A debugging method for a martial arts intelligent scoring system

Abstract

The application discloses a debugging method for a martial art intelligent scoring system, which adopts skeleton recognition technology to implement action capture and tracking on a practitioner, and comprises the following steps: key action decomposition is carried out according to martial art routine scoring rules, and scoring standards of each key action are set, including action parameter specifications, deduction modes and weights and continuity; video clips corresponding to key actions with large differences between system scoring and manual scoring are found out, whether key frame recognition is correct and whether manual scoring is misjudged is judged; whether action parameter specifications, deduction modes and weight distribution are reasonable is analyzed; whether interaction exists between every two adjacent key actions is judged, if yes, an interaction starting condition is additionally set at the previous key action, and an influence factor is additionally set at the latter key action; continuity between each key action in the whole martial art routine is analyzed, and lengths of time intervals between every two adjacent key actions are adjusted, so that the continuity scoring is more reasonable.

Description

Technical Field

[0001] This invention belongs to the field of computer technology, and specifically relates to a debugging method for a martial arts intelligent scoring system. Background Technology

[0002] With the progress and development of education, student education not only advocates ideological and moral education but also emphasizes students' physical health and literacy. To this end, many schools have included martial arts in their examinations. However, relying solely on manual scoring of martial arts movements performed by a large number of students is not objective enough, and is slow, inefficient, and lacks objectivity. As a result, many intelligent scoring systems have emerged. Before these intelligent scoring systems are officially put into use, a lot of debugging work is often required. Otherwise, machine scoring often differs greatly from human scoring, seriously affecting the promotion and application of intelligent scoring systems. However, the current debugging work is disorganized, time-consuming, inefficient, and prone to omissions and bugs. Problems can only be fixed when they occur in actual use, resulting in a poor user experience. Summary of the Invention

[0003] This invention provides a debugging method for a martial arts intelligent scoring system, which solves the technical problems of existing intelligent scoring systems, such as long debugging time, low efficiency, many missed bugs, and poor user experience.

[0004] This invention can be achieved through the following technical solutions:

[0005] A debugging method for a martial arts intelligent scoring system, wherein the martial arts intelligent scoring system uses skeleton recognition technology to capture and track the movements of practitioners, includes the following steps:

[0006] Step 1: Decompose key movements according to the scoring rules of martial arts routines, and set scoring criteria for each key movement, including movement parameter specifications, deduction methods and weights, and continuity.

[0007] Step 2: Identify the video segments corresponding to key actions where the system score and human score differ significantly, and determine whether the keyframe recognition is correct and whether the human score is incorrect.

[0008] Step 3: Analyze whether the action parameter specifications, deduction methods, and weight allocation are reasonable;

[0009] Step 4: Determine whether there is interaction between any two adjacent key actions. If so, add an interaction initiation condition to the previous key action and add an influencing factor to the next key action.

[0010] Step 5: Analyze the continuity between key movements in the entire martial arts routine, and adjust the time interval between each pair of adjacent key movements to make the continuity score more reasonable.

[0011] Furthermore, multiple scoring points are set for each key action, and action parameter specifications, deduction methods, and weights are set for each scoring point;

[0012] In step three, the selection of joints in the keyframe corresponding to the action scoring point is analyzed to determine whether it is correct. If it is correct, the corresponding action parameter specifications are judged to be reasonable in combination with the manual scoring rules. If they are not reasonable, they are readjusted. If they are reasonable, the deduction method is adjusted.

[0013] The initial value of the deduction method is set to linear floating deduction. Based on the manual scoring rules and the actual situation of the scoring action points, the deduction method is adjusted to exponential floating deduction or binary deduction. Otherwise, a weight allocation adjustment is performed. The linear floating deduction is set so that the deduction value and the action completion degree have an inverse trend. The exponential floating deduction is set so that the deduction value and the action completion degree have an exponential decreasing trend. The binary deduction is that no points are deducted if the action is completed, otherwise all points are deducted.

[0014] Based on the actual situation of key actions, a comprehensive analysis of the weight distribution of each action's scoring points is conducted, and the corresponding weight distribution is adjusted accordingly.

[0015] Furthermore, determine whether the preceding key action is a basic action; if so, then there is an interaction between it and the following key action.

