action indicator. i Let represent the score of the i-th action indicator, and n represent the total number of action indicators involved in the comprehensive evaluation calculation. For example, let's set the weight of hitting technique to 0.4, the weight of physical fitness to 0.3, and the weight of tactical understanding to 0.3. By calculating the scores of each indicator, for example, if the score of hitting technique is 80 points, the score of physical fitness is 75 points, and the score of tactical understanding is 85 points, then the quantitative evaluation score calculated using the above formula is 81 points.

[0074] Furthermore, the comprehensive evaluation model includes a comprehensive scoring threshold. When the comprehensive evaluation model calculates the athlete's quantitative evaluation score, it compares the quantitative evaluation score with the comprehensive scoring threshold. If the quantitative evaluation score is less than the comprehensive scoring threshold, training data is re-acquired to generate a new personalized training plan.

[0075] In this embodiment, a comprehensive scoring threshold is set within the comprehensive evaluation model. After the quantitative evaluation score is calculated according to the above embodiment, it is compared with the comprehensive scoring threshold. If the quantitative evaluation score is less than the comprehensive scoring threshold, training data is re-acquired to generate a new personalized training plan. That is, for example, if the quantitative evaluation score calculated in the above embodiment is 81 points, and the comprehensive scoring threshold is 80 points, then a new personalized training plan is generated again according to steps 100 to 300. If the quantitative evaluation score is 82 points, it indicates that the personalized training plan is reasonable and can be directly applied to the athlete for training.

[0076] This implementation also provides a volleyball training system based on AI posture capture, including a data acquisition module, an analysis module, a training plan generation module, and a training module. The data acquisition module utilizes a hybrid posture capture system composed of a fisheye camera, a high-definition camera, and an inertial sensor to collect real-time training data of athletes during volleyball training. The training data includes at least three-dimensional posture information, motion trajectory information, and joint angle change information. The analysis module, connected to the data acquisition module, analyzes the training data based on a pre-set AI processing unit containing a database of volleyball movements from athletes of different skill levels and physical conditions. The analysis data is obtained by analyzing the force application method and hitting point information of the hitting action based on a mechanical model, as well as the athlete's running and positioning information in conjunction with tactical scenarios. The training plan generation module, connected to the analysis module, generates personalized training plans based on the analyzed data and combined with the athlete's historical data and individual characteristics. The training module, connected to the training plan generation module, provides real-time feedback of the personalized training plan to the athlete in a visual interactive manner. Simultaneously, it explains and guides the athlete through voice interaction based on the personalized training. The athlete conducts volleyball training based on both visual and voice interaction.

[0077] The data acquisition module is used to collect real-time training data of athletes during volleyball training using a hybrid attitude capture system composed of a fisheye camera, a high-definition camera, and an inertial sensor. Specifically, the fisheye camera has a horizontal viewing angle greater than or equal to 180° and a vertical viewing angle greater than or equal to 150°, and is deployed at the top of the training court to acquire panoramic information. The high-definition camera has a resolution greater than or equal to 4K and a frame rate greater than or equal to 60fps, and is positioned around the perimeter of the training court, focusing on the details of the athletes' movements. The inertial sensor has a measurement accuracy with an angular error of less than ±0.5° and an acceleration error of less than ±2.1 m / s². 2 This includes, but is not limited to, wearing on key joints such as the shoulders, elbows, and wrists of athletes.

[0078] It should be noted that this training data includes at least three-dimensional posture information, motion trajectory information, and joint angle change information. The three-dimensional posture information is obtained by fusing triangulation data from fisheye and high-definition cameras with information acquired by inertial sensors to construct the three-dimensional coordinates of the athlete's joints, thus acquiring three-dimensional posture data. The motion trajectory information is obtained by tracking the athlete's movement trajectory in the training field in real time using a fisheye camera combined with a field coordinate system. The joint angle change information is obtained by calculating the change in joint angle based on the angular velocity ω and angular acceleration α collected in real time by inertial sensors. The formula for calculating the change in joint angle Δθ is: Δθ=ω·Δt.

