WO2024096167A1

WO2024096167A1 - Method for acquiring type of target event within sports session and electronic apparatus performing same

Info

Publication number: WO2024096167A1
Application number: PCT/KR2022/017310
Authority: WO
Inventors: 김주표; 이재찬; 김현성
Original assignee: 주식회사 핏투게더
Priority date: 2022-11-04
Filing date: 2022-11-07
Publication date: 2024-05-10

Abstract

Disclosed is a method for acquiring a type of target event within a sports session. The method comprises the steps of: acquiring a plurality of location datasets respectively corresponding to a plurality of time points within the sports session; acquiring speed data of a plurality of players; identifying a target event on the basis of the speed data and acquiring information regarding the target event; on the basis of time information corresponding to the target event, identifying a plurality of location datasets corresponding to the target event from the plurality of location datasets; and acquiring the type of target event on the basis of the plurality of location datasets that have been identified.

Description

Methods for obtaining types of target events within a sports session and electronic devices for doing so

This disclosure relates to a method for obtaining the type of target event that occurred in a sports session and an electronic device for performing the same.

A sprint event refers to an event in which a sports player runs faster than a preset speed in a sports session (including sports games and training sessions). Sprint events are highly correlated, being associated with approximately 70% of all scoring events that occur in a sports session. The analysis of sprint events is very important in the modern sports industry because the more scoring events there are, the higher the probability of winning the sports session.

The present disclosure relates to a method of providing various information (eg, sprint type) for analysis of a sprint event.

The following presents a summary of specific embodiments disclosed in this disclosure. It should be understood that the aspects presented in the following summary are merely intended to provide a brief summary of specific embodiments and are not intended to limit the scope of the present disclosure. Accordingly, it should be noted in advance that this description may include various aspects not presented below.

This disclosure and the inventive concepts disclosed hereinafter provide methods, devices, systems, and computer-readable storage media for predicting the type of target event that will occur in a sports session.

One technical problem that the present invention seeks to solve is to provide a method for predicting the type of target event based on a team location dataset representing the placement distribution of a plurality of players.

Another technical problem to be solved by the present invention is to provide a method for predicting the type of target event based on ball ownership in the target event section.

Another technical problem to be solved by the present invention is to provide a method of providing a video feed corresponding to a target event to a user.

However, the problem to be solved by this description is not limited to this, and may be expanded in various ways without departing from the spirit and scope of this description.

Embodiments include methods, devices, systems, and computer-readable storage media for predicting the position of a sports ball.

According to one embodiment, in a method of obtaining the type of target event within a sports session, a plurality of location datasets each corresponding to a plurality of time points within the sports session - each location dataset is a corresponding time point. Obtaining -including location values of players; Obtaining motion data of the plurality of players; Identifying the target event based on the motion data, and obtaining information about the target event - the information includes time information corresponding to the target event and information about the player who performed the target event. step; Situation information reflecting a game situation during the target event based on time information corresponding to the target event - the situation information includes a first location dataset reflecting the location of the player who performed the target event and the player Obtaining a second location dataset that reflects the placement or movement of at least one other player other than the player on the playing field; And a method including obtaining the type of the target event based on the situation information may be provided.

According to another embodiment, an electronic device for obtaining a type of target event within a sports session, comprising: a memory storing at least one instruction; and a processor, wherein the processor executes the at least one instruction, thereby generating a plurality of location data sets, each of which corresponds to a plurality of time points within the sports session - each location data set corresponding to a corresponding time point. Including the position values of the players, obtain, obtain motion data of the plurality of players, identify the target event based on the motion data, and information about the target event, wherein the target Acquires - including time information corresponding to an event and information about the player who performed the target event - and situational information reflecting the game situation during the target event based on the time information corresponding to the target event - The situation information includes a first location dataset reflecting the location of the player who performed the target event and a second location dataset reflecting the placement or movement of at least one other player other than the player in the stadium. An electronic device may be provided that acquires a function and acquires the type of the target event based on the context information.

The above exemplary embodiments and other exemplary embodiments will be explained or clarified by the detailed description below regarding the exemplary embodiments to be read in conjunction with the attached drawings.

The disclosed technology can have the following effects. However, since it does not mean that a specific embodiment must include all of the following effects or only the following effects, the scope of rights of the disclosed technology should not be understood as being limited thereby.

According to an embodiment of the present disclosure, the type of target event can be predicted based on a team location dataset representing the placement distribution of a plurality of players.

According to an embodiment of the present disclosure, the type of target event can be predicted based on ball ownership in the target event section.

According to an embodiment of the present disclosure, a video feed corresponding to a target event may be provided to the user.

The above disclosure does not contain a complete list of all aspects of the invention. The present invention should be understood to include all methods, devices, and systems practicable from all suitable combinations of the various aspects disclosed in the foregoing summary as well as the following detailed description and claims. In addition, effects that can be obtained or expected due to the embodiments of the present disclosure will be directly or implicitly disclosed in the detailed description of the embodiments of the present disclosure. For example, various effects expected according to embodiments of the present disclosure will be disclosed in the detailed description to be described later.

Aspects, features and advantages of specific embodiments of the present disclosure will become clearer through the following description with reference to the accompanying drawings.

1 is a diagram illustrating an example target event of a sports session according to an embodiment of the present disclosure.

Figure 2 is a diagram illustrating a player tracking device according to an embodiment of the present disclosure.

Figure 3 is a flowchart showing a method of obtaining the type of target event according to an embodiment of the present disclosure.

Figure 4 is a diagram illustrating a plurality of location data sets according to an embodiment of the present disclosure.

Figure 5 is a diagram showing a team location dataset according to an embodiment of the present disclosure.

Figure 6 is a graph showing speed data of a sprinter according to an embodiment of the present disclosure.

FIG. 7 is a diagram illustrating a method of obtaining the type of target event using a neural network model according to an embodiment of the present disclosure.

Figure 8 is a diagram for explaining a method of acquiring the position of a ball according to an embodiment of the present disclosure.

Figure 9 is a flowchart showing a method of providing a target scene corresponding to a target event according to an embodiment of the present disclosure.

FIG. 10 is a diagram for explaining a method of extracting a target scene from a video feed according to an embodiment of the present disclosure.

