Anti-cheating Method and System for Sports Videos

Abstract

The present invention provides a method for preventing cheating in sports videos, comprising the steps of: S1: obtaining sampled skeleton point information from a sampled sports video; S2: obtaining a sampled feature vector according to the sampled skeleton point information; S3: determining whether a comparison feature vector matches the sampled feature vector to determine whether a user cheats, capable of determining whether the sampled sports video newly uploaded or real-time recorded by the user cheats. The method for preventing cheating in sports videos has no requirements on the user's movement actions and has universality. The present invention also provides a system for preventing cheating in sports videos.

Description

Technical Field

[0001] The present invention relates to the field of recognition, and particularly to a method and system for preventing cheating in sports videos. Background Art

[0002] Currently, there is no sports video anti-cheating system based on skeleton recognition for determining whether a sports video uploaded by a user has been repeatedly processed, and users can upload the same sports video after repeated processing for cheating.

[0003] Therefore, it is necessary to develop a new method and system for preventing cheating in sports videos to solve the above problems existing in the prior art. Summary of the Invention

[0004] The purpose of the present invention is to provide a method and system for preventing cheating in sports videos to identify whether a sports video uploaded by a user is cheating.

[0005] To achieve the above purpose, the method for preventing cheating in sports videos according to the present invention includes the steps of:

[0006] S1: Obtain sampled skeleton point information from a sampled sports video;

[0007] S2: Obtain a sampled feature vector according to the sampled skeleton point information;

[0008] S3: Determine whether a comparison feature vector matches the sampled feature vector to determine whether the user is cheating.

[0009] The beneficial effect of the method for preventing cheating in sports videos according to the present invention lies in: through S2: obtaining a sampled feature vector according to the sampled skeleton point information; S3: determining whether a comparison feature vector matches the sampled feature vector to determine whether the sampled sports video newly uploaded or recorded in real time by the user is cheating, and the method for preventing cheating in sports videos has no requirements for the user's movement actions and has universality.

[0010] Optionally, before the step of obtaining sampled skeleton point information from a sampled sports video, it includes:

[0011] Obtain comparison frame images from a comparison sports video, then obtain comparison skeleton information from the comparison frame images, then obtain comparison skeleton point information according to the comparison skeleton information, and then obtain a comparison feature vector according to the comparison skeleton point information.

[0012] Optionally, the step of obtaining sampled skeleton point information from a sampled sports video includes:

[0013] Obtain sampled frame images from a sampled sports video, then obtain sampled skeleton information from the sampled frame images, and then obtain the sampled skeleton point information according to the sampled skeleton information.

[0014] Optionally, the obtaining of the sampling feature vector according to the sampled skeleton point information includes:

[0015] The sampled skeleton information includes a sampled skeleton, the sampled skeleton includes a plurality of sampled skeleton points, the sampled skeleton point information includes relative position data between a plurality of the sampled skeleton points, and the relative position data between the plurality of sampled skeleton points is normalized to obtain the sampling feature vector.

[0016] Optionally, the determining whether the comparison feature vector matches the sampling feature vector to determine whether the user cheats includes:

[0017] Calculating the matching degree between the sampling feature vector and the comparison feature vector, and then determining whether the matching degree reaches a preset threshold to determine whether the sampling feature vector matches the comparison feature vector.

[0018] Optionally, the determining whether the matching degree between the sampling feature vector and the comparison feature vector reaches a preset threshold to determine whether the sampling feature vector matches the comparison feature vector includes: if it is determined that the matching degree between the sampling feature vector and the comparison feature vector reaches the preset threshold, adding n to the number of matches and adding m to the number of calculations; if it is determined that the matching degree between the sampling feature vector and the comparison feature vector does not reach the preset threshold, adding m to the number of calculations, and then determining whether the number of calculations is less than a preset number of calculations, where n and m are natural numbers greater than 0;

[0019] If the number of calculations is less than the preset number of calculations, execute step S1; if the number of calculations is equal to the preset number of calculations, compare the ratio of the number of matches to the preset number of calculations with a preset cheating determination ratio. If the ratio of the number of matches to the preset number of calculations is greater than the cheating determination ratio, determine that the user cheats; if the ratio of the number of matches to the preset number of calculations is less than or equal to the cheating determination ratio, determine that the user does not cheat.