[0079] It should be noted that the court coordinate system is a reference frame used to determine the positions of athletes and volleyballs within the training court. It is established as follows:

[0080] Camera calibration: By setting up calibration objects at known locations in the training area, these calibration objects are photographed using a fisheye camera and a high-definition camera. Image processing algorithms are then used to calculate the internal parameters (such as focal length, principal point, etc.) and external parameters (such as rotation and translation relationships) of the fisheye and high-definition cameras, thereby determining their position and orientation within the field coordinate system. In this way, each pixel in the image captured by the camera can be correlated with its actual position in the field coordinate system through coordinate transformation.

[0081] Field Feature Point Extraction: Identifying fixed feature points in the training field, such as the four corners of the field and the intersection of the center line and sidelines. The positions of these feature points in the field coordinate system are known. By detecting the positions of these feature points in images captured by fisheye cameras and high-definition cameras, a mapping relationship between image coordinates and field coordinates can be established.

[0082] Inertial sensor-assisted positioning: Inertial sensors worn on key joints of the athlete can provide the angular velocity and angular acceleration of the joints. By integrating and processing the angular velocity and angular acceleration, the positional changes of the athlete's joints in the local coordinate system can be obtained. Then, combined with the previously obtained field coordinates, the position of the athlete's joints is transformed into the field coordinate system, enabling precise tracking and positioning of the athlete's movements.

[0083] Analysis Module: This module utilizes an AI processing unit based on a pre-set database of volleyball movements from athletes of varying skill levels and physical conditions. This database stores a vast amount of diverse movement data, serving as a "knowledge base" for the AI ​​processing unit's analysis. It employs deep learning algorithms and mechanical models to analyze the force application and contact point information of the hitting motion. For example, by applying Newton's laws of motion, combined with information on changes in athlete joint angles and body posture data, it calculates the magnitude and direction of the force applied at the moment of impact, thereby determining the rationality of the force application method. By analyzing the position and trajectory of various parts of the athlete's body during the hit, it determines the accuracy of the contact point. Simultaneously, the AI ​​processing unit analyzes the athlete's running and positioning information in conjunction with the tactical scenario. It considers factors such as the current match situation and the opponent's tactical arrangements, comparing the athlete's actual running and positioning with the pre-set optimal tactical plan to evaluate the athlete's performance in tactical execution. Through comprehensive analysis of this multi-dimensional information, the AI ​​processing unit obtains analytical data. This analytical data can accurately pinpoint the problems athletes have in technical movements and tactical understanding and execution, providing a scientific basis for generating personalized training plans, making training more targeted, and effectively improving athletes' technical level and tactical literacy.

[0084] Training plan generation module: This module is used by the AI ​​processing unit to generate personalized training plans based on the analytical data obtained from the aforementioned analysis module, combined with the athlete's historical data and individual characteristics. Historical data records the athlete's past training performance, technical improvements, and injury recovery information. Analyzing this data allows for an understanding of the athlete's training progress and potential. Individual characteristics include the athlete's physical attributes (such as height, weight, and fitness level), technical characteristics (such as preferred offensive techniques and defensive weaknesses), and learning abilities.

[0085] Reinforcement learning algorithms play a crucial role in this process. Based on analyzed data and individual athlete differences, they select the most suitable training program from a large number of combinations of training movements and intensities. For athletes with poor physical fitness but good technical foundations, physical fitness training might be prioritized to gradually improve their endurance and strength, while maintaining appropriate technical training intensity. For athletes with significant technical weaknesses, specific training movements would be designed to address those weaknesses. For example, for athletes with inconsistent serves, a dedicated plan would be developed to correct their serve posture and improve power generation. This personalized training program fully considers the uniqueness of each athlete, maximizing their strengths, compensating for weaknesses, and improving training effectiveness and efficiency.