Figure 11 is a diagram showing a sprint scene according to an embodiment of the present disclosure.

12 is a table showing a target event log according to an embodiment of the present disclosure.

Figure 13 is a flowchart showing a method of obtaining the type of target event according to an embodiment of the present disclosure.

Figure 14 is a diagram for explaining a method of obtaining ball ownership using a neural network model according to an embodiment of the present disclosure.

Figure 15 is a diagram for explaining the structure of a third neural network model according to an embodiment of the present disclosure.

Figure 16 is a diagram for explaining a method of obtaining the type of target event according to an embodiment of the present disclosure.

Figure 17 is a block diagram showing the configuration of an electronic device according to an embodiment of the present disclosure.

Terms used in this specification will be briefly described, and the present disclosure will be described in detail.

The terms used in the embodiments of the present disclosure have selected general terms that are currently widely used as much as possible while considering the functions in the present disclosure, but this may vary depending on the intention or precedent of a technician working in the art, the emergence of new technology, etc. . In addition, in certain cases, there are terms arbitrarily selected by the applicant, and in this case, the meaning will be described in detail in the description part of the relevant disclosure. Therefore, the terms used in this disclosure should be defined based on the meaning of the term and the overall content of this disclosure, rather than simply the name of the term.

Embodiments of the present disclosure can be modified and have various embodiments, and specific embodiments will be illustrated in the drawings and described in detail in the detailed description. However, this is not intended to limit the scope to specific embodiments, and should be understood to include all transformations, equivalents, and substitutes included in the disclosed spirit and technical scope. In describing the embodiments, if it is determined that detailed description of related known technologies may obscure the point, the detailed description will be omitted.

Terms such as first, second, etc. may be used to describe various components, but the components should not be limited by the terms. Terms are used only to distinguish one component from another.

Singular expressions include plural expressions unless the context clearly dictates otherwise. In this application, terms such as “comprise” or “consist of” are intended to designate the presence of features, numbers, steps, operations, components, parts, or combinations thereof described in the specification, but are intended to indicate the presence of one or more other It should be understood that this does not exclude in advance the presence or addition of features, numbers, steps, operations, components, parts, or combinations thereof.

Below, with reference to the attached drawings, embodiments of the present disclosure will be described in detail so that those skilled in the art can easily practice them. However, the present disclosure may be implemented in many different forms and is not limited to the embodiments described herein. In order to clearly explain the present disclosure in the drawings, parts that are not related to the description are omitted, and similar parts are given similar reference numerals throughout the specification.

The method includes a plurality of team location datasets corresponding to the target event from the plurality of location datasets - the plurality of team location datasets being the stadium of players of the first team to which the player who performed the target event belongs. generating a plurality of first team location datasets reflecting the placement status in the stadium, and a plurality of second team location datasets reflecting the placement status of players of a second team in the stadium; Further comprising, the step of obtaining the type of the target event includes at least one of the plurality of first team location datasets or the plurality of second team location datasets and location values of the player who performed the target event. The type of the target event can be obtained by inputting it into a neural network model.

The method further includes obtaining the position of a ball used in the sports session, wherein the step of obtaining the type of the target event includes determining the position of the ball in a time interval corresponding to the target event. The type of the target event can be obtained by additionally inputting it into the neural network model.

In the step of acquiring the type of the target event, the type of the target event may be obtained by additionally inputting role information of the player who performed the target event into the neural network model.

The time information corresponding to the target event includes a first time period in which the instantaneous speed of the player who performed the target event is maintained within a preset range for a critical time or more, from the time when the target event is intended by the player. It may include a second time section from the start point of the first time section, and a third time section from the end point of the first time section to a preset time after the end point of the first time section.

The threshold time may be adjusted based on the distance the player moves during the first time period.

The timing at which the target event is intended may be determined based on the player's acceleration.

The processor, a plurality of team location datasets from the identified plurality of location datasets - the plurality of team location datasets, the arrangement status of players of the first team to which the player who performed the target event belongs in the stadium generating a plurality of first team location datasets reflecting a plurality of first team location datasets, and a plurality of second team location datasets reflecting placement states of players of a second team in the stadium, wherein the plurality of first team Type of the target event by inputting at least one of the location dataset or the plurality of second team location datasets and the location values of the player who performed the target event included in the identified plurality of location datasets into a neural network model. can be obtained.

The processor may acquire the position of the ball used in the sports session and obtain the type of the target event by additionally inputting the position of the ball in the time interval corresponding to the target event into the neural network model.

The processor may obtain the type of the target event by additionally inputting role information of the player who performed the target event into the neural network model.

The processor may adjust the threshold time based on the distance the player moves during the first time period.

The processor may determine when the target event is intended based on the player's acceleration.

According to another embodiment, a method of providing information about a target event occurring in a sports session includes: acquiring a video feed for the sports session; predicting the type of target event that occurred in the sports session; extracting a target scene corresponding to the target event from the video feed; And a method including the step of storing the type of the predicted target event in association with the target scene.

The method includes providing the type of the predicted target event and the target scene to a user; And when a user command for modifying the type of the predicted target event is received, the method may further include modifying the type of the predicted target event based on the user command.

The type of the predicted target event may be displayed in the target scene.

A highlight may be displayed on the player who performed the target event in the target scene, and the player's movement trajectory may be displayed in a time section corresponding to the target event.

The method further includes obtaining a plurality of location datasets, each of which corresponds to a plurality of time points within a sports session, each location dataset including location values of players at the corresponding time points, , the predicting step is to identify a plurality of location datasets corresponding to the target event from the plurality of location datasets, and a neural network model trained to receive the plurality of location datasets and predict the type of the target event. Using , the type of the target event can be obtained based on the identified plurality of location datasets.

The target scene may be extracted from the video feed based on time information corresponding to the target event.

According to another embodiment, an electronic device providing information about a target event occurring in a sports session, comprising: a memory storing at least one instruction; and a processor, wherein the processor obtains a video feed for the sports session by executing the at least one instruction, predicts the type of the target event that occurred in the sports session, and executes the at least one instruction from the video feed. An electronic device may be provided including extracting a target scene corresponding to a target event, and storing the target scene in association with the type of the predicted target event.

