CN113939837A - Win-loss prediction system - Google Patents
Win-loss prediction system Download PDFInfo
- Publication number
- CN113939837A CN113939837A CN202080002630.6A CN202080002630A CN113939837A CN 113939837 A CN113939837 A CN 113939837A CN 202080002630 A CN202080002630 A CN 202080002630A CN 113939837 A CN113939837 A CN 113939837A
- Authority
- CN
- China
- Prior art keywords
- game
- given scene
- win
- analysis target
- scene
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000004458 analytical method Methods 0.000 claims abstract description 73
- 238000001514 detection method Methods 0.000 claims abstract description 11
- 238000010801 machine learning Methods 0.000 claims description 69
- 238000010191 image analysis Methods 0.000 claims description 11
- 238000000034 method Methods 0.000 claims description 8
- 238000010586 diagram Methods 0.000 description 14
- 238000013528 artificial neural network Methods 0.000 description 12
- 230000002596 correlated effect Effects 0.000 description 3
- 238000013135 deep learning Methods 0.000 description 3
- 230000000977 initiatory effect Effects 0.000 description 3
- 210000002569 neuron Anatomy 0.000 description 3
- 230000005540 biological transmission Effects 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000003384 imaging method Methods 0.000 description 2
- 238000010219 correlation analysis Methods 0.000 description 1
- 230000007123 defense Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/10—Services
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- A—HUMAN NECESSITIES
- A63—SPORTS; GAMES; AMUSEMENTS
- A63B—APPARATUS FOR PHYSICAL TRAINING, GYMNASTICS, SWIMMING, CLIMBING, OR FENCING; BALL GAMES; TRAINING EQUIPMENT
- A63B71/00—Games or sports accessories not covered in groups A63B1/00 - A63B69/00
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N5/00—Computing arrangements using knowledge-based models
- G06N5/04—Inference or reasoning models
- G06N5/041—Abduction
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/40—Scenes; Scene-specific elements in video content
- G06V20/41—Higher-level, semantic clustering, classification or understanding of video scenes, e.g. detection, labelling or Markovian modelling of sport events or news items
- G06V20/42—Higher-level, semantic clustering, classification or understanding of video scenes, e.g. detection, labelling or Markovian modelling of sport events or news items of sport video content
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Software Systems (AREA)
- Tourism & Hospitality (AREA)
- Mathematical Physics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Engineering & Computer Science (AREA)
- Computing Systems (AREA)
- Economics (AREA)
- Strategic Management (AREA)
- Evolutionary Computation (AREA)
- Human Resources & Organizations (AREA)
- Data Mining & Analysis (AREA)
- Artificial Intelligence (AREA)
- Medical Informatics (AREA)
- General Business, Economics & Management (AREA)
- Marketing (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Primary Health Care (AREA)
- Computational Linguistics (AREA)
- Multimedia (AREA)
- Quality & Reliability (AREA)
- Development Economics (AREA)
- Operations Research (AREA)
- Game Theory and Decision Science (AREA)
- Entrepreneurship & Innovation (AREA)
- Physical Education & Sports Medicine (AREA)
- Image Analysis (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Image Processing (AREA)
Abstract
The invention provides a win/loss prediction system capable of predicting a win/loss result of a match based on an image of the match to be analyzed. The win/loss prediction system (1) is provided with: an image acquisition unit (2) for acquiring an image of a match to be analyzed; a first system (4) for estimating a ball control in a game from an image in the game to be analyzed; a second system (5) for estimating an event occurring in the game from the image of the analysis target in the game; and a third system (6) for predicting the win or loss result of the race based on the detection result of the first system (4) and the detection result of the second system (5).
Description
Technical Field
The present invention relates to a win/loss prediction system capable of predicting a win/loss result of a game.
Background
In the prior art, systems for analyzing the warfare of a football team have been proposed. In a conventional system, for example, win and loss result codes as reference indexes reflecting qualitative strengths of division into a main court and a guest court for each soccer team and other associated data items for each soccer team are collected and stored from an existing database related to the soccer team, the collected win and loss result codes as reference indexes and other associated data items are qualitatively associated by correlation analysis, and qualitative fighting ratings based on linear sum operations of the division into the main court and the guest court for each soccer team are performed based on the associated indexes (for example, see patent document 1).
Documents of the prior art
Patent document
Patent document 1: japanese patent laid-open publication No. 2004-110826
Disclosure of Invention
Problems to be solved by the invention
However, in the existing system, although the fighting power of the soccer team can be analyzed, the win or loss result of the game cannot be predicted.
The present invention has been made in view of the above-described problems, and an object thereof is to provide a win/loss prediction system capable of predicting a win/loss result of a race.