[0086] Training Module: This module provides real-time feedback to athletes through a visually interactive approach to the generated personalized training plans, while also offering explanations and guidance via voice interaction. Visual interaction includes, but is not limited to, presentations on displays, projection devices, or smart wearable devices worn by athletes in the training area. It uses intuitive graphics and animations to display standard postures for training movements, the progress of the training plan, and the differences between the athlete's current movements and the standard movements, allowing athletes to quickly and clearly understand their training progress and areas for improvement.

[0087] Voice interaction provides athletes with detailed explanations and real-time guidance. During training, the voice interaction module issues timely instructions and suggestions based on the athlete's performance, such as "The arm angle isn't high enough, raise it another 10 degrees" or "Pay attention to the rhythm of power at the moment of impact," helping athletes better understand and execute the training plan. Athletes train volleyball using both visual and voice interaction, receiving timely feedback and adjusting their movements, forming a closed-loop system of "training-feedback-improvement." This makes training more efficient, helps athletes quickly master correct techniques and tactical awareness, and continuously improves their training level and competitive ability.

[0088] Furthermore, it also includes a scoring calculation module; wherein, the scoring calculation module is connected to the training plan generation module, and is used to generate a training effect prediction report based on the personalized training plan and the athlete's historical data and personalized characteristics, and to use the analytic hierarchy process (AHP) to assign weights to various action indicators of the athlete during volleyball training in order to construct a comprehensive evaluation model, and to calculate a quantitative evaluation score of the athlete's training effect based on the comprehensive evaluation model. The scoring calculation formula is as follows: Among them, w i Let x represent the weight of the i-th action indicator. i Let represent the score of the i-th action indicator, and n represent the total number of action indicators involved in the comprehensive evaluation calculation.

[0089] Following the training plan generation module, a scoring calculation module is also included. This module generates a training effect prediction report based on the personalized training plan generated by the training plan generation module, the athlete's historical data, and personalized characteristics. It then uses the analytic hierarchy process (AHP) to assign weights to various performance indicators during volleyball training, thereby constructing a comprehensive prediction model. Finally, it calculates a quantitative evaluation score for the athlete's training effect based on this comprehensive evaluation model. The scoring formula is as follows: Among them, w i Let x represent the weight of the i-th action indicator. iLet represent the score of the i-th action indicator, and n represent the total number of action indicators involved in the comprehensive evaluation calculation. For example, let's set the weight of hitting technique to 0.4, the weight of physical fitness to 0.3, and the weight of tactical understanding to 0.3. By calculating the scores of each indicator, for example, if the score of hitting technique is 80 points, the score of physical fitness is 75 points, and the score of tactical understanding is 85 points, then the quantitative evaluation score calculated using the above formula is 81 points.

[0090] Furthermore, the comprehensive evaluation model includes a comprehensive scoring threshold. The scoring calculation module is also used to compare the quantitative evaluation score with the comprehensive scoring threshold when the comprehensive evaluation model calculates the athlete's quantitative evaluation score. When the quantitative evaluation score is less than the comprehensive scoring threshold, training data is re-acquired to generate a new personalized training plan.

[0091] In this embodiment, a comprehensive scoring threshold is set within the comprehensive evaluation model. After the quantitative evaluation score is calculated according to the above embodiment, it is compared with the comprehensive scoring threshold. If the quantitative evaluation score is less than the comprehensive scoring threshold, training data is re-acquired to generate a new personalized training plan. That is, for example, if the quantitative evaluation score calculated in the above embodiment is 81 points, and the comprehensive scoring threshold is 80 points, then a new personalized training plan is generated again according to steps 100 to 300. If the quantitative evaluation score is 82 points, it indicates that the personalized training plan is reasonable and can be directly applied to the athlete for training.