[0020] The present invention also provides a sports video anti-cheating system, including: a sampling key point obtaining module, configured to obtain sampled skeleton point information from a sampled sports video; a sampling feature vector obtaining module, configured to obtain a sampling feature vector according to the sampled skeleton point information; and a determining module, configured to determine whether a comparison feature vector matches the sampling feature vector to determine whether the user cheats.

[0021] Optionally, the sports video anti-cheating system further includes a comparison frame image acquisition module, a comparison skeleton information acquisition module, a comparison skeleton point information acquisition module, and a comparison feature vector acquisition module. The comparison frame image acquisition module is configured to acquire comparison frame images from a comparison sports video. The comparison skeleton information acquisition module is configured to acquire comparison skeleton information from the comparison frame images. The comparison skeleton point information acquisition module is configured to acquire comparison skeleton point information based on the comparison skeleton information. The comparison feature vector acquisition module is configured to acquire comparison feature vectors based on the comparison skeleton point information.

[0022] Optionally, the sampling key point acquisition module includes a sampling frame image acquisition module, a sampling skeleton information acquisition module, and a sampling skeleton point information acquisition module. The sampling frame image acquisition module is configured to acquire sampling frame images from a sampling sports video. The sampling skeleton information acquisition module is configured to acquire sampling skeleton information from the sampling frame images. The sampling skeleton point information acquisition module is configured to acquire sampling skeleton point information from the sampling skeleton information.

[0023] Optionally, the judgment module includes a matching degree calculation module, a first judgment module, a matching count module, a calculation count module, a second judgment module, a sampling key point repeated acquisition module, and a third judgment module. The matching degree calculation module is configured to calculate the matching degree between the sampling feature vector and the comparison feature vector. The first judgment module is configured to judge whether the matching degree reaches a preset threshold. The matching count module is configured to record the number of times the matching degree reaches the preset threshold. The calculation count module is configured to record the number of times the matching degree calculation module calculates. The second judgment module is configured to judge whether the number of calculation times is less than a preset number of calculation times. If the number of calculation times is less than the preset number of calculation times, the sampling key point repeated acquisition module sends an instruction to cause the sampling key point acquisition module to acquire the sampling skeleton point information from the sampling sports video again. The third judgment module is configured to judge whether the ratio of the number of matching times to the preset number of calculation times is greater than a preset cheating judgment ratio.

[0024] The beneficial effects of the sports video anti-cheating system according to the present invention are as follows: a sampling key point acquisition module, configured to acquire sampling skeleton points and sampling body part information from a sampling sports video; a sampling feature vector acquisition module, configured to acquire sampling feature vectors based on the sampling skeleton points and the sampling body part information; a judgment module, configured to judge whether a comparison feature vector matches the sampling feature vector to determine whether a user cheats. By setting the sampling key point acquisition module, the sampling feature vector acquisition module, and the judgment module, it is determined whether the sampling sports video newly uploaded or recorded in real time by the user cheats. The sports video anti-cheating system has no requirements for the user's movement actions and has universality. Description of the Drawings

[0025] Figure 1 Schematic flowchart of the method for preventing cheating in sports videos according to an embodiment of the present invention;

[0026] Figure 2 Block diagram of the structure of the system for preventing cheating in sports videos according to an embodiment of the present invention. Detailed implementation manners

[0027] To make the objectives, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below. Apparently, the described embodiments are some but not all of the embodiments of the present invention. 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. Unless otherwise defined, the technical terms or scientific terms used herein shall have the ordinary meanings as understood by those of ordinary skill in the art in the field to which the present invention belongs. The terms such as "including" used herein are intended to mean that the elements or objects appearing before this term cover the elements or objects listed after this term and their equivalents, without excluding other elements or objects.

[0028] To solve the problems existing in the prior art, the embodiments of the present invention provide a method and a system for preventing cheating in sports videos to identify whether the sports videos uploaded by users are cheating.

[0029] Figure 1 Schematic flowchart of the method for preventing cheating in sports videos according to an embodiment of the present invention.

