Exercise action standard guidance method based on computer vision

Abstract

The invention relates to the technical field of crossing of machine vision and motion guidance, in particular to a motion action standard guidance method based on computer vision, which comprises the following steps of: firstly, acquiring an image frame sequence with a timestamp, and extracting actual physical coordinates of key points related to force application; the method comprises the following steps: firstly, generating a force, then quantifying the force through'speed-acceleration deduction '(dynamic motion) or'displacement variance + attitude deviation calculation' (static motion) according to motion types, and finally, generating guidance suggestions in combination with a standard force interval, so that accurate quantification and targeted guidance of the force are realized, and force evaluation requirements of different motion types are met.

Description

Technical Field

[0001] This invention relates to the field of interdisciplinary technology of machine vision and motion guidance, and in particular to a method for guiding motion specifications based on computer vision. Background Technology

[0002] Currently, computer vision technology has been applied in many fields. In the field of motion guidance, numerous motion guidance solutions based on this technology have been proposed and implemented. These solutions mostly collect motion images of users, use computer vision to identify key feature points such as human joints, and then compare them with the feature point positions of standard movements to determine the standardization of the user's movements and generate corresponding guidance.

[0003] However, existing solutions have significant shortcomings in recognizing the intensity of movement. They not only have a vague perception of the force exerted, but also lack unified and practical quantitative evaluation standards. For sports where the movements themselves are easy to complete but require specific force exertion, the limitations of existing solutions are particularly pronounced, making it difficult to achieve accurate recognition and scientific guidance of movement intensity.

[0004] Therefore, there is an urgent need for a method of guiding movement standards that focuses on recognizing the force of movement. Summary of the Invention

[0005] The technical problem to be solved by this invention is to provide a computer vision-based method for guiding the standardization of movement, which can achieve precise quantification and targeted guidance of force exertion, and adapt to the force assessment needs of different types of sports.

[0006] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows: A computer vision-based method for guiding motion standardization includes the following steps: S1. Based on a preset frequency, user motion images are used, and a unique timestamp is assigned to each frame image to obtain an image frame sequence; S2. The pixel coordinates of the image frame sequence are calibrated and key points are extracted to obtain the actual physical coordinates of the key points in each frame image. The key points include the joints of the limbs exerting force and the mechanical contact points; S3. Obtain the user's current movement type. If it is dynamic force exertion movement, proceed to the dynamic force exertion movement evaluation step; otherwise, proceed to the static force exertion evaluation step. Dynamic force exertion exercise assessment steps: For two consecutive frames in the image frame sequence, the instantaneous linear velocity of each key point is calculated to obtain the local motion velocity of each key point. Then, according to the key point weight corresponding to the motion type, the local motion velocity of each key point is weighted and fused to obtain the overall motion velocity. Calculate the overall acceleration based on the individual overall motion velocities; The force exerted is calculated based on the overall acceleration, combined with the effective mass of the force-generating part, the motion contact coefficient, and the motion type coefficient. Static force exertion assessment steps: Calculate the three-dimensional coordinate variance of the key points in a preset number of consecutive frames of the image frame sequence to obtain the inter-frame displacement variance; The force exertion is calculated based on the inter-frame displacement variance combined with the basic maintenance force and attitude deviation coefficient of the static force exertion motion. S4. Based on the force exertion and the standard force range corresponding to the current movement type, generate action specification guidance for the user.

[0007] To solve the above-mentioned technical problems, another technical solution adopted by the present invention is as follows: A storage medium storing a computer program thereon, which, when executed, implements the steps in the above-described computer vision-based motion guidance method.

[0008] The beneficial effects of this invention are as follows: This invention provides a computer vision-based method for guiding motion specifications. First, it collects a sequence of image frames with timestamps, extracts the actual physical coordinates of key points related to force application, and then quantifies the force application intensity according to the type of motion by either "velocity-acceleration deduction" (dynamic motion) or "displacement variance + posture deviation calculation" (static motion). Finally, it generates guidance opinions by combining standard force ranges, thus achieving accurate quantification and targeted guidance of force application intensity, and adapting to the force assessment needs of different types of motion. Attached Figure Description

[0009] Figure 1 This is a flowchart of a computer vision-based motion guidance method according to an embodiment of the present invention. Detailed Implementation

[0010] To explain in detail the technical content, objectives, and effects of the present invention, the following description is provided in conjunction with the embodiments and accompanying drawings.