[0092] It should be noted that the embodiments of the present invention have better implementability and are not intended to limit the present invention in any way. Any person skilled in the art may use the above-disclosed technical content to change or modify it into equivalent effective embodiments. However, any modifications or equivalent changes and modifications made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solution of the present invention shall still fall within the scope of the technical solution of the present invention.

Claims

1. A volleyball training method based on AI pose capture, characterized in that, Comprising the following steps: A hybrid posture capture system is composed of a fisheye camera, a high-definition camera and an inertial sensor, the fisheye camera is used to obtain panoramic data of the training ground, the high-definition camera is used to obtain action data of the athlete, and the inertial sensor is used to obtain joint movement data of the athlete; After preprocessing the panoramic data, the action data, and the joint movement data, three-dimensional posture information, motion trajectory information, and joint angle change information are obtained according to the preprocessed panoramic data, and / or the action data, and / or the joint movement data; the preprocessing includes denoising and data compression processing; A unified three-dimensional coordinate system is established by a space calibration algorithm during installation of the volleyball training ground, so that the three-dimensional posture information, the motion trajectory information, and the joint angle change information collected by the fisheye camera, the high-definition camera, and the inertial sensor are in the same space reference; the space calibration algorithm is based on Zhang's calibration method combined with the calibration parameters of the inertial sensor for joint calibration; and the training data of the athlete during the volleyball training process is collected in real time under the unified three-dimensional coordinate system based on the hybrid posture capture system, and the training data at least includes three-dimensional posture information, motion trajectory information, and joint angle change information; An AI processing unit, based on a pre-set database of volleyball movements from athletes of varying skill levels and physical conditions, analyzes the training data to obtain analytical data. This analytical data is obtained by analyzing information on the force application method and contact point of the hitting action based on a mechanical model, as well as the athlete's positioning and running information in conjunction with tactical scenarios. The analysis also includes analyzing the angular velocity of the athlete at the moment of impact when hitting the volleyball. and angular acceleration According to the angular velocity and the angular acceleration Assess the stability of the athlete's hitting motion when the angular velocity is greater than an angular velocity threshold, and / or when the angular acceleration... If the angular acceleration exceeds the threshold, the athlete's hitting motion is deemed to be unstable. The angular velocity The calculation formula of the angular acceleration The calculation formula of the angular acceleration ;​ wherein is expressed as a joint angle change amount, is expressed as a time interval; An individualized training program is generated according to the analysis data combined with the historical data and individualized characteristics of the athlete; A training effect prediction report is produced according to the individualized training program and the historical data and individualized characteristics of the athlete, and the hierarchical analysis method is used to weight and distribute each action index of the athlete during the volleyball training to construct a comprehensive evaluation model, and the training effect of the athlete is calculated and quantitatively evaluated according to the comprehensive evaluation model, and the score calculation formula is: ； wherein, a weight assigned to the i-th action indicator, a score assigned to the i-th action indicator, n denotes the total number of action indicators participating in the calculation of the composite score. The quantitative evaluation score is compared with the comprehensive score threshold set in the comprehensive evaluation model, when the quantitative evaluation score is less than the comprehensive score threshold, the training data is reacquired to generate a new individualized training program; The individualized training program is fed back to the athlete in real time in a visual interactive manner, and the athlete is explained and guided according to the individualized training in a voice interaction, and the athlete trains the volleyball according to the visual interaction and the voice interaction.

2. The volleyball training method based on AI posture capture according to claim 1, wherein The horizontal viewing angle of the fisheye camera is greater than or equal to 180°, the vertical viewing angle is greater than or equal to 150°, and an image distortion correction algorithm is configured in the fisheye camera, which corrects the fisheye image in real time when the fisheye camera collects the fisheye image; The resolution of the high-definition camera is greater than or equal to 4K, and the frame rate is greater than or equal to 60fps; The angle error of the measurement accuracy of the inertial sensor is less than ±0.5°, and the acceleration error is less than ±2.1m / s².