[0030] In the embodiments of the present invention, referring to Figure 1 , the method for preventing cheating in sports videos includes the following steps:

[0031] S1: Obtain sampling skeleton point information from the sampled sports video;

[0032] S2: Obtain a sampling feature vector according to the sampling skeleton point information;

[0033] S3: Determine whether the comparison feature vector matches the sampling feature vector to determine whether the user is cheating. In some embodiments, before the step of obtaining sampling skeleton point information from the sampled sports video, the following steps are included:

[0034] Obtain comparison frame images from the comparison sports video, then obtain comparison skeleton information from the comparison frame images, then obtain comparison skeleton point information according to the comparison skeleton information, and then obtain a comparison feature vector according to the comparison skeleton point information.

[0035] In some embodiments, the comparison skeleton information includes a comparison skeleton, the comparison skeleton includes a number of comparison skeleton points, the comparison skeleton point information includes relative position data between a number of the comparison skeleton points, and the relative position data between the number of the comparison skeleton points is normalized to obtain the comparison feature vector. Specifically, the comparison skeleton includes x comparison skeleton points, the comparison skeleton point information includes relative position data between the x comparison skeleton points, x is a positive integer, and the relative position data between the x comparison skeleton points is normalized to obtain an x-dimensional vector as the comparison feature vector.

[0036] In some specific embodiments, the comparison skeleton points include the points corresponding to the positions of the nose of the comparison skeleton, the neck of the comparison skeleton, the right shoulder of the comparison skeleton, the right elbow of the comparison skeleton, the right wrist of the comparison skeleton, the left shoulder of the comparison skeleton, the left elbow of the comparison skeleton, the left wrist of the comparison skeleton, the right arm of the comparison skeleton, the right knee of the comparison skeleton, the right ankle of the comparison skeleton, the left hip of the comparison skeleton, the left knee of the comparison skeleton, and the left ankle of the comparison skeleton in the comparison frame image, but are not limited thereto.

[0037] In some specific embodiments, a first photo is obtained from the comparison motion video in the database as a first comparison frame image, then a first comparison skeleton is obtained from the first comparison frame image, then 14 first comparison skeleton points are obtained from the first comparison skeleton, and the relative position data between the 14 first comparison skeleton points is normalized to obtain a 14-dimensional first comparison feature vector.

[0038] In some embodiments, the comparison motion video is all the motion videos in the database to prevent users from switching different accounts to cheat.

[0039] In some embodiments, the step of obtaining the sampled skeleton point information from the sampled motion video includes:

[0040] Obtain a sampled frame image from the sampled motion video, then obtain sampled skeleton information from the sampled frame image, and then obtain the sampled skeleton point information according to the sampled skeleton information.

[0041] In some embodiments, the sampled skeleton information includes a sampled skeleton, the sampled skeleton includes a number of sampled skeleton points, the sampled skeleton point information includes relative position data between a number of the sampled skeleton points, and the relative position data between the number of the sampled skeleton points is normalized to obtain the sampled feature vector. Specifically, the sampled skeleton includes y sampled skeleton points, the sampled skeleton point information includes relative position data between the y sampled skeleton points, where y is a positive integer, and the relative position data between the y sampled skeleton points is normalized to obtain a y-dimensional vector as the sampled feature vector.

[0042] In some specific embodiments, the sampled skeleton points include the points corresponding to the positions of the nose of the sampled skeleton, the neck of the sampled skeleton, the right shoulder of the sampled skeleton, the right elbow of the sampled skeleton, the right wrist of the sampled skeleton, the left shoulder of the sampled skeleton, the left elbow of the sampled skeleton, the left wrist of the sampled skeleton, the right arm of the sampled skeleton, the right knee of the sampled skeleton, the right ankle of the sampled skeleton, the left hip of the sampled skeleton, the left knee of the sampled skeleton, and the left ankle of the sampled skeleton in the sampled frame image, but are not limited thereto.

