An automatic evaluation method for aerial technical movements of upright long jump

Abstract

The present invention discloses a method for automatically evaluating aerial techniques in a straight-up long jump, comprising the following steps: S1. determining camera placement and shooting parameters; S2. estimating human posture and extracting characteristic angles; S3. establishing a standard library of aerial techniques; S4. matching aerial techniques; and S5. evaluating aerial techniques. This method quantitatively evaluates a long jumper's take-off and take-off postures, providing quantitative data on the long jumper's straight-up long jump aerial techniques. This data serves as an important reference indicator for performance assessment, reducing the impact of subjective factors in the evaluation. During regular training, long jumpers can also refer to the evaluation results to promptly correct any irregularities in various techniques during the long jump. The quantitative feedback data can also be used to provide targeted guidance to long jumpers, enhancing the potential for personalized training.

Description

Technical Field

[0001] The present invention relates to the field of machine vision technology, in particular to an automatic evaluation method for aerial technical movements of a straight-ahead long jump. Background Art

[0002] As a complex track and field sport, the straight-legged long jump is generally divided into four technical phases: run-up, take-off, flight, and landing. From the moment of take-off from the step to the moment of landing, every body movement has a crucial impact on the long jump performance. Therefore, regular training and assessments often focus on the long jumper's relevant technical movements. However, in the teaching of straight-legged long jump courses in universities, due to differences in teachers' understanding of each detailed movement and the lack of unified standards, student performance assessments often rely on personal experience and lack objectivity. This also lacks quantitative feedback data, making it difficult to implement personalized training for students. Furthermore, professional long jump technical analysis equipment is expensive and difficult to widely use. To address these issues, this method aims to analyze and evaluate student posture based on a human posture estimation algorithm, provide quantitative feedback data, and promptly identify and correct problems with students' long jump technical movements. It also implements an automated assessment system to improve the objectivity of performance assessments.

[0003] In recent years, the use of machine vision for sports analysis has become a research hotspot. For example, methods based on human posture estimation are used to evaluate and count squats and push-ups, assess yoga pose accuracy, analyze swimming posture, and analyze standing long jump posture. This paper addresses these issues in teaching the push-up long jump course by combining machine vision algorithms such as human posture estimation to analyze the aerial technique of the push-up long jump and establish an automated method for evaluating aerial technique in the long jump based on human posture estimation. Summary of the Invention

[0004] The purpose of the present invention is to provide an automatic evaluation method for the aerial technical movements of the straight-up long jump in response to the need for automatic evaluation of students' aerial technical movements in college teaching scenarios. This method can quantitatively evaluate the long jumper's take-off and aerial technical movement postures, and provide quantitative data on the long jumper's aerial technical movements of the straight-up long jump, which serves as an important reference indicator for performance assessment and reduces the influence of human subjective factors brought about by the evaluation. During normal training, the long jumper can also refer to the evaluation results to promptly correct non-standard problems in various techniques during the long jump process. The long jumper can also refer to these quantitative feedback data to provide targeted guidance to the long jumper, thereby improving the possibility of personalized training.

[0005] The technical solution for achieving the purpose of the present invention is:

[0006] A method for automatically evaluating aerial technical movements of a straight-legged long jump comprises the following steps:

[0007] S1. Determine camera placement and shooting parameters: The camera placement and shooting parameters are determined based on the characteristics of the campus long jump performance assessment venue. The camera is D meters from the edge of the run-up area and is adjusted to an appropriate angle to record the entire long jumper's process from takeoff to landing.

[0008] S2. Human pose estimation and feature angle extraction: Obtaining the coordinates of the human joints in the image provides a basis for subsequent analysis of the long jumper's mid-air movements.

[0009] S2-1. Use the BlazePose human pose estimation algorithm to extract the coordinates of key points of the human body, removing key points that are not relevant to studying the long jumper's posture and movements, such as ears and eyes;

[0010] S2-2. Joint angle calculation method. Assuming the angle of the long jumper's right leg, BlazePose detects and outputs the right ankle joint A, right knee joint B, and right hip joint C. The angle formed by the line connecting B and A and C is θ. Let the coordinates of A, B, and C be (x A ,y A )、(x B ,y B )、(x C ,y C ), then the radian angle between AB and BC is:

[0011]

[0012] The obtained rad is radians. The radians calculated by formula (1) are positive and negative. To convert them into angles, the absolute value must be taken and the limb angle θ∈[0,180°]. The final θ is obtained by formula (2):

[0013]