[0011] Please refer to Figure 1 One embodiment of the present invention is as follows: A computer vision-based method for guiding motion standardization includes the following steps: In this embodiment, the user selects the type of exercise (such as squat / plank / badminton swing / dumbbell curl) on the device, and the system retrieves the basic parameters of the corresponding movement from the standard database: including key detection joints, standard frame rate, movement phase division (such as "squat down → push off the ground → stand up" for squat), and basic force standard range.

[0012] S1. Based on a preset frequency, user motion images are used, and a unique timestamp is assigned to each frame image to obtain an image frame sequence.

[0013] In this embodiment, a continuous video stream of user movement is acquired through an edge vision device (mobile phone camera, motion sensor camera, action camera), and continuous image frames are extracted at a set frame rate to form a frame sequence. F ={ F 1, F 2, F 3,..., F n}, where n is the number of frames captured in real time, and the time difference between frames is a fixed value Δt (determined by the standard frame rate, such as Δt=1 / 50s for 50fps).

[0014] In this embodiment, each frame in the frame sequence undergoes light preprocessing to eliminate interference from the acquisition environment and ensure the accuracy of subsequent keypoint detection. The step preceding step S2 also includes the following steps: Preprocessing is performed on each frame of the image frame sequence, including: Background cropping is performed on the image based on human detection technology; Gaussian filtering is used to eliminate ambient light and shadow noise, and histogram equalization is used to improve the contrast of human body contours and joints.

[0015] S2. The pixel coordinates of the image frame sequence are calibrated and key points are extracted to obtain the actual physical coordinates of the key points in each frame image. The key points include the joints of the limbs exerting force and the mechanical contact points; The key points are extracted using a preset key point detection model and then filtered according to the motion type.

[0016] In this embodiment, depending on the device performance, a lightweight keypoint detection model (such as MobilePose / HRNet-Lite) or a high-precision model (such as HRNet / ViTPose) can be used to achieve keypoint detection.

[0017] Filter key points according to the type of exercise. For example, for squats, select the key points corresponding to "hip joint, knee joint, ankle joint, sacrum". For dumbbell curls, select the key points corresponding to "shoulder joint, elbow joint, wrist joint, dumbbell center of gravity". For badminton swings, select the key points corresponding to "shoulder joint, elbow joint, wrist joint, racket head".

[0018] Continuous tracking is performed on each keypoint in the valid image frame sequence, recording the real-time 3D coordinates of each keypoint from the pixel coordinate system to the world coordinate system (the transformation from pixel coordinates to world coordinates is achieved through perspective transformation of a monocular camera / stereo matching of a binocular camera), resulting in the inter-frame coordinate sequence of each keypoint P={P1(x1,y1,z1),P2(x2,y2,z2),...,P n (x n ,y n ,z n )}.

[0019] S3. Obtain the user's current movement type. If it is dynamic force exertion movement, proceed to the dynamic force exertion movement evaluation step; otherwise, proceed to the static force exertion evaluation step. Dynamic force exertion exercise assessment steps: For two consecutive frames in the image frame sequence, the instantaneous linear velocity of each key point is calculated to obtain the local motion velocity of each key point. Then, according to the key point weight corresponding to the motion type, the local motion velocity of each key point is weighted and fused to obtain the overall motion velocity.

[0020] In this embodiment, the calculation is based on the time difference between frames and the three-dimensional coordinate changes of key points. It is divided into two layers: local joint velocity and overall motion velocity. It retains the velocity characteristics of local force-generating parts and integrates them into the velocity index of the overall motion, providing a quantitative basis for force inference.

[0021] The instantaneous linear velocity is calculated as follows: ; in, v i Let be the instantaneous linear velocity of the key point in the i-th frame, ( x i , y i , z i )as well as( x i+1 , y i+1 , z i+1 ) represent the actual physical coordinates of the key point in the i-th frame and the (i+1)-th frame, respectively, and ∆t represents the time difference between frames; Based on the keypoint weights corresponding to the motion type, the local joint velocities of each keypoint are weighted and fused to obtain the overall motion velocity, expressed by the formula: ; in, V Indicates the overall speed of the movement. m This indicates the total number of key points. w k This represents the keypoint weight of the k-th keypoint based on the preset keypoint weight according to the motion type. v' ik The smooth speed representing the local key is obtained by performing a moving average filter on the local motion speed of a consecutive preset number of frames.