Means for solving the problems
The win/loss prediction system of the present invention comprises: an image acquisition unit that acquires an image of an analysis target in a match; a first system for estimating a ball control in the game; a second system for estimating events occurring in said match; and a third system for predicting a win or loss result of the race based on a detection result of the first system and a detection result of the second system, the first system including: a first machine learning unit that analyzes a relationship between an image of a given scene of a past game and information on positions of players and balls in the given scene of the past game by machine learning; a first estimation unit configured to estimate and output positions of the player and the ball in the given scene of the analysis target game, based on the relationship analyzed by the first machine learning unit, with the image of the given scene in the analysis target game acquired by the image acquisition unit as an input; a first determination unit that determines and outputs a ball control in a predetermined scene of the game to be analyzed based on the positions of the player and the ball output from the first estimation unit, the second system including: a second machine learning unit that analyzes a relationship between an image of a given scene of a past race and information related to an event occurring in the given scene in the past race by machine learning; and a second estimation unit configured to estimate and output an event occurring in a predetermined scene in the race of the analysis target, based on the relationship analyzed by the second machine learning unit, with the image of the predetermined scene in the race of the analysis target acquired by the image acquisition unit as an input, the third system including: a third machine learning unit that analyzes, by machine learning, a relationship between information on a ball control in a given scene of a past game, an event occurring in the given scene, an elapsed time from the start of the game to the given scene, a point of win or loss at a time point of the given scene, and a result of win or loss of the past game; and a third estimating unit that estimates and outputs, based on the relationship analyzed by the third machine learning unit, a win/lose result of the race expected at the time point of the given scene of the race of the analysis target, with information on the ball control in the given scene of the race of the analysis target output from the first determining unit, the event occurring at the given scene of the race of the analysis target output from the second estimating unit, the elapsed time from the start of the race of the analysis target to the given scene, and the win/lose point at the time point of the given scene of the race of the analysis target as input.
According to this configuration, first, when the image of the given scene in the game to be analyzed is input, the positions of the player and the ball in the given scene are estimated using the relationship (relationship between the image of the given scene in the past game and the information on the positions of the player and the ball in the given scene in the past game) analyzed by the machine learning. Then, based on the estimated positions of the player and the ball, the ball control in the predetermined scene is determined. In addition, when the image of the given scene in the game as the analysis target is input, the event occurring in the given scene (for example, the goal is made by initiating an attack before the goal, or the like) is estimated using the relationship analyzed by the machine learning (the relationship between the image of the given scene in the past game and the information on the event occurring in the given scene in the past game). Then, when information on a ball control (determined ball control) in a given scene of a match to be analyzed, an event occurring in the given scene (estimated event), an elapsed time until the given scene, and a point of win or loss in the given scene is input, a win or loss of the match expected in real time at a time point of the given scene is estimated using a relationship (a relationship between a ball control in a given scene of a past match, an event occurring in the given scene, an elapsed time until the given scene, a point of win or loss at a time point of the given scene, and a win or loss result of the past match) analyzed by machine learning. In this way, the outcome of the match can be predicted in real time based on the image of the target of analysis during the match.
In the win/loss prediction system according to the present invention, the third system may further include an image analysis unit that performs image analysis on a video of a predetermined scene in the race of the analysis target, and acquires information on an elapsed time from a start of the race of the analysis target of the predetermined scene to the predetermined scene, and a point of win/loss at a time point of the predetermined scene in the race of the analysis target.
According to this configuration, by performing image analysis on a video of a given scene in a race to be analyzed, it is possible to acquire information on the elapsed time until the given scene and the point of gain or loss at the time point of the given scene. The third estimation unit inputs information (information on the elapsed time to the predetermined scene and the point of gain or loss at the time point of the predetermined scene) acquired by the image analysis to the third estimation unit, and can be used for estimation of the win or loss result of the match.
In the win/loss prediction system according to the present invention, the third estimating unit may estimate and output a win probability, a loss probability, and a tie probability as win/loss results expected in real time at a time point of a predetermined scene of the analysis target race.
According to this configuration, the probability of success, the probability of failure, and the probability of tie are output as the expected success or failure result, and therefore the success or failure result can be understood at a glance.