3. A volleyball training system applying the volleyball training method based on the AI posture capture according to claim 1, characterized in that, Comprising a collection module, an analysis module, a training program generation module, a training module, and a score calculation module; wherein, The collection module is used for collecting training data of the player in the volleyball training process in real time based on a mixed gesture capture system composed of a fisheye camera, a high-definition camera and an inertial sensor, and the training data at least includes three-dimensional gesture information, motion trajectory information and joint angle change information. The analysis module is connected with the collection module and is used for analyzing the training data based on an AI processing unit of a preset volleyball action database containing players with different technical levels and physical conditions to obtain analysis data, which is obtained by analyzing force mode information, hitting point information of a hitting action based on a mechanical model and running position information and standing position information of the player in combination with a tactical scene. The training scheme generation module is connected with the analysis module and is used for generating a personalized training scheme according to the analysis data and in combination with historical data and individual characteristics of the player. The training module is connected with the training scheme generation module and is used for feeding back the personalized training scheme to the player in a visual interactive mode in real time, and explaining and guiding the player according to the personalized training in a voice interactive mode, so that the player performs volleyball training according to the visual interaction and the voice interaction. The scoring calculation module is connected with the training scheme generation module and is used for generating a training effect estimation report according to the personalized training scheme and the historical data and individual characteristics of the player, performing weight distribution on various action indexes of the player in the volleyball training by using an analytic hierarchy process, constructing a comprehensive evaluation model, and calculating a training effect of the player according to the comprehensive evaluation model to quantitatively evaluate a score, and a scoring calculation formula is as follows: ； wherein, a weight assigned to the i-th action indicator, a score assigned to the i-th action indicator, n denotes the total number of action indicators participating in the calculation of the composite score. The comprehensive evaluation model is internally provided with a comprehensive score threshold, and the scoring calculation module is further used for comparing the quantitative evaluation score with the comprehensive score threshold when the comprehensive evaluation model calculates the quantitative evaluation score of the player, and when the quantitative evaluation score is less than the comprehensive score threshold, the training data is reacquired to generate a new personalized training scheme. The collection module is used for collecting training data of the player in the volleyball training process in real time based on a mixed gesture capture system composed of a fisheye camera, a high-definition camera and an inertial sensor, and the training data at least includes three-dimensional gesture information, motion trajectory information and joint angle change information. The analysis module is connected with the collection module and is used for analyzing the training data based on an AI processing unit of a preset volleyball action database containing players with different technical levels and physical conditions to obtain analysis data, which is obtained by analyzing force mode information, hitting point information of a hitting action based on a mechanical model and running position information and standing position information of the player in combination with a tactical scene. The training scheme generation module is connected with the analysis module and is used for generating a personalized training scheme according to the analysis data and in combination with historical data and individual characteristics of the player. The training module is connected with the training scheme generation module and is used for feeding back the personalized training scheme to the player in a visual interactive mode in real time, and explaining and guiding the player according to the personalized training in a voice interactive mode, so that the player performs volleyball training according to the visual interaction and the voice interaction. The scoring calculation module is connected with the training scheme generation module and is used for generating a training effect estimation report according to the personalized training scheme and the historical data and individual characteristics of the player, performing weight distribution on various action indexes of the player in the volleyball training by using an analytic hierarchy process, constructing a comprehensive evaluation model, and calculating a training effect of the player according to the comprehensive evaluation model to quantitatively evaluate a score, and a scoring calculation formula is as follows: The comprehensive evaluation model is internally provided with a comprehensive score threshold, and the scoring calculation module is further used for comparing the quantitative evaluation score with the comprehensive score threshold when the comprehensive evaluation model calculates the quantitative evaluation score of the player, and when the quantitative evaluation score is less than the comprehensive score threshold, the training data is reacquired to generate a new personalized training scheme.

Images