[0022] In this embodiment, the smooth velocity of each local joint is weighted and summed according to the force exertion weight of the movement type to obtain the overall movement velocity V. For example, the core force exertion of squat is the lower limb, and the weights assigned to the hip joint / knee joint / ankle joint are 0.4, 0.35, and 0.25, respectively; the core force exertion of dumbbell curl is the upper limb, and the weights assigned to the elbow joint / wrist joint / dumbbell center of gravity are 0.5, 0.3, and 0.2, respectively.

[0023] The entire phase of the movement (from squatting to standing up) is divided into frame intervals, and the overall movement speed of each phase is calculated separately, such as the "squatting phase speed" and "pushing off and standing up phase speed" of the squat, to adapt to the force judgment requirements of different phases.

[0024] Calculate the overall acceleration based on the individual overall motion velocities.

[0025] In this embodiment, the inter-frame acceleration is calculated from the overall motion speed difference between adjacent frames and the inter-frame time difference: a =( V i+1 - V i ) / ∆ t ; The overall acceleration can be the average of all inter-frame accelerations.

[0026] The force exerted is calculated based on the overall acceleration, combined with the effective mass of the force-generating part, the motion contact coefficient, and the motion type coefficient. Based on the overall acceleration, and combined with the effective mass of the force-generating part, the motion contact coefficient, and the motion type coefficient, the force is calculated and expressed as: F = m × a × k 1× k 2+ Ff ; in, F Indicates the force exerted, where m represents the effective mass of the exerting body part, automatically estimated from the user's body parameters (e.g., for squats, the effective mass is the lower limbs, approximately 60%-70% of body weight; for dumbbell curls, it's the upper limbs plus the total mass of the machine). a Indicates overall acceleration. k 1 indicates a preset exercise type coefficient based on the type of exercise (e.g., bodyweight exercises without equipment). k 1=1, exercise with equipment k 1 = 1.2-1.5, explosive power sports k 1 = 1.5 - 2.0). k 2 indicates the preset contact coefficient based on the type of exercise (e.g., push-off exercise). k 2=1.1, gripping the equipment exercise k 2 = 1.05, non-contact motion such as swinging the racket k 2=1), F f The resistance level is determined by the equipment parameters (e.g., resistance band resistance, dumbbell weight). If no equipment is available, then... F f It is 0.

[0027] In this embodiment, based on the extension of Newton's second law (F=ma), and combined with the limb mass, rate of change of velocity (acceleration), and contact coefficient of motion (such as the contact damping of pushing off the ground and gripping the equipment) of human movement, the force F is finally obtained.

[0028] Static force exertion assessment steps: Static force exertion shows no obvious inter-frame velocity changes; the core is to infer the force through the stability of inter-frame displacement at key points and the posture deviation maintained by muscle exertion.

[0029] Calculate the three-dimensional coordinate variance of the key points in a preset number of consecutive frames of the image frame sequence to obtain the inter-frame displacement variance; The force exertion is calculated based on the inter-frame displacement variance combined with the basic maintenance force and attitude deviation coefficient of the static force exertion motion. Based on the inter-frame displacement variance, combined with the basic maintenance force and attitude deviation coefficient of the static force generation motion, the force generation force is calculated and expressed as: F = F 0×1 / S × k 3; in, F Indicates the force exerted. F 0 represents the basic strength required to maintain a static movement, which is determined based on the type of exercise (e.g., the basic strength required for an adult male to maintain a plank is approximately 40% of his body weight).S This represents the variance of inter-frame displacement. k 3 represents the attitude deviation coefficient, which is obtained by mapping the deviation between the user's attitude and the standard attitude.

[0030] The determination of the attitude deviation coefficient includes the following steps: Based on the pre-entered user torso length, the actual physical coordinates of the key points are normalized to obtain normalized coordinates.

[0031] In this embodiment, eliminating individual differences caused by user height, limb proportions, and camera acquisition distance / angle is the core premise for posture deviation quantification, and normalization using the human body local coordinate system is adopted: (1) Take the core reference joint of the static movement as the origin O (e.g., select the sacral joint for plank and the hip joint for wall squat). (2) Establish a user-specific local three-dimensional coordinate system with the trunk axis as the Y-axis (e.g., using the vector from the sacral joint to the shoulder joint), the horizontal direction of the coronal plane of the body as the X-axis, and the direction of the sagittal plane perpendicular to the XY plane as the Z-axis; (3) Transform the world coordinates of all core nodes of the user into this local coordinate system to obtain normalized coordinates; (4) Using the user's torso length (e.g., the Euclidean distance from the sacrum to the shoulder joint) as the scaling factor, scale all normalized coordinates proportionally (scaled to 1 unit length) to obtain standardized normalized coordinates without individual differences, and establish a coordinate set. Ensure that the postures of users with different heights / limb proportions can be directly compared.