The win/loss prediction method of the present invention is a win/loss prediction method executed by a win/loss prediction system, and the win/loss prediction system includes: an image acquisition unit that acquires an image of an analysis target in a match; a first system for estimating a ball control in the game; a second system for estimating events occurring in said match; and a third system for predicting a win or loss result of the race based on the detection result of the first system and the detection result of the second system, wherein the first system performs the steps of: a first machine learning step of analyzing, by machine learning, a relationship between an image of a given scene of a past game and information related to positions of players and balls in the given scene of the past game; a first estimation step of estimating and outputting positions of the player and the ball in the given scene of the analysis target game, based on the relationship analyzed in the first machine learning step, with the image of the given scene in the analysis target game acquired by the image acquisition unit as an input; and a first determination step of determining and outputting a ball control in a given scene of the game to be analyzed based on the positions of the player and the ball output from the first estimation step, wherein the second system executes: a second machine learning step of analyzing, by machine learning, a relationship between an image of a given scene of a past game and information related to an event occurring in the given scene in the past game; and a second estimation step of estimating and outputting an event occurring in a predetermined scene in the race of the analysis target, based on the relationship analyzed in the second machine learning step, with the image of the predetermined scene in the race of the analysis target acquired by the image acquisition unit as an input, wherein the third system executes: a third machine learning step of analyzing, by machine learning, a relationship between information on a ball control in a given scene of a past game, an event occurring at the given scene, an elapsed time from a start of the game to the given scene, a point of win or loss at a point of time of the given scene, and a win or loss result of the past game; and a third estimation step of estimating and outputting, based on the relationship analyzed in the third machine learning step, a win/lose result of the race expected at the time point of the given scene of the race of the analysis target, with information on the ball control in the given scene of the race of the analysis target output from the first determination step, the event occurring at the given scene of the race of the analysis target output from the second estimation step, the elapsed time from the start of the race of the analysis target to the given scene, and the win/lose point at the time point of the given scene of the race of the analysis target as input.
With this method, as with the system described above, first, an image of a given scene in a game to be analyzed is input, and the positions of players and balls in the given scene are estimated using the relationship (relationship between the image of the given scene of the past game and information on the positions of the players and balls in the given scene of the past game) analyzed by machine learning. Then, based on the estimated positions of the player and the ball, the ball control in the predetermined scene is determined. In addition, when the image of the given scene in the game as the analysis target is input, the event occurring in the given scene (for example, the goal is made by initiating an attack before the goal, or the like) is estimated using the relationship analyzed by the machine learning (the relationship between the image of the given scene in the past game and the information on the event occurring in the given scene in the past game). Then, information on a ball control (determined ball control) in a given scene of a game to be analyzed, an event occurring in the given scene (estimated event), an elapsed time until the given scene, and a point of win or loss in the given scene is input, and a win or loss result of the game expected in real time at a time point of the given scene is estimated using a relationship (a relationship between a ball control in a given scene of a past game, an event occurring in the given scene, an elapsed time until the given scene, information on a point of win or loss at a time point of the given scene, and a win or loss result of the past game) analyzed by machine learning. In this way, the win or loss result of the match can be predicted in real time from the image of the match to be analyzed.
Effects of the invention
According to the present invention, the win or loss result of the match can be predicted from the image in the match to be analyzed.
Drawings
Fig. 1 is a diagram showing a configuration of a win/loss prediction system according to an embodiment of the present invention.
FIG. 2 is a block diagram of a first system of an embodiment of the invention.
Fig. 3 is a diagram showing an example of determination of the ball control according to the embodiment of the present invention.
Fig. 4 is a block diagram of a second system of an embodiment of the present invention.
Fig. 5 is a diagram showing an example of events in the embodiment of the present invention.
Fig. 6 is a block diagram of a third system of an embodiment of the present invention.
Fig. 7 is a diagram showing an example of win/loss prediction according to the embodiment of the present invention.
Fig. 8 is a diagram showing the operation (flow of processing) of the win/loss prediction system according to the embodiment of the present invention.
Detailed Description
Hereinafter, a win/loss prediction system according to an embodiment of the present invention will be described with reference to the drawings. In the present embodiment, a win/loss prediction system for predicting win/loss in a game such as soccer is exemplified.
The configuration of the win/loss prediction system according to the embodiment of the present invention will be described with reference to the drawings. Fig. 1 is a diagram showing a configuration of a win/loss prediction system according to the present embodiment. As shown in fig. 1, the win or loss prediction system 1 is connected to a match shooting system 10 via a network N. The match shooting system 10 includes a shooting unit 11 that shoots a video in a match such as a soccer game, and a video transmission unit 12 that transmits the shot video via the network N.
As shown in fig. 1, the win/loss prediction system 1 includes: an image acquisition unit 2 that acquires an in-game image (an in-game image to be analyzed) sent from the game imaging system 10; a first system 4 for estimating a ball control in a game; a second system 5 for estimating an event occurring in the race; and a third system 6 for predicting the win or loss of the match based on the detection results of the first system 4 and the second system 5. The win/loss prediction system 1 includes a video storage unit 3 for storing video data of a past race. The video storage unit 3 may store the video during the race (the video during the race to be analyzed) sent from the race shooting system 10.
Fig. 2 is a block diagram of the first system 4. As shown in fig. 2, the first system 4 includes a first input/output unit 40, a first machine learning unit 41, a first estimation unit 42, and a first determination unit 43.