[0043] In some specific embodiments, the duration of the sampled motion video newly uploaded by the user is 30 minutes. A second photo is obtained from the part of the sampled motion video from 1 to 15 minutes as the first sampled frame image, then a first sampled skeleton is obtained from the first sampled frame image, then 14 first sampled skeleton points are obtained from the first sampled skeleton, and the relative position data between the 14 first sampled skeleton points is normalized to obtain a 14-dimensional first sampled feature vector.

[0044] In some specific embodiments, the sampled motion video is a motion video uploaded by the user or a motion video recorded by the user in real time.

[0045] In some embodiments, determining whether the comparison feature vector matches the sampled feature vector to determine whether the user cheats includes:

[0046] Calculating the matching degree between the sampled feature vector and the comparison feature vector, and then determining whether the matching degree reaches a preset threshold to determine whether the sampled feature vector matches the comparison feature vector.

[0047] In some embodiments, determining whether the matching degree between the sampled feature vector and the comparison feature vector reaches a preset threshold to determine whether the sampled feature vector matches the comparison feature vector includes:

[0048] If it is determined that the matching degree between the sampled feature vector and the comparison feature vector reaches a preset threshold, then add n to the number of matches and add m to the number of calculations. If it is determined that the matching degree between the sampled feature vector and the comparison feature vector does not reach the preset threshold, then add m to the number of calculations, and then determine whether the number of calculations is less than a preset number of calculations. Both n and m are natural numbers greater than 0;

[0049] If the number of calculations is less than the preset number of calculations, then execute step S1. If the number of calculations is equal to the preset number of calculations, then compare the ratio of the number of matches to the preset number of calculations with a preset cheating judgment ratio. If the ratio of the number of matches to the preset number of calculations is greater than the cheating judgment ratio, then it is determined that the user is cheating. If the ratio of the number of matches to the preset number of calculations is less than or equal to the cheating judgment ratio, then it is determined that the user is not cheating.

[0050] In some specific embodiments, calculate the first matching degree between the first sampled feature vector and the first comparison feature vector. The first matching degree is 70%. Then it is determined that the first matching degree of 70% does not reach the preset threshold of 90%. The first sampled feature vector does not match the first comparison feature vector. The number of matches is 1, and the number of calculations is 5. Then add 1 to the number of calculations, and the number of calculations is updated to 6. The preset number of calculations is 6, and the number of calculations is equal to the preset number of calculations. Then it is determined that the ratio of the number of matches 1 to the preset number of calculations 6, which is one-sixth, is less than the preset cheating judgment ratio of one-half. Then it is determined that the user is not cheating.

[0051] In some other specific embodiments, the first matching degree between the first sampled feature vector and the first comparison feature vector is calculated. The first matching degree is 93%. Then it is determined that the first matching degree of 93% reaches the preset threshold of 90%, and the first sampled feature vector matches the first comparison feature vector. The number of matching times is 3, and the number of calculation times is 4. Then the number of matching times is incremented by 1, and the number of matching times is updated to 4. The number of calculation times is incremented by 1, and the number of calculation times is updated to 5. The preset number of calculation times is 6, and the number of calculation times is less than the preset number of calculation times. Then a third photo is obtained from the 1 - 7.5 - minute part of the sampled motion video as the second sampled frame image. Then a second sampled skeleton is obtained from the second sampled frame image. Then 14 second sampled skeleton points are obtained from the second sampled skeleton. The relative position data between the 14 second sampled skeleton points is normalized to obtain a 14 - dimensional second sampled feature vector. Then the second matching degree between the second sampled feature vector and the first comparison feature vector is calculated. The second matching degree is 96%. Then it is determined that the second matching degree of 96% reaches the preset threshold of 90%, and the second sampled feature vector matches the first comparison feature vector. Then the number of matching times is incremented by 1, the number of matching times is updated to 5, the number of calculation times is incremented by 1, and the number of calculation times is updated to 6. The number of calculation times is equal to the preset number of calculation times. Then it is determined that the ratio of the number of matching times 5 to the preset number of calculation times 6, which is five - sixths, is greater than the preset cheating judgment ratio of one - half. Then it is determined that the user cheats.

[0052] Figure 2 The structural block diagram of the motion video anti - cheating system according to the embodiment of the present invention.