[0032] Based on the normalized coordinates of the key points, a three-dimensional vector is constructed, and the three-dimensional joint angles are calculated. For each three-dimensional joint angle, the average of each frame is calculated, and the average joint angle set of each joint is constructed based on the average calculation results.

[0033] In this embodiment, the calculation of three-dimensional joint angles is achieved through the vector angle formula. For example, for each core joint pair (such as joint A - joint B - joint C, where B is called the core pivot joint), normalized coordinates are extracted. P A , P B as well as P C Vectors are constructed based on normalized coordinates. Then, the three-dimensional joint angle of joint B is calculated using the formula for the angle between three-dimensional vectors.

[0034] Perform the above calculations on all core joint pairs and average them across frames to obtain the average joint angle set of the user's real-time pose, such as... A user ={a user1 , a user2 , a user3 ,..., a usern}

[0035] The deviation rate of each three-dimensional joint angle in the average joint angle set is calculated by comparing it with the standard joint angle in the preset standard joint angle set, thus obtaining the local posture deviation rate.

[0036] In this embodiment, the user's actual joint angles are compared one by one with the standard posture joint angles, and the local deviation rate of each group of joints is calculated to eliminate the absolute value difference of single angle deviations (e.g., the allowable deviation of a 90° joint is 4.5°, while the allowable deviation of a 120° joint is 6°; unifying these into a deviation rate makes it easier to quantify). The formula is: ; in, r i This represents the local attitude deviation rate of the i-th joint. a useri This represents the average joint angle of the i-th joint. a stdi This represents the standard joint angle of the i-th joint. T i This represents the allowable deviation threshold for the i-th joint (preset in a standard database and determined by professional sports standards, such as 90° for the elbow joint in a plank exercise). T i =4.5°; hip joint 180°, T i =9°).

[0037] like r i ≤100%, take the actual calculated value; if r i >100%, uniformly taken r i =100% (indicating that the joint posture deviates significantly from the standard).

[0038] The overall posture deviation rate is calculated by combining the local posture deviation rate corresponding to each joint with the local weights preset based on the motion type.

[0039] In this embodiment, the local deviation rates of each joint are weighted and summed based on the static force application weights to obtain the overall deviation rate of the user's real-time posture (core logic: different joints contribute differently to posture stability during static force application, and the weights are tilted towards the core force application joints; this weight is a common setting in existing methods and can be retrieved from a standard database). The formula is expressed as: ; in, R This represents the overall attitude deviation rate. w i This represents the force exertion weight of the i-th joint (determined by the type of exercise, such as in plank where the force exertion weight of "shoulder and elbow joints" is 0.3, "trunk and shoulder joints" is 0.1, "hip and knee joints" is 0.05, and "knee and ankle joints" is 0.05).

[0040] The attitude deviation coefficient is obtained by mapping according to the value of the overall attitude deviation rate through a preset mapping rule.

[0041] In this embodiment, if R ≤5% (overall posture deviation ≤5%) k 3 = 1.0 (Standard posture, no reduction in power efficiency); If 5% < R ≤15% (overall posture deviation 5%~15%) k 3. Use linear interpolation of 0.8~0.9 (the larger the deviation, the lower the value). k The smaller 3 is, such as R =10% k 3 = 0.85 R =15% k 3=0.8); like R >15% (overall posture deviation >15%) k 3. Take a range of values ​​from 0.5 to 0.7 (graded according to the degree of deviation, such as 15% < R ≤30% k 3 = 0.7, 30% < R ≤50% k When 3 = 0.6, R > 50% k 3 = 0.5).

[0042] Linear interpolation calculation (if 5% < R When ≤15%, the formula is expressed as: k 3 = 0.9 - [( R -5%)×0.1] / 10%.

[0043] S4. Based on the force exertion and the standard force range corresponding to the current movement type, generate action specification guidance for the user.

[0044] In this embodiment, the calculated force F is smoothed over time (by moving average over 10 consecutive frames) to eliminate the force jump caused by instantaneous speed fluctuations, thus obtaining the effective force F', which serves as the final basis for subsequent standard interval comparisons.