The processor provides the type of the predicted target event and the target scene to the user, and when a user command for modifying the type of the predicted target event is received, the predicted target event is changed based on the user command. The type can be modified.

The processor acquires a plurality of location datasets, each of which corresponds to a plurality of time points within a sports session, each location dataset including location values of players at the corresponding time points, and the plurality of locations. Identifying a plurality of location datasets corresponding to the target event from a dataset, receiving the plurality of location datasets and using a neural network model learned to predict the type of the target event, the identified plurality of location datasets Based on this, the type of the target event can be obtained.

According to another embodiment, in a method of obtaining the type of target event in a sports session, the target event is identified based on the speed of a plurality of players participating in the sports session, and information about the identified target event is - Obtaining the information, including time information corresponding to the target event and information about a target player who performed the target event; Obtaining a plurality of location data sets, each of which corresponds to a plurality of time points included in the time interval of the identified target event, each location data set including position values of the plurality of players at the corresponding time point; determining ball ownership in a time interval of the identified target event based on the plurality of location data sets; and obtaining the type of the target event based on the plurality of location datasets and the determined ball ownership.

In the step of acquiring the type of the target event, when the first team to which the target player belongs has possession of the ball in the time section of the identified target event, the method is learned to obtain the type of target event corresponding to the attack situation. Using a first neural network model, based on the position value of the target player, a first team position dataset related to the first team, and a second team location dataset associated with a second team that is an opponent of the first team The type of the target event can be obtained.

The step of acquiring the type of the target event includes, when a second team that is an opponent of the first team to which the target player belongs has possession of the ball in the time section of the identified target event, a target event corresponding to a defense situation The type of the target event can be obtained based on the location value of the target player and the second team location dataset related to the second team, using a second neural network model learned to obtain the type.

The step of obtaining the type of the target event includes, when the first team to which the target player belongs has possession of the ball in the time section of the identified target event, the location of the ball in the time section of the identified target event. The reflecting ball position dataset may be input to the first neural network model.

The method includes a plurality of team location datasets from the plurality of location datasets, wherein the plurality of team location datasets are generated from location datasets corresponding to players of a first team and a stadium of players of the first team. A first team location dataset reflecting the placement status in the stadium, and a second team location data generated from a location dataset corresponding to the players of the second team and reflecting the placement status of the players of the second team in the stadium. It further includes the step of generating a set, wherein the step of obtaining the type of the target event includes: a location value of the target player, using a third neural network model learned to obtain the type of the target event; The type of the target event may be obtained based on a first team location dataset related to the first team and a second team location dataset related to the second team that is the opposing team of the first team.

The ball ownership is to determine the ball ownership by receiving a plurality of team position data sets generated from the plurality of position data sets and a ball position data set reflecting the position of the ball in the time interval of the identified target event. It can be obtained using the learned fourth neural network model.

The fourth neural network model may be a neural network model of the recurrent neural network series that receives the plurality of team location data sets in segments corresponding to a predetermined time length.

According to another embodiment, an electronic device for obtaining a type of target event within a sports session, comprising: a memory storing at least one instruction; and a processor; wherein the processor, by executing the at least one instruction, identifies a target event based on speeds of a plurality of players participating in the sports session, and provides information about the identified target event - the information. includes time information corresponding to the target event and information about the target player who performed the target event, and obtains a plurality of time points each corresponding to a plurality of time points included in the time section of the identified target event. Obtain location datasets, each location dataset comprising location values of the plurality of players at a corresponding point in time, and ball possession in a time interval of the identified target event based on the plurality of location datasets. An electronic device may be provided including the step of determining and obtaining the type of the target event based on the plurality of location datasets and the determined ball ownership.

The processor uses a first neural network model learned to obtain the type of target event corresponding to an attack situation when the first team to which the target player belongs has possession of the ball in the time interval of the identified target event. , Obtaining the type of the target event based on the location value of the target player, a first team location dataset related to the first team, and a second team location dataset related to the second team that is the opposing team of the first team. can do.

The processor is trained to obtain the type of target event corresponding to the defensive situation when the second team, which is an opponent of the first team to which the target player belongs, has possession of the ball in the time section of the identified target event. Using a second neural network model, the type of the target event can be obtained based on the location value of the target player and a second team location dataset related to the second team.

The processor generates a ball position dataset reflecting the position of the ball in the time interval of the identified target event when the first team to which the target player belongs has possession of the ball in the time interval of the identified target event. It can be input into the first neural network model.

According to another embodiment, a computer-readable recording medium on which a program for executing the method on a computer is recorded may be provided.

Referring to FIG. 1, in a soccer game, a player 1 may run with the ball 2 at a preset speed (eg, 25 km/h) or higher. At this time, the target event is a sprint event, and the type of target event may be ‘Run with ball’. The electronic device 100 of the present disclosure can identify a target event within a sports session and obtain the type of the target event. For example, when an event as shown in FIG. 1 occurs in a sports session, the electronic device 100 may determine that a ‘hit and run’ event has been performed by the player 1.

Meanwhile, for convenience of explanation, the following description will focus on the case where the target event is a sprint event, but the target event of the present disclosure may be one of events other than the sprint event. For example, the target event could be explosive acceleration or change of direction (COD). Additionally, the target event may be High Speed Running (HSR) or a set piece. High-speed running refers to running at a lower speed than sprinting.

The various methods described in this disclosure are not limited to soccer and can be applied to all ball sports, including hockey and ice hockey. And, in this disclosure, a player performing a target event may be referred to as a target player. In particular, when the target event is a sprint event, the target player may be referred to as a sprinter.

Referring to Figure 2, the player tracking device (T) may be inserted into a pocket (4) formed in clothing (3) worn by the player (1). The player tracking device T may include at least one sensor. For example, the player tracking device T may include at least one of a GPS sensor, an IMU sensor, or an LPS sensor. The player tracking device T can measure data related to the player 1 during a sports session and transmit the measured data to a server. For example, data related to player 1 may include position data, velocity data, and acceleration data of player 1 during a sports session. Meanwhile, the player tracking device T may be installed on another item worn by the player 1 rather than the pocket 4. For example, the player tracking device T may be provided on a shoe, bracelet or hair band worn by the player 1.