[0014] Formula (2) gives the angles between joints. The angles between the limbs and the horizontal and vertical directions are assumed to be the right calf of the long jumper. D is the intersection of the straight line passing through point A and parallel to the x-axis and the straight line passing through point B and perpendicular to the x-axis. The coordinates of D are (x B ,y A ), D′ is the intersection of the straight line passing through point A and perpendicular to the x-axis and the x-axis, and the coordinates of D′ are (x A ,0),θ1 is the angle between BA and AD, AD is parallel to the x-axis, θ2 is the angle between BA and AD′, AD′ is perpendicular to the x-axis, θ1 and θ2 can be obtained by formulas (1) and (2);

[0015] S3. Establishing a Standard Library of Aerial Technical Movements: As a complex sport, standardizing the technical movements of the long jump has always been a challenge. Due to the uncertainty of technical movement standards, it is difficult to achieve consistent understanding and evaluation during teaching and practice. To this end, three important technical movements, namely, take-off, straightening and extending the hips, and tucking the abdomen, were selected for evaluation. Based on this, standardized definitions and scoring criteria for technical movements were established. Through comprehensive evaluation and consideration, standardized definitions and scoring criteria for these three movements were obtained. The key point coordinates of the human body extracted in step S2-1 were used to extract corresponding posture features, and a standard movement posture library was established.

[0016] S4. Aerial Technical Posture Matching: A matching algorithm is used to extract keyframes from the long jump video that are most similar to the standard posture. These keyframes include takeoff, hip extension, and landing with the abdomen tucked in. These keyframes are used for subsequent posture evaluation. The specific posture matching process is as follows:

[0017] S4-1. Extract the characteristic angle of posture, extract the characteristic angle of posture, define the characteristic angles of the three actions of take-off, straightening the hips and tightening the abdomen, and the characteristic angle of the take-off posture is θ 11 ,θ 12 ,θ 13 ; The characteristic angle of the posture of straightening and extending the hips is θ 21 ,θ 22 ,θ 23 ; The characteristic angle of the abdomen posture is θ 31 ,θ 32 ,θ 33 ;

[0018] S4-2. Calculate the weighted Euclidean distance between the characteristic angles of the posture to be tested and the standard posture. Extract the posture to be tested from the long jump video frame by frame. Then, perform weighted Euclidean distance calculation on the characteristic angle vectors composed of the characteristic angles corresponding to the three postures of take-off, straightening the body and extending the hips, and tucking the abdomen, respectively, and the characteristic angle vectors corresponding to the three standard postures to evaluate the similarity between the posture to be tested and the three standard postures. Assume that the characteristic angle vector of the posture to be tested is The corresponding characteristic angle vector of the standard posture is The weighted Euclidean distance for:

[0019]

[0020] Where i = 1, 2, 3, 1 represents the take-off posture, 2 represents the posture of straightening the body and extending the hips, 3 represents the posture of contracting the abdomen, p (i) ,q (i) Represents the characteristic angle vector of the corresponding posture; is the weighted Euclidean distance between the characteristic angle vector of the take-off posture to be tested and the characteristic angle vector of the corresponding standard take-off posture, is the weighted Euclidean distance between the characteristic angle vector of the posture to be tested and the characteristic angle vector of the corresponding standard posture, is the weighted Euclidean distance between the characteristic angle vector of the abdomen-tucked posture to be tested and the characteristic angle vector of the corresponding standard abdomen-tucked posture; n is the number of characteristic angles; represents the weight of the j-th feature angle of posture i;

[0021] S4-3. Calculate the cosine distance and set the threshold t to improve the accuracy of posture matching. (i) ,right Calculate the cosine distance between the coordinate vectors of all key points of the posture to be measured and the coordinate vectors of the key points of the corresponding standard posture. Assume that the coordinate vector of the posture to be measured is The coordinate vector of the standard posture is The cosine distance between the coordinate vectors of the posture to be measured and the standard posture is:

[0022]

[0023] S4-4. Calculate the comprehensive distance. Consider the characteristic angle Euclidean distance and the cosine distance of the coordinate point vector between the posture to be measured and the standard posture, and obtain the comprehensive distance d between the posture to be measured and the standard posture, as shown in formula (5):

[0024]

[0025] where μ i , ρ i The coefficient weights for matching the corresponding action postures, where i = 1, 2, 3, 1 represents the take-off posture, 2 represents the straight-back posture, and 3 represents the abdominal posture;

[0026] S4-5. The comprehensive distance result calculated between the posture to be measured and the three key action standard postures is min(d (i) ) as the final matching result;