[0045] Based on a standard database and personalized calibration, the standard range [F_min, F_max] of the user's force for this movement is determined. The effective force F' is compared with the range to accurately judge whether the force meets the standard, and at the same time, the type of force abnormality is identified. Establishment of standard intensity range: Baseline range: The intensity value derived from the standard movements of professional athletes / experienced coaches is used as the benchmark. The baseline value is ±10% as the basic standard range (competitive training), and ±15% as the standard range for general fitness. Personalized adjustment: The adaptive calibration module adjusts the base range based on the user's body parameters (e.g., children / elderly reduce F_min), exercise level (beginners reduce F_min), and training goals (muscle gain increases F_min) to obtain the final personalized standard range; Assessment of whether the intensity meets the standard: Qualification: F'∈[F_min,F_max], and inter-frame intensity fluctuation ≤10%; Slight deficiency: F'∈[0.8F_min,F_min), the intensity does not meet the minimum standard; Severe deficiency: F' < 0.8F_min, indicating a significant lack of power generation; Mild overextension: F'∈(F_max,1.2F_max], the force exceeds the upper limit, which can easily lead to sports injuries; Severely exceeding the limit: F'>1.2F_max, excessive force applied, needs immediate correction; Abnormal force fluctuation: F'∈[F_min,F_max], but inter-frame force fluctuation >10%, indicating uneven force application.

[0046] Based on the force assessment results, the system provides users with multi-dimensional, personalized, real-time guidance and feedback. The feedback format is adapted to the output capabilities of the edge-side acquisition device (e.g., visual + auditory + text for mobile devices, and voice for action cameras). The guidance suggestions are specific to the force application area and the method of action adjustment, avoiding vague descriptions. Visual feedback: In the real-time video on the device, key points of force application are marked with colors (green = meeting the standard, yellow = slightly insufficient / exceeding the standard, red = severely insufficient / exceeding the standard / abnormal fluctuation), and the real-time force value + standard range is displayed on the side of the screen, presenting the force change trend in a dynamic bar chart; Auditory feedback: Concise voice prompts are used to match the real-time nature of the movements, such as "slightly insufficient force in squat push-off", "excessive force in dumbbell curl elbow joint", "even force in plank, force meets standard". Text / image feedback: Detailed adjustment suggestions are displayed on the end interface, specifying "force-generating parts + movement adjustment methods", such as "the force of squatting is seriously insufficient: it is recommended to increase the force of the quadriceps muscles in the legs, increase the speed of hip sinking, and quickly push off the ground when squatting down to 90° of the knee joint"; Action phase review: After a user completes a set of actions, a force analysis report can be generated for the user, including force change curves at each stage, abnormal frame locations, and weak points in force exertion, providing a basis for review for subsequent training.

[0047] Embodiment 2 of the present invention is as follows: A storage medium storing a computer program thereon, which, when executed, implements the steps of the computer vision-based motion guidance method described in Embodiment 1 above.

[0048] This invention provides a computer vision-based method for guiding proper movement, focusing on the accurate identification and quantification of movement force. Through image preprocessing, targeted key point extraction, and coordinate calibration, it constructs scientific force calculation models for both dynamic and static scenarios. By combining movement type-specific weights, coefficients, and standard ranges, it achieves accurate quantification of force exertion for different movement types. Simultaneously, it provides real-time guidance and movement review from multiple dimensions, including visual, auditory, and textual feedback. This addresses the shortcomings of existing solutions in terms of ambiguous force perception while also considering personalization and practicality. It adapts to different users' body parameters, exercise levels, and training goals, improving the scientific rigor and safety of exercise guidance and helping users standardize their movements and avoid injuries.

[0049] The above description is merely an embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent modifications made based on the content of the present invention specification and drawings, or direct or indirect applications in related technical fields, are similarly included within the patent protection scope of the present invention.

Claims

1. A method for guiding motion standardization based on computer vision, characterized in that, Including the following steps: S1. Based on a preset frequency, user motion images are used, and a unique timestamp is assigned to each frame image to obtain an image frame sequence; S2. The pixel coordinates of the image frame sequence are calibrated and key points are extracted to obtain the actual physical coordinates of the key points in each frame image. The key points include the joints of the limbs exerting force and the mechanical contact points; S3. Obtain the user's current movement type. If it is dynamic force exertion movement, proceed to the dynamic force exertion movement evaluation step; otherwise, proceed to the static force exertion evaluation step. Dynamic force exertion exercise assessment steps: For two consecutive frames in the image frame sequence, the instantaneous linear velocity of each key point is calculated to obtain the local motion velocity of each key point. Then, according to the key point weight corresponding to the motion type, the local motion velocity of each key point is weighted and fused to obtain the overall motion velocity. Calculate the overall acceleration based on the individual overall motion velocities; The force exerted is calculated based on the overall acceleration, combined with the effective mass of the force-generating part, the motion contact coefficient, and the motion type coefficient. Static force exertion assessment steps: Calculate the three-dimensional coordinate variance of the key points in a preset number of consecutive frames of the image frame sequence to obtain the inter-frame displacement variance; The force exertion is calculated based on the inter-frame displacement variance combined with the basic maintenance force and attitude deviation coefficient of the static force exertion motion. S4. Based on the force exertion and the standard force range corresponding to the current movement type, generate action specification guidance for the user.