Each of a plurality of players participating in a sports session may wear a player tracking device (T). Accordingly, a plurality of location data sets corresponding to each of a plurality of players can be obtained. Meanwhile, some of the plurality of players participating in the sports session may not be wearing the player tracking device (T). For example, a goalie or goaltender may not wear a player tracking device (T).

In the present disclosure, the electronic device 100 may identify a sprint event using data related to the player 1 and determine the type of the identified sprint event. For example, the electronic device 100 may identify a sprint event using the player's speed data. Additionally, the electronic device 100 may obtain the type of sprint event using the player's location data. At least some of the various data used by the electronic device 100 may be measured by the player tracking device (T). The electronic device 100 may receive data related to the player 1 from a server or a player tracking device T.

Meanwhile, the location data of the present disclosure may be data processed by the electronic device 100 rather than raw data measured by various sensors. For example, the electronic device 100 may obtain location data by performing signal processing on a signal measured by a GPS sensor and a signal measured by an IMU sensor. Additionally, the electronic device 100 may convert the position value measured by the GPS sensor to correspond to the coordinate system of the stadium where the sports session is taking place.

Referring to FIG. 3, the electronic device 100 may acquire a plurality of location datasets each corresponding to a plurality of time points within a sports session (S310). Each location dataset may include location values of players at a corresponding point in time. For example, a plurality of location datasets may be as shown in FIG. 4. Figure 4 is a diagram illustrating a plurality of location data sets according to an embodiment of the present disclosure. Referring to FIG. 4, the plurality of location datasets 40 may include a first location dataset corresponding to a first time point (t=1). The first location data set is a plurality of position values ((x1_A1, y1_A1), (x1_A2, y1_A2), ..., (x1_B11, y1_B11)) respectively representing the positions of a plurality of players at a first time point (t=1). may include.

The electronic device 100 may acquire motion data of a plurality of players (S320). Motion data may include speed data, acceleration data, and angular velocity data. Motion data may represent various actions of the player. For example, motion data may represent shooting, jumping, changing direction, etc. Motion data of multiple players may be measured by multiple player tracking devices (T) worn by multiple players. For example, a player's speed data may be measured by a GPS sensor included in the tracking device (T). Additionally, the player's acceleration data and angular velocity data may be measured by an IMU sensor included in the tracking device (T).

The electronic device 100 may receive speed data of a plurality of players from a server or a player tracking device (T). Alternatively, the electronic device 100 may acquire speed data based on position data or acceleration data measured by the player tracking device T. For example, the electronic device 100 may obtain speed data by differentiating position data or obtain speed data by integrating acceleration data. Additionally, the electronic device 100 may acquire speed data measured by an external device. For example, an external device may perform video analysis on a video feed of a sports session to obtain speed data.

The electronic device 100 may identify a target event based on the acquired motion data and obtain information about the target event (S330). For example, if the speed of the first player is maintained within a preset range (e.g., 25 (km/h) to 50 (km/h)) for more than a threshold time (e.g., 3 seconds), the electronic device 100 1 It can be determined that the sprint event has been performed by the player.

Information about the target event may include time information corresponding to the target event and information about the target player who performed the target event. Additionally, information about the target event may include ball possession information or the distance traveled by the target player during the time interval corresponding to the target event. Meanwhile, in the present disclosure, the time section corresponding to the target event may be briefly referred to as the target event section. In particular, when the target event is a sprint event, the time section corresponding to the sprint event may be referred to as a sprint section.

The electronic device 100 may obtain the length of the target event section based on speed data of the target player who performed the identified target event. The target event section may include a first time section in which the instantaneous speed of the target player is maintained within a preset range for more than a threshold time. The electronic device 100 may identify the first point in time when the target player's instantaneous speed enters a preset range as the starting point of the first time period. The electronic device 100 may identify a second point in time at which the target player's instantaneous speed deviates from the preset range after the first point in time as the end point of the first time period.

Meanwhile, the critical time for identifying the target event may vary depending on the distance the target player moves while performing the target event. For example, the greater the distance the target player moves, the shorter the threshold time may be. As the distance traveled by the target player decreases, the critical time may increase. Meanwhile, the moving distance of the target player in the target event section may be calculated based on the speed data of the target player in the target event section and the time length of the target event section.

Meanwhile, the target event section of the present disclosure may be larger than the time section in which the target player actually performs the target event. For example, the sprint section of the present disclosure may be larger than the section where the target player actually sprints. This is because the type of target event can be obtained more accurately by using information before and after the time when the actual target event was performed.

The target event section may include a second time section from the time when the target event is intended by the target player to the starting point of the first time section (ie, the first time point). The electronic device 100 may determine when a target event is intended (or when an intention occurs) based on the acceleration of the target player. For example, the electronic device 100 may determine the time when the target player's acceleration becomes greater than a preset value as the time when the intent for the target event occurs.

The target event section may include a third time section from the end point of the first time section (i.e., the second time point) to the third time point that is a preset time (e.g., 2 seconds) after the end point of the first time section. .

Meanwhile, an identification value may be assigned to each player tracking device (T) worn by each player. The electronic device 100 may identify the target player based on speed data measured by the player tracking device (T) and an identification value assigned to the player tracking device (T).

The electronic device 100 may acquire situation information reflecting the game situation during the target event based on time information corresponding to the target event (S340). The situation information may include a first location dataset reflecting the location of the target player and a second location dataset reflecting the placement or movement of at least one other player other than the target player in the stadium. In addition to this, the situational information may include context information that reflects various tactical situations within a sports session.

The situation information may include a plurality of team location datasets corresponding to the target event. The electronic device 100 may generate a plurality of team location datasets from a plurality of location datasets corresponding to the target event section. Here, a plurality of team location datasets may each correspond to a plurality of time points within the target event section. The team location dataset includes a first team location dataset reflecting the disposition status of players of the first team on the stadium to which the player who performed the target event belongs, and a first team location dataset reflecting the disposition status of players of the second team opposing the first team on the stadium. It may include a second team location dataset that reflects the deployment status of. As an example, the team location dataset may be as shown in Figure 5.

Figure 5 is a diagram showing a team location dataset according to an embodiment of the present disclosure. Specifically, the first team location dataset 51 may reflect the arrangement status of players of the first team in the stadium at a point in time within the target event section. The second team location data set 52 may reflect the arrangement status of players of the second team in the stadium at a point in time within the target event section.