[0027] S5. Aerial Technical Maneuver Posture Evaluation: Score the three postures matched in step S4;

[0028] S5-1. Determine the scoring function. Use the posture matching algorithm to match the three key action postures of take-off, straightening, and tucking the abdomen for evaluation. Output the score of each posture. Use the similarity between the posture to be tested and the standard posture to score. Use the comprehensive distance d calculated by formula (5) between the posture to be tested and the standard posture for measurement. The smaller d is, the more similar it is, and the higher the score is. Use the linear function to convert the comprehensive distance into a score:

[0029] score=k×f+c (6)

[0030] Where k and c are constants;

[0031] S5-2. Calculate the posture score. To obtain the values ​​of k and c and improve the accuracy and objectivity of the score, select n groups of representative long jump videos. In the absence of mutual interference or influence, score the three key action postures of take-off, straightening, and tucking the abdomen in the video. Each key posture score adopts a 100-point system, and the average score is taken as the final score of the posture. For the scored key postures in the n groups of videos, calculate their comprehensive distance d from the corresponding standard posture. Finally, three types of samples are obtained. Where i = 1, 2, 3, 1 represents the take-off posture, 2 represents the upright posture, and 3 represents the abdominal posture. The video number t = 1, 2, ..., n; It represents the comprehensive distance between the posture i in video number t and its corresponding standard posture. The comprehensive distance d is calculated by formula (5). Score is the average score of the posture. The least square fitting method is used to fit the three types of samples respectively, and the optimal k and c corresponding to the take-off, straightening, and abdomen-tucked postures are obtained respectively. For each matched posture, the comprehensive distance d is obtained and substituted into formula (6) to obtain the corresponding score.

[0032] The key points of this technical solution are:

[0033] (1) The human posture estimation algorithm is used to extract the coordinates of the key points of the long jumper, and the characteristic angle is defined according to the characteristics of the long jump. The weighted Euclidean distance is calculated by the characteristic angle vector of the measured posture and the characteristic angle vector of the standard posture to perform posture matching. In order to improve the matching accuracy, the cosine distance calculation of the posture coordinates is added to comprehensively consider the results of posture matching.

[0034] (2) By scoring several groups of videos, a scoring dataset is constructed to fit the constant of the scoring function to improve the accuracy and objectivity of the scoring.

[0035] This method quantitatively evaluates the long jumper's take-off and aerial technical movements, and provides quantitative data on the long jumper's aerial technical movements in the upright long jump, which serves as an important reference indicator for performance assessment and reduces the influence of human subjective factors brought about by the assessment. During normal training, long jumpers can also refer to the results of the assessment to promptly correct any irregularities in the various techniques during the long jump process. They can also refer to these quantitative feedback data to provide targeted guidance to long jumpers and increase the possibility of personalized training. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 is a flow chart of an embodiment;

[0037] Figure 2This is a diagram showing the placement of cameras and markers in the embodiment;

[0038] Figure 3 A topological diagram of key points of the human body in the embodiment;

[0039] Figure 4 : is a diagram of the right leg joint angle of the long jumper in the embodiment;

[0040] Figure 5 : A diagram showing the angles between the limbs and the horizontal and vertical directions in the embodiment;

[0041] Figure 6 It is the characteristic angle diagram of the key action joints in the embodiment;

[0042] Figure 7 : The key actions extracted in the embodiment and the corresponding score graph. DETAILED DESCRIPTION

[0043] The present invention will be further described below with reference to the accompanying drawings and embodiments, but the present invention is not limited thereto.

[0044] Example:

[0045] Reference Figure 1 A method for automatically evaluating aerial technical movements of a straight-legged long jump comprises the following steps:

[0046] S1. Determine the camera placement and shooting parameters: Determine the camera placement and shooting parameters based on the characteristics of the long jump performance test site on campus. The camera placement is as follows: Figure 2 As shown, the camera is D meters away from the edge of the run-up area, and the camera is adjusted to a suitable angle to record the entire process of the long jumper from take-off to landing;

[0047] S2. Human pose estimation and feature angle extraction: Obtaining the coordinates of the human joints in the image provides a basis for subsequent analysis of the long jumper's mid-air movements.