The images of the past race stored in the image storage unit 3 and the images of the race to be analyzed acquired by the image acquisition unit 2 are input to the first input/output unit 40. The information of the determination result in the first determination unit 43 (information of the ball control in a given scene of the game to be analyzed) is output from the first input/output unit 40.
The first machine learning unit 41 analyzes the relationship between the image of the given scene of the past game stored in the image storage unit 3 and the information on the positions of the players and the ball in the given scene of the past game by machine learning. In this machine learning, any method such as deep learning by a neural network is used. For example, in the case of a neural network, an image of a given scene of a past game is input to an input layer, and information on the positions of players and balls in the given scene of the past game is output from an output layer. Then, the weighting coefficients between the neurons of the neural network are optimized by supervised learning using analysis data in which data input to the input layer and data output from the output layer are correlated.
The first estimation unit 42 estimates and outputs the positions of the player and the ball in the given scene of the analysis target game, based on the relationship analyzed by the first machine learning unit 41, with the image of the given scene in the analysis target game acquired by the image acquisition unit 2 as an input. For example, in the case of the neural network described above, the estimation is performed by inputting the image of the predetermined scene in the game of the analysis target acquired by the image acquisition unit 2 into the input layer and outputting the information on the positions of the player and the ball in the predetermined scene in the game of the analysis target from the output layer.
The first determination unit 43 determines and outputs the ball control (which team holds the ball) in a predetermined scene of the game to be analyzed, based on the positions of the player and the ball output from the first estimation unit 42. For example, the first determination unit 43 determines, based on the distance between the ball and the player, that the player side team closest to the distance between the ball is the team holding the ball.
Fig. 3 is a diagram showing an example of determination of the ball control. In fig. 3, the player of the team (team a) who is shot from left to right is shown by "hatched triangle", and the player of the team (team B) who is shot from right to left is shown by "white triangle". Additionally, the ball is illustrated with a "black circle". In the example of fig. 3, since the player closest to the ball is the player of team a, team a is determined to be the ball team holding the ball.
Fig. 4 is a block diagram of the second system 5. As shown in fig. 4, the second system 5 includes a second input/output unit 50, a second device learning unit 51, and a second estimating unit 52.
The video of the past game stored in the video storage unit 3 is input to the second input/output unit 50. The information of the estimation result in the second estimation unit 52 (information of an event occurring in a given scene of the game to be analyzed) is output from the second input/output unit 50.
The second machine learning portion 51 analyzes the relationship between the image of the given scene of the past race and the information related to the event occurring in the given scene in the past race by machine learning. In this machine learning, any method such as deep learning by a neural network is used. For example, in the case of a neural network, the input layer is configured to input images of a given scene of a past game, and the output layer is configured to output information related to an event occurring in the given scene of the past game. Then, the weighting coefficients between the neurons of the neural network are optimized by supervised learning using analysis data in which data input to the input layer and data output from the output layer are correlated.
The second estimating unit 52 estimates and outputs an event occurring in a given scene in the race to be analyzed, based on the relationship analyzed by the second machine learning unit 51, with the video of the given scene in the race to be analyzed acquired by the video acquiring unit 2 as an input. For example, in the case of the neural network described above, the estimation is performed by inputting the video of a predetermined scene in the race to be analyzed, which is acquired by the video acquisition unit 2, into the input layer and outputting information on an event occurring in the predetermined scene in the race to be analyzed from the output layer.
Fig. 5 is a diagram showing an example of an event occurring in a given scene in a race. In fig. 5, an example of an event occurring with respect to a team holding a ball is shown. Examples of the event in the case where the attack is successful include "dribble-breaking success", "pass success", and "shooting success". Examples of the event in the case of failure of the attack include "dribbling-through failure", "pass failure", and "shot failure".
Fig. 6 is a block diagram of the third system 6. As shown in fig. 6, the third system 6 includes a third input/output unit 60, a third device learning unit 61, and a third estimation unit 62.
The third input/output unit 60 is input with ball control in a given scene of a past game, information on an event occurring in the given scene, an elapsed time from the start of the game to the given scene, a point of gain and loss at the time point of the given scene, ball control in the given scene of the game of the analysis target output from the first determination unit 43, an event occurring in the given scene of the game of the analysis target output from the second estimation unit 52, an elapsed time from the start of the game of the analysis target to the given scene, and information on a point of gain and loss at the time point of the given scene of the game of the analysis target. The information of the estimation result in the third estimation unit 62 (information of the win or lose result of the game expected at the time point of the given scene of the game as the analysis target) is output from the third input/output unit 60.
