A Physical Fitness Event Action Quality Assessment Method Based on Label Distribution Learning

Abstract

The present invention relates to the technical field of physical movement quality assessment, and specifically, to a method for assessing the quality of physical event movements based on marker distribution learning. It at least includes obtaining physical event videos and evaluation images in the original database; clipping the competition videos to obtain multiple main key frame images of limb movements; comparing and evaluating the multiple main key frame images with the evaluation images inside the original database to obtain the evaluation information on the standard degree of the main key frame image movements; obtaining other competition videos in the original database and generating the evaluation information on the standard degree of the secondary key frame image movements; comparing the evaluation information on the standard degree of the main key frame image movements with the evaluation information on the standard degree of the secondary key frame image movements; predicting the standard degree of the personnel's physical movements. By making an auxiliary predictive assessment of the standard degree of the physical movements of the subsequent personnel during a physical event with the same movements as the main key frame images, the present system can have a preliminary understanding of the current physical event.

Description

Technical Field

[0001] The present invention relates to the technical field of physical movement quality assessment, and more specifically, to a method for assessing the quality of physical event movements based on labeled distribution learning. Background Art

[0002] With the progress and development of technology, more and more physical events appear in life, such as a series of physical events like gymnastics, swimming, and dancing.

[0003] After a person participates in a physical event, the referee will score the person's physical movements according to the standard degree of the person's limb performance to judge the person's physical standard score. However, with the increasing number of current physical events, the method of manual scoring and assessment greatly affects the assessment efficiency of the overall physical event;

[0004] Secondly, using the above scoring method can only score the current physical event of the person, and cannot pre-analyze and score the subsequent physical event performance of the person based on the performance of the physical events the person has participated in, which affects the pre-assessment of the person's performance in the physical event, and thus cannot objectively predict and evaluate the current situation of the physical event. Summary of the Invention

[0005] The purpose of the present invention is to provide a method for assessing the quality of physical event movements based on labeled distribution learning to solve the problems raised in the above background art.

[0006] To achieve the above purpose, the present invention aims to provide a method for assessing the quality of physical event movements based on labeled distribution learning, at least including:

[0007] S1. Obtain the physical event video and evaluation reference image in the original database;

[0008] S2. Determine the evaluation reference; mark a torso reference point O(X0, Y0, Z0) and multiple limb reference points B j (X j , Y j , Z j ) in the evaluation reference image, and establish the limb reference vector

[0009] S3. Create a reference database and store the limb reference vector in the reference database;

[0010] S4. Establish the movement quality vector; including:

[0011] S41. Clip the physical event video to obtain multiple main key-frame images, and mark a torso comparison point and limb comparison points with the same number and limb positions as the limb reference points in each main key-frame;

[0012] S42. Obtain the trunk comparison point coordinates o(x i0k , y i0k , z i0k ) and the set of limb comparison point coordinates {b j (x ijk , y ijk , z ijk )}, where i is the subscript of the main key frame, j is the subscript of the comparison point, and k is the subscript of the camera device; {(x ijk , y ijk , z ijk )} represents the coordinates in the x, y, and z directions of the jth limb key point in the ith main key frame obtained by the kth camera device;

[0013] S43. Average the trunk comparison point coordinates (x i0k , y i0k , z i0k ) and the set of limb comparison point coordinates {b j (x ijk , y ijk , z ijk )} to obtain the trunk comparison point coordinates and the set of average limb comparison point coordinates satisfying:

[0014]

[0015]

[0016] where n is the total number of camera devices;

[0017] S44. Establish a limb comparison vector; the limb comparison vector

[0018] S5. Vector comparison; for any limb comparison point b j , obtain the spatial angle α j between the limb reference vector and the limb comparison vector, and arrange them in descending order to obtain an action deviation sequence {α j};

[0019] S6. Evaluation; the evaluator scores the participating personnel based on the action deviation sequence {α j}.

[0020] Further, when initially obtaining the physical fitness event video, obtain physical fitness event videos from multiple perspectives in the original database.

[0021] Further, when clipping the physical fitness event video, set the clipping interval according to the evaluation reference image.