[0053] Referring to Figure 2 As shown in, an embodiment of the present invention provides a motion video anti - cheating system. The motion video anti - cheating system includes: a sampled key point acquisition module 100, configured to acquire sampled skeleton point information from a sampled motion video; a sampled feature vector acquisition module 200, configured to acquire a sampled feature vector according to the sampled skeleton point information; and a judgment module 300, configured to judge whether a comparison feature vector matches the sampled feature vector to determine whether a user cheats.

[0054] In some embodiments, referring to Figure 2, the motion video anti-cheating system further includes a comparison frame image acquisition module 400, a comparison skeleton information acquisition module 401, a comparison skeleton point information acquisition module 402, and a comparison feature vector acquisition module 403. The comparison frame image acquisition module 400 is configured to acquire comparison frame images from a comparison motion video. The comparison skeleton information acquisition module 401 is configured to acquire comparison skeleton information from the comparison frame images. The comparison skeleton point information acquisition module 402 is configured to acquire comparison skeleton point information according to the comparison skeleton information. The comparison feature vector acquisition module 403 is configured to acquire comparison feature vectors according to the comparison skeleton point information.

[0055] In some embodiments, referring to Figure 2 , the sampling key point acquisition module 100 includes a sampling frame image acquisition module 101, a sampling skeleton information acquisition module 102, and a sampling skeleton point information acquisition module 103. The sampling frame image acquisition module 101 is configured to acquire sampling frame images from a sampling motion video. The sampling skeleton information acquisition module 102 is configured to acquire sampling skeleton information from the sampling frame images. The sampling skeleton point information acquisition module 103 is configured to acquire sampling skeleton point information from the sampling skeleton information.

[0056] In some embodiments, referring to Figure 2 , the judgment module 300 includes a matching degree calculation module 301, a first judgment module 302, a matching count module 303, a calculation count module 304, a second judgment module 305, a sampling key point repeated acquisition module 306, and a third judgment module 307. The matching degree calculation module 301 is configured to calculate the matching degree between the sampling feature vector and the comparison feature vector. The first judgment module 302 is configured to judge whether the matching degree reaches a preset threshold. The matching count module 303 is configured to record the number of matches where the matching degree reaches the preset threshold. The calculation count module 304 is configured to record the number of calculations of the matching degree calculation module. The second judgment module 305 is configured to judge whether the number of calculations is less than a preset number of calculations. If the number of calculations is less than the preset number of calculations, the sampling key point repeated acquisition module 306 sends an instruction to cause the sampling key point acquisition module 100 to acquire the sampling skeleton point information from the sampling motion video again. The third judgment module 307 is configured to judge whether the ratio of the number of matches to the preset number of calculations is greater than a preset cheating judgment ratio.

[0057] In some embodiments, the operation method of the motion video anti-cheating system of the present invention is as described in the motion video anti-cheating method of the present invention.

[0058] In some embodiments, referring to Figure 2, the comparison frame image acquisition module 400 acquires a second comparison frame image from the comparison motion video, then the comparison skeleton information acquisition module 401 acquires second comparison skeleton information from the second comparison frame image, the comparison skeleton point information acquisition module 402 acquires second comparison skeleton point information according to the second comparison skeleton information, and the comparison feature vector acquisition module 403 is used to acquire a second comparison feature vector according to the second comparison skeleton point information.

[0059] In some embodiments, referring to Figure 2 , the sampling frame image acquisition module 101 acquires a third sampling frame image from the sampling motion video, then the sampling skeleton information acquisition module 102 acquires third sampling skeleton information from the third sampling frame image, then the sampling skeleton point information acquisition module 103 acquires third sampling skeleton point information from the third sampling skeleton information, then the sampling feature vector acquisition module 200 acquires a third sampling feature vector according to the third sampling skeleton point information, and then the judgment module 300 determines whether the second comparison feature vector matches the third sampling feature vector to determine whether the user cheats.