The third machine learning section 61 analyzes the relationship between the information on the point of win or loss at the time point of the given scene and the win or loss result of the past game by machine learning, the ball control in the given scene of the past game, the event occurring at the given scene, the elapsed time from the start of the game to the given scene. In the machine learning, any method such as deep learning by a neural network can be used. For example, if it is a neural network, the composition i inputs information on a ball control in a given scene of a past game, an event occurring in the given scene, an elapsed time from the start of the game to the given scene, a point of win or loss at a time point of the given scene into the input layer, and outputs information on a win or loss result of the past game from the output layer. Then, the weighting coefficients between the neurons of the neural network are optimized by supervised learning using analysis data in which data input to the input layer and data output from the output layer are correlated.
The third estimating unit 62 estimates and outputs the result of winning or losing of the game expected at the time point of the given scene of the game of the analysis target, based on the relationship analyzed by the third machine learning unit 61, by inputting the information on the ball control in the given scene of the game of the analysis target output from the first determining unit 43, the event occurring at the given scene of the game of the analysis target output from the second estimating unit 52, the elapsed time from the start of the game of the analysis target to the given scene, and the point of winning or losing at the time point of the given scene of the game of the analysis target. For example, in the case of the neural network described above, the estimation is performed by inputting information on the ball control in the predetermined scene of the race to be analyzed output from the first determination unit 43, the event occurring in the predetermined scene of the race to be analyzed output from the second estimation unit 52, the elapsed time from the start of the race to the predetermined scene of the race to be analyzed, and the point of gain and loss at the time point of the predetermined scene of the race to be analyzed into the input layer, and outputting information on the result of the win and loss of the race expected at the time point of the predetermined scene of the race to be analyzed from the output layer.
The third system 6 may further include an image analysis unit (not shown) that performs image analysis of a video of a predetermined scene in the race to be analyzed, and thereby, information on an elapsed time from the start of the race to be analyzed of the predetermined scene to the predetermined scene and a point of gain and loss at the time point of the predetermined scene in the race to be analyzed. In this image analysis, a known technique can be used. The third estimation unit 62 estimates and outputs the probability of winning, the probability of negation, and the probability of tie, as the result of winning or negation expected at the time point of a given scene of the game to be analyzed.
The third estimating unit 62 may be configured to input the past scores of the players (first place, halfway place), the attack statistical data (dribbling distance, passing success rate, shooting success rate, etc.) of each team, the defense statistical data (number of deliquesces, number of goals broken, etc.) of each team, the physical statistical data (number of decisions, etc.) of each team, the number of violations (number of violations, number of yellow/red boards, etc.) of each team, the past game scores between the team and the opponent team, the nearest scores (negative/score/lost score) between the team and the opponent team at the time of the game, the order difference between the team and the game at the time of the game, the joint game/region/meeting information, weather (clear, cloudy, rainy, etc.), air temperature/humidity, court number, number of views, weather, and the number of players, Time of opening, date of opening (year/month), etc.
The third estimation unit 62 may output information of the final score of the game, the middle score of the game, the team/player who scores the next, the elapsed time until the next score, the team/player who performs the next action (the action such as shooting, dribbling, passing, or throwing), the elapsed time until the next action, the next player who replaces the next action, and the like.
The operation of the win/loss prediction system 1 configured as described above will be described with reference to the flowchart of fig. 8.
In the win/loss prediction system 1 of the present embodiment, first, as a preliminary preparation, the first system 4 analyzes the relationship between the video of the given scene of the past game and the information on the positions of the players and the ball in the given scene of the past game by machine learning (first machine learning step). In addition, in the second system 5, the relationship between the image of the given scene of the past race and the information on the event occurring in the given scene in the past race is analyzed by machine learning (second machine learning step). Further, in the third system 6, the relationship between the information on the ball control in a given scene of the past game, the event occurring in the given scene, the elapsed time from the start of the game to the given scene, the point of win or loss at the time point of the given scene, and the win or loss result of the past game is analyzed by machine learning (third machine learning step).
Then, as shown in fig. 8, in the case of predicting the win or loss result of the game as the analysis target, the image captured by the game capturing system 10 (the image in the game as the analysis target) is acquired (S1), the positions of the player and the ball in the given scene of the game as the analysis target are estimated and output with the image of the given scene in the game as the input acquired by the image acquiring unit 2 based on the relationship analyzed in the first machine learning step (S2), and the ball control in the given scene of the game as the analysis target is determined and output based on the positions of the player and the ball as the output (S3).
Next, in the second system 5, based on the relationship analyzed in the second machine learning step, the video of the given scene in the race to be analyzed, which is acquired by the video acquisition unit 2, is input, and an event occurring in the given scene in the race to be analyzed is estimated and output (S4). Then, the third system 6 receives as input the information on the ball control in the given scene of the race to be analyzed output from the first determination step, the event occurring in the given scene of the race to be analyzed output from the second estimation step, the elapsed time from the start of the race to the given scene of the race to be analyzed, and the point of gain or loss at the time point of the given scene of the race to be analyzed, and estimates and outputs the result of win or loss of the race expected at the time point of the given scene of the race to be analyzed (S5).