Each element of the team location datasets 51 and 52 may correspond to a location within the stadium and may indicate whether a player is located at each location. If the value of each element is 0, it may mean that there is no player in the location corresponding to the element. If the value of each element is 1, this may mean that a player belonging to the first team (Team A) is located in the position corresponding to the element. Of course, here, the values "0" or "1" are just arbitrary indices, so you can express them as "-1" and "1", or use any other form of value that can semantically distinguish between teams. Meanwhile, in Figure 5, the position is expressed using a coordinate plane in the form of a grid map with the field expressed as 120x90, but the size of the grid map is not necessarily limited to the example in Figure 5, and the position is expressed with vector coordinates instead of grid coordinates. It is also possible.

Meanwhile, the team location dataset may be a merged team location dataset of two teams. For example, if an element of the team location dataset is "0", there may be no player in the location corresponding to that element. If it is “1” in the team location dataset, a player belonging to the first team may exist in the location corresponding to the element. If it is "2" in the team location dataset, a player belonging to the second team may exist in the location corresponding to the element. Here again, it should be noted that the index number is arbitrary.

As described above, the electronic device 100 may process location data in the form of a heatmap as shown in FIG. 5 to obtain permutation invariant data. Alternatively, the electronic device 100 may acquire a team location dataset using a neural network model. A neural network model can convert input data into permutation invariant data. For example, a neural network model may be a 'set transformer'. Meanwhile, this neural network model may be included inside a neural network model learned to infer the type of target event. For example, the first neural network model (M1) may include a 'set transformer' that is trained to receive a plurality of location datasets and generate a team location dataset.

Referring again to FIG. 3, the electronic device 100 may obtain the type of target event based on context information (S350). The electronic device 100 includes at least one of a plurality of first team location datasets or a plurality of second team location datasets corresponding to the target event section in the first neural network model (M1), and a target player corresponding to the target event section. You can obtain the type of target event by entering the location data set. Here, the target player's location data set refers to a set of location values of the target player in the target event section.

Meanwhile, the input data of the first neural network model (M1) may vary depending on ball ownership in the target event section. For example, when the first team belonging to the target player has possession of the ball in the target event section, the electronic device 100 stores the target player's location dataset, the first team's team location dataset, and the target player's team location dataset against the first team. The team location dataset of the second team may be input into the first neural network model (M1). When the second team has possession of the ball in the target event section, the electronic device 100 may input the target player's location dataset and the second team's team location dataset into the first neural network model (M1).

In order to improve the accuracy of predicting the target event type using the first neural network model (M1), additional information may be input to the first neural network model (M1). For example, the electronic device 100 may input the ball position dataset corresponding to the target event section or the role information of the target player into the first neural network model M1. The target player's role information may mean the target player's position (eg, attacker). Additionally, the electronic device 100 may input identification information of the first team to which the target player belongs and/or identification information of the second team opposing the first team to the first neural network model (M1).

Meanwhile, the first neural network model (M1) will be described in more detail later with reference to FIG. 7.

Referring to FIG. 6, the sprint section 60 may include a first time section 61, a second time section 62, and a third time section 63. The first time section 61 may be a time section in which the instantaneous speed of the sprinter is maintained above the critical speed (sth) for more than a critical time. The second time section 62 may be a time section from the point in time (t1) when the sprinter's intention to sprint occurs to the starting point (t2) of the first time section 61. The third time section 63 may be a time section from a point in time (t3) when the instantaneous speed of the sprinter becomes less than the critical time to a time point (t4) a preset time after the time point (t3).

As described above, the first time section 61 may be defined as a time section in which the instantaneous speed of the sprinter is maintained within a preset range for a critical time. Here, the preset range may be defined by a lower limit threshold and an upper limit threshold. The lower threshold may be the threshold speed (sth).

Meanwhile, speed data acquired by the electronic device 100 may include noise. For example, the instantaneous speed of the sprinter may momentarily become less than the critical speed (sth) for a very short time (for example, 0.1 second) in the first time section 61 and then become greater than the critical speed (sth) again. At this time, the electronic device 100 may judge the sudden decrease in the sprinter's instantaneous speed as noise and ignore it. That is, even though the sprinter's instantaneous speed becomes smaller than the critical speed (sth), the electronic device 100 can determine the entire first time section 61 as one sprint section.

Referring to FIG. 7, the electronic device 100 may select one of the target player's location dataset 71, the first team location dataset 72, the second team location dataset 73, or the ball location dataset 74. The type 75 of the target event can be obtained by inputting at least one into the first neural network model (M1). The first neural network model (M1) is based on at least one of the target player's position dataset 71, the first team position dataset 72, the second team position dataset 73, and the ball position dataset 74. It may be a model learned to obtain the type 75 of the target event.

According to the first embodiment, the electronic device 100 inputs the target player's location dataset 71 and the plurality of second team location datasets 73 corresponding to the target event section into the first neural network model (M1). can do. Each second team location dataset 73 refers to a team location dataset of the second team opposing the first team to which the target player belongs at a corresponding point in time within the target event section.

According to the second embodiment, the electronic device 100 includes a target player's location dataset 71 corresponding to the target event section, a plurality of first team location datasets 72, and a plurality of second team location datasets ( 73) can be input into the first neural network model (M1). Each first team location dataset 72 refers to a team location dataset of the first team to which the target player belongs at a corresponding point in time within the target event section.

According to the third embodiment, the electronic device 100 includes a target player's location dataset 71 corresponding to the target event section, a plurality of first team location datasets 72, and a plurality of second team location datasets ( 73) and the ball position dataset 74 can be input to the first neural network model (M1).

Meanwhile, the first neural network model (M1) processes the input data and can infer the type of target event (75) using only part of the input data based on the processing results. For example, the first neural network model (M1) may include a ball ownership determination module that determines ball ownership and an event type prediction module. The ball ownership determination module may determine ball ownership based on a plurality of first team location datasets 72, a plurality of second team location datasets 73, and a ball location dataset 74. The event type prediction module may predict the type of target event based on ball ownership determined by the ball ownership determination module. For example, when the first team to which the target player belongs owns the ball in the target event section, the event type prediction module includes the target player's location dataset 71, a plurality of first team location datasets 72, and a plurality of The type of target event can be predicted based on the second team location dataset 73. When the second team possesses the ball in the target event section, the event type prediction module can predict the type of target event based on the target player's location dataset 71 and the plurality of second team location datasets 73. there is.