[0016] The interaction is initiated when the overall deduction of the previous key action exceeds the first threshold or the deduction of a certain action score point in the previous key action exceeds the second threshold. If the interaction is initiated, the score of the next key action is multiplied by the influence factor to obtain the final score.

[0017] Furthermore, in step two, the video is first reviewed again to check if the manual scoring is correct. If it is correct, the keyframe recognition is checked to see if it is correct. If the recognition is incorrect, the keyframe recognition is performed again. If the recognition is correct, step three is executed.

[0018] Furthermore, the scoring criteria are set to professional and amateur versions. It is determined whether the version used for scoring is correct. If so, step two is executed.

[0019] The beneficial technical effects of this invention are as follows:

[0020] 1. For martial arts routines, this invention has established a complete scoring standard, extending from each key movement to the entire martial arts routine. Then, based on the difference between system scoring and manual scoring, it first checks the specifications of movement parameters, deduction methods and weights of movement scoring points, and then checks the interactivity between key movements and the coherence of the entire martial arts routine. This can identify problems as much as possible, reduce bugs, and ensure the stable operation of the subsequent scoring system.

[0021] 2. Clarify the debugging process and methods of the martial arts intelligent scoring system to make the debugging process clearer and more transparent. Even engineers who are not familiar with martial arts training can make corresponding adjustments according to the debugging methods of this invention, so that the martial arts intelligent scoring system can score martial arts routines more quickly and accurately, and reduce the influence of human subjective factors on the scoring. Attached Figure Description

[0022] Figure 1 This is the overall flowchart of the present invention;

[0023] Figure 2(a) and 2(b) These are schematic diagrams illustrating the scoring criteria for the two key movements of this invention: the front kick and the flying kick.

[0024] Figure 3 This is an example of the difference between the system scoring and the manual scoring of this invention;

[0025] Figure 4 This is a schematic diagram of the scoring criteria for the airborne flying kick after debugging according to the present invention. Detailed Implementation

[0026] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings and preferred embodiments.

[0027] With the development of artificial intelligence technology, human posture recognition algorithms have become increasingly sophisticated, such as the RMPE model, Mask R-CNN model, and skeleton recognition algorithms. Among these, OpenPose-based posture recognition technology is the most widely used. Martial arts intelligent scoring systems can use skeleton recognition technology to capture and track the movements of practitioners, and then analyze the capture and tracking results to provide corresponding scores. However, from the completion of the scoring system's construction to its formal deployment, a significant amount of debugging work is required. Currently, there are no practical and feasible debugging methods to facilitate the rapid promotion and application of the scoring system. Figure 1 As shown, this invention provides a debugging method for a martial arts intelligent scoring system. First, a comprehensive scoring standard is designed, considering not only the specific scoring of each key movement but also the coherence assessment of the entire martial arts routine, making it more closely aligned with the actual situation of human scoring. Then, a debugging strategy from details to the whole is adopted, systematically checking for potential scoring discrepancies to reduce omissions and resolve problems as early as possible during the debugging phase. This reduces the probability of bugs during actual use, improves user experience, and provides a standardized debugging method for the martial arts intelligent scoring system. This method effectively guides developers in performing debugging work, significantly shortens debugging time, improves debugging efficiency, and accelerates the widespread application of the martial arts intelligent scoring system.

[0028] Specifically as follows:

[0029] Step 1: Decompose the key movements according to the martial arts routine, and set the scoring criteria for each key movement, including movement parameter specifications, deduction methods and weights, and continuity.

[0030] Through consultation and discussion between R&D personnel and martial arts experts, key movements were divided according to the existing scoring standards of martial arts routines. Then, multiple scoring points for each key movement were identified, and movement parameter specifications, deduction methods, and weights were set. At the same time, the integrity of the martial arts routine, including the continuity between key movements, was considered, and the initial setting of the scoring standards for each key movement was completed.

[0031] As shown in Figure 2, the scoring criteria for two key movements, the front kick and the flying kick, are given. Based on the discussion, weights are assigned to each scoring point of the movement. According to the position of the scoring movement, different levels of completion assessment and their deduction methods are set, corresponding to the upper and lower limit columns of parameters and the upper and lower limit columns of interval deduction. For example, if the air time of the flying kick is greater than 1 second, no points are deducted; if it is less than 0.5 seconds, all points are deducted (1 point is deducted); if it is between 0.5 seconds and 1 second, 0.5 points are deducted. This is a linear deduction method.