[0060] In some embodiments, referring to Figure 2 , the judgment module 300 includes a matching degree calculation module 301, a first judgment module 302, a calculation count module 304, a second judgment module 305, and a third judgment module 307. The matching degree calculation module 301 calculates a third matching degree between the third sampling feature vector and the second comparison feature vector, and the third matching degree is 70%. Then the first judgment module 302 determines that the third matching degree of 70% does not reach the preset threshold of 90%, and the third sampling feature vector does not match the second comparison feature vector. The number of matches is 1, and the number of calculations is 5. Then the calculation count module 304 adds 1 to the number of calculations and records the number of calculations as 6. Then the second judgment module 305 determines that the number of calculations 6 is equal to the preset number of calculations 6. Then the third judgment module 307 determines that the ratio of the number of matches 1 to the preset number of calculations 6, which is one-sixth, is less than the preset cheating judgment ratio of one-half. Then the third judgment module 307 determines that the user has not cheated.

[0061] In still other embodiments, referring to Figure 2, the judgment module 300 includes a matching degree calculation module 301, a first judgment module 302, a matching count module 303, a calculation count module 304, a second judgment module 305, a sampling key point repeated acquisition module 306, and a third judgment module 307. The matching degree calculation module 301 calculates the third matching degree between the third sampling feature vector and the second comparison feature vector. The third matching degree is 93%. Then, the first judgment module 302 determines that the third matching degree of 93% reaches the preset threshold of 90%, and the third sampling feature vector matches the second comparison feature vector. The number of matches is 3, and the number of calculations is 4. Then, the matching count module 303 adds 1 to the number of matches and records the number of matches as 4. Then, the calculation count module 304 adds 1 to the number of calculations and records the number of calculations as 5. Then, the second judgment module 305 determines that the number of calculations 5 is less than the preset number of calculations 6. Then, the sampling key point repeated acquisition module 306 sends an instruction to cause the sampling key point acquisition module 100 to obtain the fourth sampling skeleton point information from the sampling motion video again. Then, the sampling feature vector acquisition module 200 obtains the fourth sampling feature vector according to the fourth sampling skeleton point information.

[0062] Refer to Figure 2 , then the matching degree calculation module 301 calculates the fourth matching degree between the fourth sampling feature vector and the second comparison feature vector. The fourth matching degree is 96%. Then, the first judgment module 302 determines that the fourth matching degree of 96% reaches the preset threshold of 90%, and the fourth sampling feature vector matches the second comparison feature vector. Then, the matching count module 303 adds 1 to the number of matches and records the number of matches as 5. Then, the calculation count module 304 adds 1 to the number of calculations and records the number of calculations as 6. Then, the second judgment module 305 determines that the number of calculations 6 is equal to the preset number of calculations 6. Then, the third judgment module 307 determines that the ratio of the number of matches 5 to the preset number of calculations 6, which is five-sixths, is greater than the preset cheating judgment ratio of one-half. Then, the third judgment module 307 determines that the user is cheating.

[0063] Although the embodiments of the present invention have been described in detail above, it is obvious to those skilled in the art that various modifications and changes can be made to these embodiments. However, it should be understood that such modifications and changes are all within the scope and spirit of the present invention described in the claims. Moreover, the present invention described herein may have other embodiments and can be implemented or realized in various ways.

Claims

1. A method for preventing cheating in sports videos, characterized in that, Including the steps: S0: Obtain comparison frame images from the comparison motion video, then obtain comparison skeleton information from the comparison frame images, then obtain comparison skeleton point information based on the comparison skeleton information, and then obtain comparison feature vectors based on the comparison skeleton point information; S1: Obtain sampling frame images from the sampling motion video, then obtain sampling skeleton information from the sampling frame images, and then obtain sampling skeleton point information based on the sampling skeleton information; S2: Obtain sampling feature vectors based on the sampling skeleton point information; S3: Calculate the matching degree between the sampling feature vector and the comparison feature vector, and then determine whether the matching degree reaches a preset threshold to determine whether the sampling feature vector and the comparison feature vector match; if it is determined that the matching degree between the sampling feature vector and the comparison feature vector reaches the preset threshold, add n to the number of matches and add m to the number of calculations; if it is determined that the matching degree between the sampling feature vector and the comparison feature vector does not reach the preset threshold, add m to the number of calculations, and then determine whether the number of calculations is less than the preset number of calculations. n and m are natural numbers greater than 0; If the number of calculations is less than the preset number of calculations, execute step S1; if the number of calculations is equal to the preset number of calculations, compare the ratio of the number of matches to the preset number of calculations with the preset cheating judgment ratio. If the ratio of the number of matches to the preset number of calculations is greater than the cheating judgment ratio, determine that the user is cheating; if the ratio of the number of matches to the preset number of calculations is less than or equal to the cheating judgment ratio, determine that the user is not cheating.