Meanwhile, in FIG. 7, the ball position dataset 74 is shown as being input to the first neural network model (M1), but according to another embodiment, the ball position dataset 74 is obtained inside the first neural network model (M1). It can be. For example, the first neural network model (M1) may include a ball position prediction module trained to predict the ball position based on a plurality of team position datasets.

Referring to FIG. 8, the electronic device 100 may obtain the ball position 82 by inputting the team position dataset 81 into the second neural network model (M2). Specifically, the electronic device 100 may input a plurality of team location datasets 81, each corresponding to a plurality of viewpoints within the target event section, into the second neural network model (M2). The second neural network model (M2) can be trained to predict the position of the ball 82 by receiving a plurality of team position datasets 81 corresponding to a plurality of viewpoints in segments of a predetermined time length. The ball position 82 predicted by the second neural network model (M2) can be used as input data of the first neural network model (M1). Alternatively, the second neural network model (M2) may be embedded within the first neural network model (M1).

Referring to FIG. 9, the electronic device 100 may obtain a video feed for a sports session (S910). The video feed may be footage of all or part of a sports session. For example, the video feed may be a broadcast of a Real Madrid vs. Barcelona match. The electronic device 100 may receive a video feed from an external device or external server. The electronic device 100 may store the received video feed in the memory 120 .

The electronic device 100 may predict the type of target event that occurred in the sports session (S920). For example, the target event may be a sprint event. At this time, the electronic device 100 can predict the type of sprint event. The method for predicting the type of target event has been described above with reference to FIG. 3, so detailed description thereof will be omitted.

The electronic device 100 may extract a target scene corresponding to the target event from the video feed (S930). For example, the electronic device 100 may extract a scene of 'Lionel Messi' sprinting from a video feed. The electronic device 100 may identify the target scene based on time information (eg, time stamp) corresponding to the target event.

The electronic device 100 may store the type of predicted target event in relation to the extracted target scene (S940). The electronic device 100 may generate a target event log by matching the type of target event and the target scene based on the time stamp corresponding to the target event. The electronic device 100 may store the target event log in the memory 120 .

The target event log may include various information related to the target event in addition to the type of target event. For example, the target event log may include data related to identification information, affiliated team, and position of the target player who performed the target event. When the target event is a sprint event, the target event log may include data related to the distance the target player sprints. Additionally, the target event log may include data regarding the results of the target event. The result of a target event can mean success or failure. The target event log may include a link for executing an application or web page that outputs the target scene.

The electronic device 100 may provide the type of predicted target event and the extracted target scene to the user (S950). The electronic device 100 may transmit a target event log in which information about the target event, including the type of the target event, and a link for playing the target scene are recorded, to the user terminal. The user can watch the target scene by selecting a link to play the target scene.

The type of predicted target event may be displayed in the target scene provided to the user. Additionally, a visual indicator regarding the player who performed the target event in the target scene may be displayed. For example, a highlight may be displayed on the target player in the target scene, or the target player's movement trajectory may be displayed.

Referring to FIG. 10, the video feed (VF) may include a plurality of frames including a first frame 101, a third frame 103, and a fifth frame 105. For example, the video feed (VF) is a video of a soccer game, and the target scene to be extracted may be a sprint scene corresponding to a sprint event.

The electronic device 100 may identify a sprint event. For example, the electronic device 100 may identify a sprint event based on speed data measured by a player tracking device (T) worn by players included in the video feed (VF).

The electronic device 100 may define a sprint section corresponding to a sprint event. For example, the sprint section is a first time section in which the instantaneous speed of the sprinter is maintained within a preset range for more than a critical time, a second time section from the point when the sprinter intends to sprint to the starting point of the first time section, and a second time section. It may include a third time interval from the end point of the 1-hour interval to a preset time after the end point of the first time interval.

The electronic device 100 may extract the sprint scene 106 from the video feed (VF) based on the start and end points of the sprint section. For example, the sprint scene 106 may include a second frame 102, a third frame 103, and a fourth frame 104.

Referring to FIG. 11, in a sprint scene, a highlight may be displayed around the sprinting player 1 (i.e., the sprinter) to emphasize the sprinter 1. For example, a bounding box 111 may be displayed around the sprinter 1. Additionally, text 115 indicating the sprint type may be displayed in the sprint scene.

In the sprint scene, an indicator 112 indicating the movement trajectory of the sprinter 1 may be displayed. The indicator 112 may include a first indicator 113 and a second indicator 114. The first indicator 113 and the second indicator 114 may have different visual characteristics. For example, the first indicator 113 and the second indicator 114 may have different thickness, shape, color, design, or pattern.

The area where the first indicator 113 and the second indicator 114 are displayed may be determined according to the speed of the sprinter 1. For example, the first indicator 113 may be displayed in an area where the sprinter 1 moves at a speed within a preset range. On the other hand, the second indicator 114 may be displayed in an area where the sprinter 1 moves at a speed outside the preset range.

Meanwhile, the visual characteristics of the displayed indicator may vary depending on the sprint type. Sprint types can be broadly divided into offensive sprints and defensive sprints. An offensive sprint refers to a sprint performed when the sprinter's team is in an offensive situation (i.e., the sprinter's team has possession of the ball). A defensive sprint refers to a sprint performed when the sprinter's team is in a defensive situation (i.e., the opponent of the sprinter's team has possession of the ball). For example, when the sprint type is an offensive sprint, a first indicator may be displayed, and when the sprint type is a defensive sprint, a second indicator may be displayed. At this time, the visual characteristics of the first indicator and the visual characteristics of the second indicator may be different. Of course, even within an offensive sprint or defensive sprint, the visual characteristics of the indicator may vary depending on the detailed type.

Meanwhile, the electronic device 100 may perform image processing on the sprint scene to display the various indicators described above.