[0048] S2-1. Use the BlazePose human pose estimation algorithm to extract the key point coordinates of the human body, remove the key point information that is not related to the long jumper's posture and movement, such as ears, eyes, etc., and the extracted key point topology is as follows Figure 3 As shown;

[0049] S2-2. Joint angle calculation method, such as Figure 4 As shown, taking the right leg angle of the long jumper as an example, BlazePose detects the right ankle joint A, right knee joint B and right hip joint C. The angle formed by the line connecting B and A and C is θ. Let the coordinates of A, B and C be (x A ,y A )、(x B ,yB )、(x C ,y C ), then the radian angle between AB and BC is:

[0050]

[0051] The obtained rad is radians. The radians calculated by formula (1) are positive and negative. To convert them into angles, the absolute value must be taken and the limb angle θ∈[0,180°]. The final θ is obtained by formula (2):

[0052]

[0053] Formula (2) gives the angles between joints, the angles between limbs and the horizontal and vertical directions. Taking the right calf of a long jumper as an example, D is the intersection of a straight line passing through point A and parallel to the x-axis and a straight line passing through point B and perpendicular to the x-axis. The coordinates of D are (x B ,y A ), D′ is the intersection of the straight line passing through point A and perpendicular to the x-axis and the x-axis, and the coordinates of D′ are (x A ,0), such as Figure 5 As shown, Figure 5 In (a), θ1 is the angle between BA and AD, and AD is parallel to the x-axis. Figure 5 In (b), θ2 is the angle between BA and AD′. AD′ is perpendicular to the x-axis. θ1 and θ2 can be obtained from formulas (1) and (2).

[0054] S3. Establishing a Standard Library of Aerial Technical Movements: As a complex sport, standardizing the technical movements of the long jump has always been a challenge. Due to the uncertainty of technical movement standards, it is difficult to achieve consistent understanding and evaluation during teaching and practice. To this end, three important technical movements, namely, take-off, straightening and extending the hips, and tucking the abdomen, were selected for evaluation. Based on this, standardized definitions and scoring criteria for technical movements were established. Through comprehensive evaluation and consideration, standardized definitions and scoring criteria for these three movements were obtained. The key point coordinates of the human body extracted in step S2-1 were used to extract corresponding posture features, and a standard movement posture library was established.

[0055] S4. Aerial Technical Posture Matching: A matching algorithm is used to extract keyframes from the long jump video that are most similar to the standard posture. These keyframes include takeoff, hip extension, and landing with the abdomen tucked in. These keyframes are used for subsequent posture evaluation. The specific posture matching process is as follows:

[0056] S4-1. Extract the posture feature angle, such as Figure 6 As shown in the figure, the characteristic angles of the three actions of taking off, stretching the hips and tightening the abdomen are defined as follows: Figure 6 θ in (a) 11,θ 12 ,θ 13 、 Figure 6 θ in (b) 21 ,θ 22 ,θ 33 and Figure 6 θ in (c) 31 ,θ 32 ,θ 33 ;

[0057] S4-2. Calculate the weighted Euclidean distance between the characteristic angles of the posture to be tested and the standard posture, extract the posture to be tested in the image frame by frame from the long jump video, and assign the posture to be tested to the jump position. Figure 6 (a) Lift your body and extend your hips Figure 6 (b) Tighten your abdomen Figure 6 (c) The characteristic angle vector composed of the characteristic angles corresponding to the three postures is weighted Euclidean distance calculated with the characteristic angle vectors corresponding to the three standard postures to evaluate the similarity between the posture to be tested and the three standard postures. Assume that the characteristic angle vector of the posture to be tested is The corresponding characteristic angle vector of the standard posture is The weighted Euclidean distance for:

[0058]

[0059] Where i = 1, 2, 3, 1 represents the take-off posture, 2 represents the posture of straightening the body and extending the hips, 3 represents the posture of contracting the abdomen, p (i) ,q (i) Represents the characteristic angle vector of the corresponding posture; is the weighted Euclidean distance between the characteristic angle vector of the take-off posture to be tested and the characteristic angle vector of the corresponding standard take-off posture, is the weighted Euclidean distance between the characteristic angle vector of the posture to be tested and the characteristic angle vector of the corresponding standard posture, is the weighted Euclidean distance between the characteristic angle vector of the abdomen-tucked posture to be tested and the characteristic angle vector of the corresponding standard abdomen-tucked posture; n is the number of characteristic angles; represents the weight of the j-th feature angle of posture i;

[0060] S4-3. Calculate the cosine distance and set the threshold t to improve the accuracy of posture matching. (i) ,right Calculate the cosine distance between the coordinate vectors of all key points of the posture to be measured and the coordinate vectors of the key points of the corresponding standard posture. Assume that the coordinate vector of the posture to be measured is The coordinate vector of the standard posture is The cosine distance between the coordinate vectors of the posture to be measured and the standard posture is:

[0061]

[0062] S4-4. Calculate the comprehensive distance. Consider the characteristic angle Euclidean distance and the cosine distance of the coordinate point vector between the posture to be measured and the standard posture, and obtain the comprehensive distance d between the posture to be measured and the standard posture, as shown in formula (5):

[0063]

[0064] where μ i , ρ i The coefficient weights for matching the corresponding action postures, where i = 1, 2, 3, 1 represents the take-off posture, 2 represents the straight-back posture, and 3 represents the abdominal posture;

[0065] S4-5. The comprehensive distance result calculated between the posture to be measured and the three key action standard postures is min(d (i) ) as the final matching result;

[0066] S5. Aerial Technical Maneuver Posture Evaluation: Score the three postures matched in step S4;

[0067] S5-1. Determine the scoring function. Use the posture matching algorithm to match the three key action postures of take-off, straightening, and tucking the abdomen for evaluation. Output the score of each posture. Use the similarity between the posture to be tested and the standard posture to score. Use the comprehensive distance d calculated by formula (5) between the posture to be tested and the standard posture for measurement. The smaller d is, the more similar it is, and the higher the score is. Use the linear function to convert the comprehensive distance into a score:

[0068] score=k×d+c (6)

[0069] Where k and c are constants;

[0070] S5-2. Calculate the posture score. To obtain the values ​​of k and c and improve the accuracy and objectivity of the score, select n groups of representative long jump videos. In the absence of mutual interference or influence, score the three key action postures of take-off, straightening, and tucking the abdomen in the video. Each key posture score adopts a 100-point system, and the average score is taken as the final score of the posture. For the scored key postures in the n groups of videos, calculate their comprehensive distance d from the corresponding standard posture. Finally, three types of samples are obtained. Where i = 1, 2, 3, 1 represents the take-off posture, 2 represents the upright posture, and 3 represents the abdominal posture. The video number t = 1, 2, ..., n; It represents the comprehensive distance between the posture i in video number t and its corresponding standard posture. The comprehensive distance d is calculated by formula (5). Score is the average score of the posture. The least square fitting method is used to fit the three types of samples respectively, and the optimal k and c corresponding to the take-off, hip extension, and abdomen contraction postures are obtained respectively. For each matched posture, the comprehensive distance d is obtained and substituted into formula (6) to get the corresponding score.

[0071] Simulation experiment: In a college physical education teaching scenario, in this experiment, the camera is placed according to step S1 and a class of students' straight long jump videos are recorded. Through 30 representative long jump videos selected by full-time physical education teachers, the key action images of take-off, straightening the body and extending the hips, and contracting the abdomen in each long jump video are intercepted. Three physical education teachers are asked to score the intercepted key actions, and the best fitting results for each posture action are obtained according to step S5-2. The new long jump video sample is tested, and the extracted key postures and corresponding scores are shown as follows: Figure 7 shown.