[0032] In addition, considering that the martial arts scoring system has many application scenarios, such as experiential courses for primary school students or high-precision martial arts single-subject examinations, we can design scoring standards for different assessment levels, such as professional and amateur versions. The amateur version can reduce some weights to increase scores, thereby increasing participants' interest, while the professional version needs to refine the weight rules to achieve precise scoring and improve the fairness of the examination. Therefore, during debugging, we can first determine whether the version used for scoring is correct based on the actual situation. If so, proceed to step two.

[0033] Step 2: Identify the video segments corresponding to key actions where the system score and human score differ significantly, and determine whether the keyframe recognition is correct and whether the human score is incorrect.

[0034] Filter and extract original videos with discrepancies, such as those with high action completion but low system scores, or low action completion but high system scores. Figure 3 As shown, first review the video to check if the manual scoring is correct. If it is correct, check if the keyframe recognition is correct. If the recognition is incorrect, re-recognize the keyframes. For example, if the keyframe corresponding to the take-off point of the flying kick is selected as the landing point or the impact point, it needs to be manually modified. If the recognition is correct, proceed to step three.

[0035] The original video can also be re-skeletalized to check for errors in the skeleton recognition object that may cause scoring errors. If so, the video should be cleaned up, such as removing background and interference, before re-skeletalizing. If not, continue with the subsequent steps.

[0036] Step 3: Analyze whether the action parameter specifications, deduction methods, and weight allocation are reasonable.

[0037] 1. Analyze whether the joint selection in the keyframe corresponding to the action scoring point is correct. If correct, combine the manual scoring rules to judge whether the corresponding action parameter specifications are reasonable. If unreasonable, readjust them. If reasonable, implement the deduction method adjustment.

[0038] The initial value of this deduction method is set to linear floating deduction. Based on the manual scoring rules and the actual situation of the scoring action points, the deduction method is adjusted to exponential floating deduction or binary deduction. Otherwise, the weight allocation is adjusted. Specifically, linear floating deduction is set so that the deduction value and the action completion rate are inversely proportional; exponential floating deduction is set so that the deduction value and the action completion rate show an exponential decreasing trend; binary deduction is no deduction if the action is completed, otherwise all points are deducted.

[0039] Specifically, first analyze whether the selection of joints in the keyframes corresponding to the motion scoring points is correct, for example... Figure 4 In parameter 1 of the flying kick, the judges do not deduct points based on the time it takes for the swing leg to land first. Instead, they judge based on which leg lands first. Therefore, the scoring action of parameter 1 needs to be redefined. Parameter 1 should be changed to land foot as take-off foot. The corresponding action parameter specifications and deduction methods, i.e., the upper and lower limits of the deduction value, also need to be adjusted.

[0040] Then, a non-linear adjustment of the deduction method is made: For some technical movements with high completion requirements, the weight adjustment does not have to be limited to linear changes. A non-linear function can be used to adjust the weight to more precisely simulate the judge's scoring behavior. For example, the weight can be adjusted exponentially or logarithmically according to the difficulty of the movement or the quality of execution. For example, the hang time of each person's flying kick will be different, but using a fixed scoring range cannot accurately reflect this. Using the exponential or logarithmic conversion of the hang time means that the closer it is to the scoring standard of 1 second, the fewer points will be deducted, rather than simply stipulating that the hang time is 0.5-1 second and deducting the same number of points. This makes the score of hang time more significant and distinct.

[0041] 2. When adjusting the weights, based on the actual situation of the key movements, the weight distribution of each movement's scoring point is comprehensively analyzed, and the corresponding weight distribution is adjusted accordingly. For example, parameters 3 and 4 in the flying kick are relatively important scoring points in actual judging, but they will not be deducted more than 1 point because the kicking leg is not straight or does not go over the shoulder. Therefore, based on the comparison between the system score and the judge's score, if the weight deduction of a certain weight in the system score exceeds the judge's usual deduction range (e.g., more than 1 point), the weight score will be modified accordingly.

[0042] Often, the adjustment of motion parameter specifications, deduction methods, and weight allocation are interrelated and can be flexibly combined and executed according to the actual situation.

[0043] Step 4: Determine whether there is interaction between any two adjacent key actions. If so, add an interaction initiation condition to the previous key action and add an influencing factor to the next key action.