12 is a table showing a target event log according to an embodiment of the present disclosure. Specifically, Figure 12 shows the target event log 1200 when the sports session is a soccer game and the target event is a sprint event.

Referring to FIG. 12, the target event log 1200 may include information about a soccer game, information about a sprinter, and information about a sprint event.

Information about the soccer game may include the date of the soccer game (1201) and identification information (1202) of the opposing team of the team to which the sprinter belongs. Information about the sprinter may include the sprinter's name (1203), position (1204), and identification information of the affiliated team.

Information about the sprint event may include the distance the sprinter runs in each sprint event (1206) and the ball ownership (1207) in the sprint event section. If the sprinter's team possessed the ball, it was recorded as In-possession (abbreviated In-pos), and if the sprinter's team's opponent possessed the ball, it was recorded as Out-possession (abbreviated Out-pos).

Information about the sprint event may include information about the start point (1208) and end point (1209) of the sprint section, the type of sprint event (1210), the result of the sprint event (1211), and the video feed corresponding to the sprint section. there is. Information about the video feed may include a link 1212 for running the video feed.

An example of the type 1210 of the sprint event is 'Run with ball'. 'Run with ball' refers to the movement of running with a ball attached. Another example of the type of sprint event (1210) is 'Covering'. 'Covering' refers to the movement of a sprinter positioning himself towards the goal side to prevent an opposing player who has or will have the ball from heading towards the goal side. Another example of the type of sprint event (1210) is 'Break into Box'. 'Break into Box' refers to a movement in which a teammate on the side has the ball and heads toward the opponent's goal in the opponent's penalty box, aiming to score by receiving a cross from the side.

The electronic device 100 may provide the target event log 1200 to the user. A user may modify at least part of the target event log 1200. For example, a user can modify the type 1210 of a sprint event. As described above, the type of sprint event 1210 may be predicted by the first neural network model M1 and may therefore include some error. The user can run the video PD through the link 1212 to determine whether the type 1210 of the sprint event predicted by the first neural network model (M1) is accurate.

When a user inputs a command to modify the type 1210 of a sprint event into the user terminal, the command entered by the user may be transmitted to the electronic device 100 through a communication interface. At this time, the electronic device 100 may modify and update the type 1210 of the sprint event based on the command input by the user. The electronic device 100 may store the target event log 1200 in which the type 1210 of the sprint event has been modified in the memory 120 .

Meanwhile, FIG. 12 illustrates an embodiment in which the electronic device 100 provides the target event log 1200 to the user, and the user modifies the type 1210 of the sprint event included in the target event log 1200. However, this is only an example, and the electronic device 100 can provide the type of target event to the user in any form, and the user cannot necessarily modify the type of target event only through the target event log. reveal.

Referring to FIG. 13, the electronic device 100 may identify a target event based on the speed of a plurality of players participating in a sports session and obtain information about the identified target event (S1310). Information about the identified target event may include time information corresponding to the target event and information about the target player who performed the target event.

The electronic device 100 may acquire a plurality of location data sets each corresponding to a plurality of time points included in the time interval of the target event (S1320). Each location dataset may include a plurality of location values each assigned to a plurality of player identifiers at a corresponding point in time.

The electronic device 100 may generate a plurality of team location datasets from a plurality of location datasets (S1330). The plurality of team location datasets may include a plurality of first team location datasets that are generated from location values assigned to player identifiers corresponding to the first team and reflect the placement status of players of the first team in the stadium. there is. The plurality of team location datasets may include a plurality of second team location datasets that are generated from location values assigned to player identifiers corresponding to the second team and reflect the placement status of players of the second team in the stadium. there is.

The electronic device 100 may determine ball ownership in the time section of the target event based on a plurality of team location data sets (S1340). Ball ownership in the time interval of the target event (or target event interval) can be determined using a third neural network model (M3) learned to predict ball ownership based on multiple team position datasets and ball position datasets. You can. The electronic device 100 may determine ball ownership in the target event section by inputting a plurality of team location datasets and ball location datasets corresponding to the target event section into the third neural network model (M3).

Meanwhile, in this disclosure, ball ownership can be said to generally mean the team's ball ownership, unless specified as a player's ball ownership. When a team “has the ball,” it means that a player on that team is in control of the ball. Meanwhile, the output of the third neural network model (M3) may be probability values indicating the probability that the two teams will possess the ball. For example, if the probability that the first team possesses the ball is 80% and the probability that the second team possesses the ball is indicated is 20%, the electronic device 100 may determine that the first team possesses the ball. .

Meanwhile, the third neural network model (M3) may be a neural network model of the recurrent neural network series that receives a plurality of team location data sets in segments corresponding to a predetermined time length. A more detailed description of the third neural network model (M3) will be described later with reference to FIG. 15.

The electronic device 100 may obtain the type of target event based on a plurality of team location datasets and ball ownership (S1350). When the first team to which the target player belongs has possession of the ball in the target event section, the electronic device 100 determines the type of target event using an attack event classification model learned to obtain the type of target event corresponding to the attack situation. It can be obtained. Specifically, the electronic device 100 may input the target player's location values corresponding to the target event section, a plurality of first team location datasets, and a plurality of second team location datasets into the attack event classification model. Here, the first team location dataset refers to a team location dataset related to the first team to which the target player belongs. The second team location dataset refers to a team location dataset related to the second team, which is the opposing team of the first team.

When the second team has possession of the ball in the time section of the target event, the electronic device 100 may obtain the type of the target event using a defense event classification model learned to obtain the type of the target event corresponding to the defense situation. You can. Specifically, the electronic device 100 may input the target player's location value and a plurality of second team location datasets into a defense event classification model.

In this way, the neural network model for obtaining the type of target event can be implemented as a dual network consisting of an attack event classification model and a defense event classification model. In this case, the type of neural network model used to obtain the type of event and the type of data input to each neural network model may vary depending on ball ownership. The electronic device 100 may select an optimal neural network model based on ball ownership and obtain the type of target event using the selected neural network model. However, this is only an example, and the neural network model for obtaining the type of target event may be implemented as a model that can classify both target events corresponding to attack situations and target events corresponding to defense situations. .

Meanwhile, the electronic device 100 may additionally input the ball position dataset into the neural network model to obtain more accurate results. For example, the electronic device 100 may additionally input a ball position dataset reflecting the ball position in the target event section into the attack event classification model and/or the defense event classification model.