According to the win/lose prediction system 1 of the present embodiment, the win/lose result of the soccer game can be predicted.
That is, in the present embodiment, first, when an image of a given scene in a game to be analyzed is input, the positions of players and balls in the given scene are estimated using a relationship (relationship between an image of a given scene in a past game and information on the positions of players and balls in the given scene in the past game) analyzed by machine learning. Then, based on the estimated positions of the player and the ball, the ball control in the predetermined scene is determined.
When the image of the given scene in the game to be analyzed is input, the event occurring in the given scene (for example, a goal shot is made by initiating an attack before a goal) is estimated using the relationship analyzed by machine learning (the relationship between the image of the given scene in the past game and the information on the event occurring in the given scene in the past game).
Then, when information on a ball control (determined ball control) in a given scene of a game to be analyzed, an event (estimated event) occurring in the given scene, an elapsed time until the given scene, and a point of win or loss in the given scene is input, a win or loss result of the game expected in real time at a time point of the given scene is estimated using a relationship (a relationship between a ball control in a given scene of a past game, an event occurring in the given scene, an elapsed time until the given scene, information on a point of win or loss at a time point of the given scene, and a win or loss result of the past game) analyzed by machine learning. In this way, the win or loss result of the match can be predicted in real time from the image in the match to be analyzed.
In the present embodiment, by performing image analysis on a video of a predetermined scene in a race to be analyzed, it is possible to acquire information on the elapsed time until the predetermined scene and the point of gain or loss at the time point of the predetermined scene. The information (information on the elapsed time to the predetermined scene and the point of gain or loss at the time point of the predetermined scene) acquired by the image analysis can be input to the third estimation unit 62, and used for estimation of the win or loss result of the match.
In the present embodiment, the probability of success, the probability of failure, and the probability of tie are output as the expected success or failure result, and therefore, the success or failure result can be easily understood at a glance.
The embodiments of the present invention have been described above by way of examples, but the scope of the present invention is not limited thereto, and modifications and variations can be made within the scope of the claims depending on the purpose.
For example, although the above description has been made on an example of predicting the outcome of a soccer game, the present invention can be similarly applied to the outcome prediction of a game other than soccer (for example, football, american football, basketball, and hockey).
(Industrial Applicability)
As described above, the win/loss prediction system according to the present invention has an effect of predicting the win/loss result of a game to be analyzed from an image of the game, and is useful as a win/loss prediction system for a game such as soccer.
(description of reference numerals)
1 win-loss prediction system
2 image acquisition part
3 video storage part
4 first system
5 second System
6 third System
10 match shooting system
11 imaging unit
12 image transmission part
40 first input/output unit
41 first machine learning unit
42 first estimating part
43 first judging section
50 second input/output unit
51 second machine learning unit
52 second estimating part
60 third input/output unit
61 third machine learning unit
62 third estimating part
And N, network.
Claims (4)
1. A win/loss prediction system is characterized by comprising:
an image acquisition unit that acquires an image of an analysis target in a match;
a first system for estimating a ball control in the game;
a second system for estimating events occurring in said match; and
a third system for predicting a win or loss result of the game based on the detection result of the first system and the detection result of the second system,
the first system includes:
a first machine learning unit that analyzes a relationship between an image of a given scene of a past game and information on positions of players and balls in the given scene of the past game by machine learning;
a first estimation unit configured to estimate and output positions of the player and the ball in the given scene of the analysis target game, based on the relationship analyzed by the first machine learning unit, with the image of the given scene in the analysis target game acquired by the image acquisition unit as an input; and
a first determination unit that determines and outputs a ball control in a predetermined scene of the game to be analyzed based on the positions of the player and the ball output from the first estimation unit,
the second system includes:
a second machine learning unit that analyzes a relationship between an image of a given scene of a past race and information related to an event occurring in the given scene in the past race by machine learning; and
a second estimation unit that estimates and outputs an event occurring in a given scene in the race of the analysis target, based on the relationship analyzed by the second machine learning unit, with the image of the given scene in the race of the analysis target acquired by the image acquisition unit as an input,
the third system includes:
a third machine learning unit that analyzes, by machine learning, a relationship between information on a ball control in a given scene of a past game, an event occurring in the given scene, an elapsed time from the start of the game to the given scene, a point of win or loss at a time point of the given scene, and a result of win or loss of the past game; and
and a third estimating unit that estimates and outputs a win/lose result of the game expected at the time point of the given scene of the game of the analysis target, based on the relationship analyzed by the third machine learning unit, with information on the ball control in the given scene of the game of the analysis target output from the first determining unit, the event occurring at the given scene of the game of the analysis target output from the second estimating unit, the elapsed time from the start of the game of the analysis target to the given scene, and the point of win/lose at the time point of the given scene of the game of the analysis target as input.