[0044] In actual scoring, there may be interaction effects between different technical movements, that is, the execution of one movement may affect the evaluation of another movement. For example, in a front kick, even if the toes touch the forehead after the swing leg bends, the quality of the movement is relatively low. If such a correlation exists, interaction terms with weights are introduced to simulate the actual complexity of this scoring.

[0045] Typically, one of the two key actions that generate an interaction is a basic action, such as a front kick. Therefore, we can first determine whether the previous key action is a basic action. If so, there is an interaction with the next key action. Then, we use the condition that the overall deduction of the previous key action is greater than the first threshold or the deduction of a certain action score point in the previous key action is greater than the second threshold as the interaction initiation condition. If the interaction initiation condition is met, the score of the next key action is multiplied by the influence factor to obtain the final score.

[0046] Step 5: Analyze the continuity between key movements in the entire martial arts routine, and adjust the time interval between each pair of adjacent key movements to make the continuity score more reasonable.

[0047] In the evaluation of boxing, there are many movements in the transition between the bow and horse stances, and this continuous movement is also a very important part of the evaluation of martial arts routines. Therefore, it is very necessary to score the continuity of a martial arts routine. Time series analysis can be used to obtain the completion time and interval time of each key movement. By setting different interval times and their corresponding weights, the difference between the system score and the human score can be tested until the difference between the two is within the specification, thus completing the debugging of the continuity score.

[0048] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A debugging method for a martial arts intelligent scoring system, wherein the martial arts intelligent scoring system uses skeleton recognition technology to capture and track the movements of a practitioner, characterized in that... Includes the following steps: Step 1: Decompose key movements according to the scoring rules of martial arts routines, and set scoring criteria for each key movement, including movement parameter specifications, deduction methods and weights, and continuity. Step 2: Identify the video segments corresponding to key actions where the system score and human score differ significantly, and determine whether the keyframe recognition is correct and whether the human score is incorrect. In step two, the video is first reviewed to check if the manual scoring is correct. If it is correct, the keyframe recognition is checked to see if it is correct. If it is incorrect, the keyframe recognition is performed again. If the identification is correct, proceed to step three; Step 3: Analyze whether the action parameter specifications, deduction methods, and weight allocation are reasonable; Multiple scoring points are set for each key action, and action parameter specifications, deduction methods, and weights are set for each scoring point; In step three, the selection of joints in the keyframe corresponding to the action scoring point is analyzed to determine whether it is correct. If it is correct, the corresponding action parameter specifications are judged to be reasonable in combination with the manual scoring rules. If they are not reasonable, they are readjusted. If they are reasonable, the deduction method is adjusted. The initial value of the deduction method is set to linear floating deduction. Based on the manual scoring rules and the actual situation of the scoring action points, the deduction method is adjusted to exponential floating deduction or binary deduction. Otherwise, a weight allocation adjustment is performed. The linear floating deduction is set so that the deduction value and the action completion degree have an inverse trend. The exponential floating deduction is set so that the deduction value and the action completion degree have an exponential decreasing trend. The binary deduction is that no points are deducted if the action is completed, otherwise all points are deducted. Based on the actual situation of key actions, comprehensively analyze the weight distribution of each action's scoring points and adjust the corresponding weight distribution accordingly; Determine if the preceding key action is a basic action; if so, it interacts with the following key action. If the overall deduction of the previous key action is greater than the first threshold or the deduction of a certain action score point in the previous key action is greater than the second threshold, the interaction activation condition is used. If the interaction activation condition is met, the score of the next key action is multiplied by the influence factor to obtain the final score. Step 4: Determine whether there is interaction between any two adjacent key actions. If so, add an interaction initiation condition to the previous key action and add an influencing factor to the next key action. Step 5: Analyze the continuity between key movements in the entire martial arts routine, and adjust the time interval between each pair of adjacent key movements to make the continuity score more reasonable. Time series analysis is used to obtain the completion time and interval of each key action. By setting different intervals and their corresponding weights, the difference between system scoring and human scoring is tested until the difference is within the specification, thus completing the debugging of the consistency scoring.

2. The debugging method for a martial arts intelligent scoring system according to claim 1, characterized in that: The scoring criteria are set to professional and amateur versions. The correct version is determined, and if so, step two is executed.

Images