[0022] Further, in step S6, when the evaluator scores the participants based on the action deviation sequence {α j}, the interval method is used for scoring. According to the size of the spatial angle α j , it is divided into five-level standards, and specific scores G are given:

[0023]

[0024] Further, after the score information is generated, the standard degree and score of the physical action of other competition videos with the same actions as the main key-frame image are predicted based on the score information.

[0025] Further, the main key-frame images with a first-level standard degree are regarded as sample images for storage to obtain a sample image library.

[0026] Further, when the sample images in the sample image library gradually increase, the redundant sample images are deleted in ascending order according to the number of limb comparison points with a first-level standard degree in the sample image, and new sample images are stored.

[0027] Further, the deletion of redundant sample images in ascending order is completed by the bubble method.

[0028] Further, when a person uses the sample image library for training, the sample images are recommended in descending order according to the number of limb comparison points with a first-level standard degree in the sample image.

[0029] Further, the recommendation of sample images in descending order is completed by the bubble method.

[0030] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0031] 1. The present invention compares the limb standard degrees of competition participants by means of vectors, calculates the deviation angles of the limbs, can intuitively display the deficiencies of the competition participants, and can achieve rapid scoring or guidance.

[0032] 2. In the present invention, high-standard physical training videos are recommended to target personnel with different physical action standard degrees based on the score information, so as to facilitate the personnel to carry out corresponding physical competition training and promote the physical action standard degrees of competition personnel. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 is the overall flowchart of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0034] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0035] Embodiment 1

[0036] Please refer to Figure 1 As shown, this embodiment provides a method for evaluating the quality of physical fitness event actions based on label distribution learning, including at least:

[0037] For the above purpose, the present invention aims to provide a method for evaluating the quality of physical fitness event actions based on label distribution learning, including at least:

[0038] S1. Obtain the physical fitness event videos and evaluation benchmark images in the original database;

[0039] S2. Determine the evaluation benchmark; mark a torso reference point O(X0, Y0, Z0) and multiple limb reference points B j (X j , Y j , Z j ) in the evaluation benchmark image, and establish limb reference vectors

[0040] S3. Create a benchmark database and store the limb reference vectors in the benchmark database;

[0041] S4. Establish action quality vectors; including:

[0042] S41. Clip the physical fitness event video to obtain multiple main key frame images, and mark a torso comparison point and limb comparison points with the same number and limb positions as the limb reference points in each main key frame;

[0043] S42. Obtain the torso comparison point coordinates o(x i0k , y i0k , z i0k ) and the set of limb comparison point coordinates {b j (x ijk , y ijk , z ijk )}, where i is the subscript of the main key frame, j is the subscript of the comparison point, and k is the subscript of the camera device; {(x ijk , y ijk , z ijk )} represents the x, y, and z coordinates of the jth limb key point in the ith main key frame obtained by the kth camera device;

[0044] S43. Equalize the coordinates of the torso comparison points (x i0k , y i0k , z i0k ) and the set of coordinates of the limb comparison points {b j (x ijk , y ijk , z ijk )} to obtain the coordinates of the torso comparison points and the set of average coordinates of the limb comparison points Satisfying:

[0045]

[0046]

[0047]

[0048]

[0049]

[0050] where n is the total number of camera devices;

[0051] S44. Establish a limb comparison vector; the limb comparison vector

[0052] S5. Vector comparison; for any limb comparison point b j , obtain the spatial angle α j between the limb reference vector and the limb comparison vector, and arrange them in descending order to obtain an action deviation sequence {α j};

[0053] S6. Evaluation; the evaluator scores the participating personnel based on the action deviation sequence {α j}.

[0054] Furthermore, when initially obtaining the physical fitness competition video, obtain physical fitness competition videos from multiple perspectives in the original database.

[0055] Furthermore, when clipping the physical fitness competition video, set the clipping interval according to the evaluation reference image.

[0056] Furthermore, in step S6, when the evaluator scores the participating personnel based on the action deviation sequence {α j}, the interval method is used for scoring. According to the size of the spatial angle α j , it is divided into five-level standards, and specific scores G are given:

[0057]

[0058] Further, after the score information is generated, the standard degree and score of the physical actions of other competition videos with the same actions as the main key-frame image are predicted based on the score information.

[0059] Further, the main key-frame images with a first-level standard degree are regarded as sample images for storage to obtain a sample image library.