2. The win or loss prediction system of claim 1, wherein,
the third system includes an image analysis unit that performs image analysis of a video of a predetermined scene in the race of the analysis target, and acquires information on an elapsed time from a start of the race of the analysis target of the predetermined scene to the predetermined scene and a point of gain or loss at a time point of the predetermined scene in the race of the analysis target.
3. The win or loss prediction system according to claim 1 or 2, wherein,
the third estimating section estimates and outputs a probability of winning, a probability of negation, and a probability of tie, respectively, as a result of winning or negation expected at a time point of a given scene of the race of the analysis target.
4. A win/loss prediction method executed by a win/loss prediction system,
the win/loss prediction system is provided with:
an image acquisition unit that acquires an image of an analysis target in a match;
a first system for estimating a ball control in the game;
a second system for estimating events occurring in said match; and
a third system for predicting a win or loss result of the game based on the detection result of the first system and the detection result of the second system,
performing the following steps in the first system:
a first machine learning step of analyzing, by machine learning, a relationship between an image of a given scene of a past game and information related to positions of players and balls in the given scene of the past game;
a first estimation step of estimating and outputting positions of the player and the ball in the given scene of the analysis target game, based on the relationship analyzed in the first machine learning step, with the image of the given scene in the analysis target game acquired by the image acquisition unit as an input; and
a first determination step of determining and outputting a ball control in a given scene of the game to be analyzed based on the positions of the player and the ball output from the first estimation step,
performing the following steps in the second system:
a second machine learning step of analyzing, by machine learning, a relationship between an image of a given scene of a past game and information related to an event occurring in the given scene in the past game; and
a second estimation step of estimating and outputting an event occurring in a predetermined scene in the race of the analysis target, based on the relationship analyzed in the second machine learning step, with the image of the predetermined scene in the race of the analysis target acquired by the image acquisition unit as an input,
performing, at the third system, the steps of:
a third machine learning step of analyzing, by machine learning, a relationship between information on a ball control in a given scene of a past game, an event occurring at the given scene, an elapsed time from a start of the game to the given scene, a point of win or loss at a point of time of the given scene, and a win or loss result of the past game; and
a third estimation step of estimating and outputting, based on the relationship analyzed in the third machine learning step, a win/lose result of the race expected at the time point of the given scene of the race of the analysis target, with information on the ball control in the given scene of the race of the analysis target output from the first determination step, the event occurring at the given scene of the race of the analysis target output from the second estimation step, the elapsed time from the start of the race of the analysis target to the given scene, and the win/lose point at the time point of the given scene of the race of the analysis target as input.
Applications Claiming Priority (3)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| JP2020-082460 | 2020-05-08 | ||
| JP2020082460A JP7078667B2 (en) | 2020-05-08 | 2020-05-08 | Win / Loss Prediction System |
| PCT/JP2020/027422 WO2021225007A1 (en) | 2020-05-08 | 2020-07-15 | Outcome prediction system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113939837A true CN113939837A (en) | 2022-01-14 |
Family
ID=78409511
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202080002630.6A Pending CN113939837A (en) | 2020-05-08 | 2020-07-15 | Win-loss prediction system |
Country Status (4)
| Country | Link |
|---|---|
| JP (1) | JP7078667B2 (en) |
| KR (1) | KR102443007B1 (en) |
| CN (1) | CN113939837A (en) |
| WO (1) | WO2021225007A1 (en) |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20170223415A1 (en) * | 2014-08-21 | 2017-08-03 | Wook Jeon | Participative sports game system and method by real-time analyzing and predicting |
| KR20180063777A (en) * | 2016-12-02 | 2018-06-12 | (주) 지티씨솔루션 | Soccer information analysis apparatus and the method thereof |
| WO2018122956A1 (en) * | 2016-12-27 | 2018-07-05 | 日本電気株式会社 | Sport motion analysis support system, method and program |
| CN109165253A (en) * | 2018-08-15 | 2019-01-08 | 宁夏大学 | A kind of method and apparatus of Basketball Tactical auxiliary |