2. The anti-cheating method for sports videos according to claim 1, wherein The obtaining of the sampling feature vector based on the sampling skeleton point information includes: The sampling skeleton information includes a sampling skeleton, the sampling skeleton includes a number of sampling skeleton points, the sampling skeleton point information includes relative position data between a number of the sampling skeleton points, and the relative position data between the number of sampling skeleton points is normalized to obtain the sampling feature vector.

3. A sports video anti-cheating system, characterized in that, Including: A comparison frame image obtaining module, configured to obtain comparison frame images from the comparison motion video; A comparison skeleton information obtaining module, configured to obtain comparison skeleton information from the comparison frame images; A comparison skeleton point information obtaining module, configured to obtain comparison skeleton point information based on the comparison skeleton information, and a comparison feature vector obtaining module, configured to obtain comparison feature vectors based on the comparison skeleton point information; A sampling key point obtaining module, configured to obtain sampling frame images from the sampling motion video, then obtain sampling skeleton information from the sampling frame images, and then obtain sampling skeleton point information based on the sampling skeleton information; A sampling feature vector obtaining module, configured to obtain sampling feature vectors based on the sampling skeleton point information; A judgment module, which is used to calculate the matching degree between the sampled feature vector and the comparison feature vector, and then determine whether the matching degree reaches a preset threshold to determine whether the sampled feature vector matches the comparison feature vector; if it is determined that the matching degree between the sampled feature vector and the comparison feature vector reaches the preset threshold, the matching count is incremented by n and the calculation count is incremented by m, if it is determined that the matching degree between the sampled feature vector and the comparison feature vector does not reach the preset threshold, the calculation count is incremented by m, and then it is determined whether the calculation count is less than a preset calculation count, where n and m are natural numbers greater than 0; If the calculation count is less than the preset calculation count, step S1 is executed. If the calculation count is equal to the preset calculation count, the ratio of the matching count to the preset calculation count is compared with a preset cheating judgment ratio. If the ratio of the matching count to the preset calculation count is greater than the cheating judgment ratio, it is determined that the user is cheating. If the ratio of the matching count to the preset calculation count is less than or equal to the cheating judgment ratio, it is determined that the user is not cheating.

4. The anti-cheating system for sports videos according to claim 3, characterized in that, The sampled key point acquisition module includes a sampled frame image acquisition module, a sampled skeleton information acquisition module, and a sampled skeleton point information acquisition module. The sampled frame image acquisition module is used to acquire a sampled frame image from the sampled motion video. The sampled skeleton information acquisition module is used to acquire sampled skeleton information from the sampled frame image. The sampled skeleton point information acquisition module is used to acquire sampled skeleton point information from the sampled skeleton information.

5. The anti-cheating system for sports videos according to claim 3, characterized in that, The judgment module includes a matching degree calculation module, a first judgment module, a matching count module, a calculation count module, a second judgment module, a sampled key point repeated acquisition module, and a third judgment module. The matching degree calculation module is used to calculate the matching degree between the sampled feature vector and the comparison feature vector. The first judgment module is used to determine whether the matching degree reaches a preset threshold. The matching count module is used to record the matching count when the matching degree reaches the preset threshold. The calculation count module is used to record the calculation count of the matching degree calculation module. The second judgment module is used to determine whether the calculation count is less than the preset calculation count. If the calculation count is less than the preset calculation count, the sampled key point repeated acquisition module sends an instruction to cause the sampled key point acquisition module to acquire the sampled skeleton point information from the sampled motion video again. The third judgment module is used to determine whether the ratio of the matching count to the preset calculation count is greater than a preset cheating judgment ratio.

Images