Claims

1. A method for automatically evaluating aerial technical movements in a straight-above-the-horse long jump, characterized in that: The steps include: S1. Determine camera placement and shooting parameters: Determine the camera placement and shooting parameters based on the characteristics of the campus long jump performance assessment venue. The camera should be D meters from the edge of the run-up area and adjusted to an appropriate angle to capture the entire jumper's process from takeoff to landing. S2. Human pose estimation and feature angle extraction: Obtain the coordinates of the human body's joints in the image: S2-1. Use the BlazePose human pose estimation algorithm to extract the coordinates of key points of the human body, removing key points that are not relevant to studying the long jumper's posture and movements. S2-2. Joint angle calculation method, right ankle joint output by BlazePose detection , right knee joint The angle formed by the line connecting the right hip joint C, B and A, C is ,set up 、 The coordinates of C are 、 、 , then the radian angle between AB and BC is: (1)； get The radian calculated by formula (1) can be positive or negative. To convert it to an angle, the absolute value must be taken and the limb angle , the final From formula (2), we can get: (2)； Formula (2) gives the angles between joints. The angles between the limbs and the horizontal and vertical directions are assumed to be the right calf of the long jumper. D is the angle passing through point A and parallel to The straight line passing through point B and perpendicular to the axis The coordinates of the intersection of the straight lines of the axis are , is a point passing through point A and perpendicular to The straight line of the axis The intersection of the axes, The coordinates are , is the angle between BA and AD, AD is parallel to axis, BA and The angle of Perpendicular to Axis, can be obtained by formula (1) and (2) 、 ; S3 establishes a standard library of aerial technical actions: using step S2-1 to extract the key point coordinates of the human body to extract the corresponding posture features and establish a standard action posture library; S4. Aerial Technical Posture Matching: A matching algorithm is used to extract keyframes from the long jump video that are most similar to the standard posture. These keyframes include the takeoff, hip extension, and landing with the abdomen tucked in. The specific posture matching process is as follows: S4-1. Extract the characteristic angles of posture and define the characteristic angles of the three actions of take-off, straightening the body and extending the hips, and contracting the abdomen. The characteristic angle of the take-off posture is 、 、 ; The characteristic angle of the posture of straightening the body and extending the hips is 、 、 ; The characteristic angle of the abdomen posture is 、 、 ; S4-2. Calculate the weighted Euclidean distance between the characteristic angles of the posture to be tested and the standard posture. Extract the posture to be tested from the long jump video frame by frame. Then, perform weighted Euclidean distance calculation on the characteristic angle vectors corresponding to the three postures of take-off, straightening and extending the hips, and tucking the abdomen. This is used to evaluate the similarity between the posture to be tested and the three standard postures. Assume that the characteristic angle vector of the posture to be tested is , the corresponding characteristic angle vector of the standard posture is , then the weighted Euclidean distance for: (3)； in , 1 represents the jumping posture, 2 represents the hip-stretching posture, 3 represents the abdominal contraction posture, 、 Represents the characteristic angle vector of the corresponding posture; is the weighted Euclidean distance between the characteristic angle vector of the take-off posture to be tested and the characteristic angle vector of the corresponding standard take-off posture, is the weighted Euclidean distance between the characteristic angle vector of the posture to be tested and the characteristic angle vector of the corresponding standard posture, is the weighted Euclidean distance between the characteristic angle vector of the abdomen-tucked posture to be tested and the characteristic angle vector of the corresponding standard abdomen-tucked posture; is the number of characteristic angles; Expressing posture No. The weight of the characteristic angle; S4-3. Calculate the cosine distance and set the threshold to improve the accuracy of posture matching ,right Calculate the cosine distance between the coordinate vectors of all key points of the posture to be measured and the coordinate vectors of the key points of the corresponding standard posture. Assume that the coordinate vector of the posture to be measured is , the coordinate vector of the standard posture is , the cosine distance between the coordinate vector of the posture to be measured and the standard posture is: (4)； S4-4. Calculate the comprehensive distance. Consider the characteristic angle Euclidean distance and the cosine distance of the coordinate point vector between the posture to be measured and the standard posture to obtain the comprehensive distance between the posture to be measured and the standard posture. , as shown in formula (5): (5)； in is the coefficient weight for matching the corresponding action posture, where , 1 represents the jumping posture, 2 represents the hip-extending posture, and 3 represents the abdominal contraction posture; S4-5. Calculate the comprehensive distance between the posture to be measured and the three key action standard postures, and take As a result of the final match; S5. Aerial Technical Maneuver Posture Evaluation: Score the three postures matched in step S4; S5-1. Determine the scoring function and evaluate the three key postures of take-off, upright, and abdominal contraction obtained by the posture matching algorithm. Output the score of each posture. Use the similarity between the posture to be tested and the standard posture to score. Use formula (5) to calculate the comprehensive distance between the posture to be tested and the standard posture To measure, The smaller the value, the more similar it is and the higher the score is. A linear function is used to convert the comprehensive distance into a score. : (6)； in, and is a constant; S5-2. Calculate posture scores. To obtain the values ​​of k and c and improve the accuracy and objectivity of the scoring, select n representative groups of straight-legged long jump videos. Score the three key postures in the videos—takeoff, straight-legged, and abdominal contraction—without interfering with each other. Each key posture is scored on a 100-point scale, and the average score is taken as the final score for that posture. For each of the n groups of videos, calculate the overall distance between the scored key postures and the corresponding standard posture. , and finally we get 3 types of samples ,in , 1 represents the jumping posture, 2 represents the upright posture, 3 represents the abdominal posture, video number ; Indicates the video number The middle posture is The comprehensive distance of the corresponding standard posture, the comprehensive distance Calculated by formula (5), is the average score of the posture. The least square fitting method is used to fit the three types of samples, and the optimal scores corresponding to the take-off, upright and abdominal postures are obtained respectively. and For each matched posture, the comprehensive distance d is obtained and substituted into formula (6) to obtain the corresponding score.

Images