| US20190224556A1 (en) * | 2018-01-21 | 2019-07-25 | Stats Llc | Method and System for Interactive, Interpretable, and Improved Match and Player Performance Predictions in Team Sports |
| CN110147524A (en) * | 2019-05-10 | 2019-08-20 | 深圳市腾讯计算机系统有限公司 | A kind of result of the match prediction technique, device and equipment based on machine learning |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2004110826A (en) | 2002-09-16 | 2004-04-08 | Asobous:Kk | Soccer team analysis system, and soccer team analysis/using method |
| KR101844874B1 (en) * | 2017-03-15 | 2018-04-04 | (주)에이피케이어플킹 | System for predicting decision using big data based on data mining techniques and method therefor |
| JP7032948B2 (en) | 2018-02-14 | 2022-03-09 | 成典 田中 | Tactical analyzer |
-
2020
- 2020-05-08 JP JP2020082460A patent/JP7078667B2/en active Active
- 2020-07-15 CN CN202080002630.6A patent/CN113939837A/en active Pending
- 2020-07-15 KR KR1020207032988A patent/KR102443007B1/en active IP Right Grant
- 2020-07-15 WO PCT/JP2020/027422 patent/WO2021225007A1/en active Application Filing
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20170223415A1 (en) * | 2014-08-21 | 2017-08-03 | Wook Jeon | Participative sports game system and method by real-time analyzing and predicting |
| KR20180063777A (en) * | 2016-12-02 | 2018-06-12 | (주) 지티씨솔루션 | Soccer information analysis apparatus and the method thereof |
| WO2018122956A1 (en) * | 2016-12-27 | 2018-07-05 | 日本電気株式会社 | Sport motion analysis support system, method and program |
| US20190224556A1 (en) * | 2018-01-21 | 2019-07-25 | Stats Llc | Method and System for Interactive, Interpretable, and Improved Match and Player Performance Predictions in Team Sports |
| CN109165253A (en) * | 2018-08-15 | 2019-01-08 | 宁夏大学 | A kind of method and apparatus of Basketball Tactical auxiliary |
| CN110147524A (en) * | 2019-05-10 | 2019-08-20 | 深圳市腾讯计算机系统有限公司 | A kind of result of the match prediction technique, device and equipment based on machine learning |
Non-Patent Citations (1)
| Title |
|---|
| LIFANG WU等: "Ontology Based Global and Collective Motion Patterns for Event Classification in Basketball Videos", 《 IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS FOR VIDEO TECHNOLOGY》, 19 March 2019 (2019-03-19), pages 1 - 13 * |
Also Published As
| Publication number | Publication date |
|---|---|
| KR20210136828A (en) | 2021-11-17 |
| JP7078667B2 (en) | 2022-05-31 |
| JP2021177316A (en) | 2021-11-11 |
| WO2021225007A1 (en) | 2021-11-11 |
| KR102443007B1 (en) | 2022-09-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11755952B2 (en) | System and method for predictive sports analytics using body-pose information | |
| KR101938364B1 (en) | Soccer information analysis apparatus and the method thereof | |
| Apostolou et al. | Sports Analytics algorithms for performance prediction | |
| Spearman et al. | Physics-based modeling of pass probabilities in soccer | |
| Erickson et al. | Global state evaluation in StarCraft | |
| Synnaeve et al. | A dataset for StarCraft AI and an example of armies clustering | |
| JP4621910B2 (en) | Dominance level determination device and dominance level determination method | |
| CN105844697A (en) | Data and event statistics implementing method for sports event on-site three-dimensional information | |
| Galli et al. | A cheating detection framework for unreal tournament iii: A machine learning approach | |
| US11908190B2 (en) | Game monitoring | |
| Umemoto et al. | Location analysis of players in UEFA EURO 2020 and 2022 using generalized valuation of defense by estimating probabilities | |
| CN114419526A (en) | Foul behavior detection method and apparatus, electronic device and storage medium | |
| CN113939837A (en) | Win-loss prediction system | |
| US11471779B2 (en) | Spectating support apparatus, spectating support method, and spectating support program | |
| Wu et al. | Evaluation of off-the-ball actions in soccer | |
| EP3968644A1 (en) | Method and data processing system for making predictions during a live event stream | |
| CN112200891A (en) | System and method for reconstructing video three-dimensional animation scene of football match | |
| Röfer et al. | B-Human 2019–complex team play under natural lighting conditions | |
| EP3797400B1 (en) | System and method for automatic detection of referee's decisions in a ball-game | |
| JP7032948B2 (en) | Tactical analyzer | |
| Joo et al. | Learning to automatically spectate games for Esports using object detection mechanism | |
| JP7246649B2 (en) | tactical analyzer | |
| KR101765432B1 (en) | Method for Determining Win-Loss of Game for Operating Process of Game Conference | |
| Louivion et al. | Analysis of Performance Measures That Affect NBA Salaries | |
| Budianto et al. | An analysis on the rush strategies of the real-time strategy game StarCraft-II |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| REG | Reference to a national code |
Ref country code: HK Ref legal event code: DE Ref document number: 40066524 Country of ref document: HK |