[0060] Further, when the sample images in the sample image library gradually increase, the redundant sample images are deleted in ascending order according to the number of limb comparison points with a first-level standard degree in the sample images, and new sample images are stored.

[0061] Further, the deletion of redundant sample images in ascending order is completed by the bubble method.

[0062] Further, when a person uses the sample image library for training, the sample images are recommended in descending order according to the number of limb comparison points with a first-level standard degree in the sample images.

[0063] Further, the recommendation of sample images in descending order is completed by the bubble method.

[0064] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited by the above embodiments. The above embodiments and the descriptions in the specification are only preferred examples of the present invention and are not used to limit the present invention. Without departing from the spirit and scope of the present invention, the present invention will have various changes and improvements, and these changes and improvements fall within the scope of the present invention claimed. The scope of protection claimed by the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for evaluating the quality of physical fitness event actions based on label distribution learning, characterized in that: At least include: S1. Obtain the physical fitness event videos and evaluation reference images in the original database; When initially obtaining the physical fitness event videos, obtain physical fitness event videos from multiple perspectives in the original database; S2. Determine the evaluation benchmark; mark a torso reference point O(X0, Y0, Z0) and multiple limb reference points B in the evaluation benchmark image j (X j , Y j , Z j ), and establish limb reference vectors S3. Create a reference database and store the limb reference vectors in the reference database; S4. Establish an action quality vector; Include: S41. Clip the physical fitness event videos to obtain multiple main key frame images, and mark a torso comparison point and limb comparison points with the same number and limb positions as the limb reference points in each main key frame; when clipping the physical fitness event videos, set the clipping interval according to the evaluation reference images; S42. Obtain the trunk comparison point coordinates o(x i0k , y i0k , z i0k ) and the set of limb comparison point coordinates {b j (x ijk , y ijk , z ijk )}, where i is the subscript of the main key frame, j is the subscript of the comparison point, and k is the subscript of the camera device; {(x ijk , y ijk , z ijk )} represents the coordinates in the x, y, and z directions of the j-th limb key point in the i-th main key frame obtained by the k-th camera device; S43. Average the coordinates (x i0k , y i0k , z i0k ) of the torso comparison points and the set of coordinates {b j (x ijk , y ijk , z ijk )} of the limb comparison points to obtain the coordinates of the torso comparison points and the set of average coordinates of the limb comparison points, satisfying: Where n is the total number of camera devices; S44. Establish a limb comparison vector; the limb comparison vector S5. Vector comparison: For any limb comparison point b j , obtain the spatial angle α between the limb reference vector and the limb comparison vector j , and arrange them in descending order to obtain an action deviation sequence {α j}; S6. Evaluation; The evaluator scores the participants based on the action deviation sequence {α j}, and uses the interval method for scoring. According to the size of the spatial angle α j it is divided into five-level standard degrees, and specific scores G are given: Where G is the specific score.

2. The method for evaluating the quality of physical fitness event actions based on marker distribution learning according to claim 1, wherein: After the score information is generated, predict the standard degree and score of the physical fitness actions of other competition videos with the same actions as the main key frame images based on the score information.

3. The method for evaluating the quality of physical fitness event actions based on marker distribution learning according to claim 2, wherein: Regard the main key frame images with a standard degree of level one as sample images for storage to obtain a sample image library.

4. The method for evaluating the quality of physical fitness event actions based on marker distribution learning according to claim 3, wherein: When the sample images in the sample image library gradually increase, delete the redundant sample images in ascending order according to the number of limb comparison points with a standard degree of level one in the sample images, and store the new sample images.

5. The method for evaluating the quality of physical fitness event actions based on marker distribution learning according to claim 4, wherein: The deletion of the redundant sample images in ascending order is completed by the bubble method.

6. The method for evaluating the quality of physical fitness event actions based on marker distribution learning according to claim 5, wherein: When a person uses the sample image library for training, recommend the sample images in descending order according to the number of limb comparison points with a standard degree of level one in the sample images.

7. The method for evaluating the quality of physical fitness event actions based on marker distribution learning according to claim 6, characterized in that: The recommendation of the sample images in descending order is completed by the bubble method.

Images