2. The method for guiding motion standardization based on computer vision according to claim 1, characterized in that, The instantaneous linear velocity is calculated as follows: ； in, v i Let be the instantaneous linear velocity of the key point in the i-th frame, ( x i , y i , z i )as well as( x i+1 , y i+1 , z i+1 ) represent the actual physical coordinates of the key point in the i-th frame and the (i+1)-th frame, respectively, and ∆t represents the time difference between frames.

3. The method for guiding motion standardization based on computer vision according to claim 1, characterized in that, Based on the keypoint weights corresponding to the motion type, the local joint velocities of each keypoint are weighted and fused to obtain the overall motion velocity, expressed by the formula: ； in, V Indicates the overall speed of the movement. m This indicates the total number of key points. w k This represents the keypoint weight of the k-th keypoint based on the preset keypoint weight according to the motion type. v' ik The smooth speed representing the local key is obtained by performing a moving average filter on the local motion speed of a consecutive preset number of frames.

4. The method for guiding motion standardization based on computer vision according to claim 1, characterized in that, Based on the overall acceleration, and combined with the effective mass of the force-generating part, the motion contact coefficient, and the motion type coefficient, the force is calculated and expressed as: F = m × a × k 1× k 2+ F f ； in, F This indicates the force applied, where m represents the effective mass of the force-applying part, automatically estimated from the user's body parameters. a Indicates overall acceleration. k 1 indicates a preset exercise type coefficient based on the exercise type. k 2 indicates the preset motion contact coefficient based on the type of motion. F f This represents resistance, obtained from the instrument's parameters; if no instrument is available, then... F f It is 0.

5. The method for guiding motion standardization based on computer vision according to claim 1, characterized in that, Based on the inter-frame displacement variance, combined with the basic maintenance force and attitude deviation coefficient of the static force generation motion, the force generation force is calculated and expressed as: F = F 0×1 / S × k 3； in, F Indicates the force exerted. F 0 represents the basic sustaining force for a static movement, which is determined based on the type of movement. S This represents the variance of inter-frame displacement. k 3 represents the attitude deviation coefficient, which is obtained by mapping the deviation between the user's attitude and the standard attitude.

6. The method for guiding motion standardization based on computer vision according to claim 5, characterized in that, The determination of the attitude deviation coefficient includes the following steps: Based on the pre-entered user torso length, the actual physical coordinates of the key points are normalized to obtain normalized coordinates; Based on the normalized coordinates of the key points, a three-dimensional vector is constructed, and the three-dimensional joint angles are calculated. For each three-dimensional joint angle, the average of each frame is calculated, and the average joint angle set of each joint is constructed based on the average calculation results. The deviation rate of each three-dimensional joint angle in the average joint angle set is calculated by comparing it with the standard joint angle in the preset standard joint angle set to obtain the local posture deviation rate. The overall posture deviation rate is calculated by combining the local posture deviation rate corresponding to each joint with the local weights preset based on the motion type. The attitude deviation coefficient is obtained by mapping according to the value of the overall attitude deviation rate through a preset mapping rule.

7. The method for guiding motion standardization based on computer vision according to claim 1, characterized in that, Step S2 is preceded by the following steps: Preprocessing is performed on each frame of the image frame sequence, including: Background cropping is performed on the image based on human detection technology; Gaussian filtering is used to eliminate ambient light and shadow noise, and histogram equalization is used to improve the contrast of human body contours and joints.

8. The method for guiding motion standardization based on computer vision according to claim 1, characterized in that, The key points are extracted using a preset key point detection model and then filtered according to the motion type.

9. A storage medium having a computer program stored thereon, characterized in that, When the computer program is executed, it implements the steps of the computer vision-based motion guidance method as described in any one of claims 1-8.

Images