Referring to FIG. 14, the electronic device 100 inputs the first team location dataset 141, the second team location dataset 142, and the ball location dataset 143 into the third neural network model (M3). Ball possession (144) can be obtained. Specifically, the electronic device 100 uses a plurality of first team location datasets 141 and a plurality of second team location datasets 142 corresponding to a specific time interval (e.g., target event interval) to form a third neural network model. It can be entered in (M3). Additionally, the ball position data set 143 may be a plurality of position values indicating the position of the ball in a specific time period.

The output of the third neural network model (M3) may be probability values representing the probability that each team possesses the ball. The output of the third neural network model (M3) may include identification information of the team in possession of the ball.

Referring to FIG. 15, the third neural network model (M3) may include a first neural network (NN1), a second neural network (NN2), and a third neural network (NN3). The first neural network (NN1) generates a plurality of first feature vectors (152-1, 152-2) based on input data (151-1, 151-2, ..., 151-100) respectively corresponding to a plurality of frames. , ..., 152-100) can be output. Input data corresponding to each frame may include team position datasets and ball position data of multiple teams. For example, the first neural network NN1 may output the first feature vector 152-1 corresponding to the first frame based on the first input data 151-1 corresponding to the first frame. The first neural network (NN1) may be a convolutional neural network (CNN).

Meanwhile, in FIG. 15, for convenience of explanation, the input data (151-1, 151-2, ..., 151-100) is shown as being input to the third neural network model (M3) on a frame basis, but this is only an embodiment. In addition, the input data (151-1, 151-2, ..., 151-100) may be input to the third neural network model (M3) in segments of a predetermined time length. At this time, some of the input data for each time section may overlap.

The second neural network NN2 generates a plurality of second feature vectors 153-1 based on a plurality of first feature vectors 152-1, 152-2, ..., 152-100 respectively corresponding to a plurality of frames. , 153-2, ..., 153-100) can be output. For example, the second neural network NN2 may output a second feature vector 153-1 corresponding to the first frame based on the first feature vector 152-1 corresponding to the first frame. The second neural network (NN2) may be a variety of recurrent neural network-based networks, including a bidirectional Long Short-Term Memory (LSTM) network including a forward network and a reverse network.

The third neural network (NN3) provides the team's ball possession information ( 154-1, 154-2, ..., 154-100) can be output. The team's ball possession information may mean the probability of each team owning the ball. For example, the third neural network NN3 may output the team's ball possession information 154-1 in the first frame based on the second feature vector 153-1 corresponding to the first frame. The third neural network (NN3) may be a deep neural network (DNN).

Referring to FIG. 16, the electronic device 100 may acquire a plurality of team location datasets 161 and ball location datasets 162. The electronic device 100 may acquire ball ownership in the target event section based on a plurality of team location datasets 161 and ball location datasets 162.

When the first team to which the target player who performed the target event belongs possesses the ball in the target event section, the electronic device 100 includes a plurality of team location datasets 161 corresponding to the target event section and the target player's location data. The type 166 of the attack event can be obtained by inputting the set 163 into the attack event classification model 164. For example, if the target event is a sprint event, the attack event may mean an attack sprint. Example types of offensive sprints include 'Run with ball', which is a movement in which a player runs with the ball, and 'Penetrate', which is a movement that aims to score by receiving a through pass while running toward the space behind the opponent's defense and heading directly to the vicinity of the opponent's goal. You can. Meanwhile, the electronic device 100 may additionally input the ball position dataset 162 into the attack event classification model 164.

When the second team, which is the opposing team of the first team, possesses the ball in the target event section, the electronic device 100 includes a plurality of team location datasets 161 corresponding to the target event section and the target player's location dataset ( 163) can be input into the defense event classification model 165 to obtain the type 166 of the defense event. For example, if the target event is a sprint event, the defense event may mean a defense sprint. Exemplary types of defensive sprints include 'Pressing', which is a movement to get close to an opposing player to prevent the opposing player with the ball from advancing the ball by dribbling or passing, and 'Pressing', which is a movement to steal the ball from an opposing player who is moving while in possession of the ball. There may be 'Chase the Opponent with Ball', etc. Meanwhile, the electronic device 100 may additionally input the ball position dataset 162 into the defense event classification model 165.

The attack event classification model 164 and the defense event classification model 165 may be implemented as embedded within the first neural network model (M1). In this case, the first neural network model (M1) determines ball ownership based on a plurality of team position datasets 161 and ball position datasets 162, and inputs input data into the classification model corresponding to ball ownership. You can.

Referring to FIG. 17 , the electronic device 100 may include a communication interface 110, a memory 120, and a processor 130. For example, the electronic device 100 may be a server. However, this is only an example, and the electronic device 100 may be a personal computer.

The communication interface 110 includes at least one communication circuit and can communicate with various types of external devices or external servers. For example, the communication interface 110 may receive location information about the player from an external device. The external device may be a server. The external device may store location information received from the player tracking device (T). Alternatively, the external device may acquire the player's location data based on location information about the player and store the obtained location data. The communication interface 110 may receive location information about the player from the player tracking device (T).

The communication interface 110 is a Wi-Fi communication module, a cellular communication module, a 3G (3rd generation) mobile communication module, a 4G (4th generation) mobile communication module, a 4th generation LTE (Long Term Evolution) communication module, and a 5G (5th generation) mobile communication module. It may include at least one of a module or wired Ethernet.

The memory 120 may store an operating system (OS) for controlling the overall operation of the components of the electronic device 100 and commands or data related to the components of the electronic device 100. Memory 120 may store information about a plurality of neural network models. Information about the plurality of neural network models may include information about parameters corresponding to each of the plurality of neural network models and learning data for learning the plurality of neural network models. The memory 120 may be implemented as non-volatile memory (ex: hard disk, solid state drive (SSD), flash memory) or volatile memory.

The processor 130 is electrically connected to the memory 120 and can control the overall functions and operations of the electronic device 100. The processor 130 may control the electronic device 100 by executing instructions stored in the memory 120.

For example, the processor 130 may acquire a player's location data set (position data set for short) corresponding to a plurality of time points within a sports session. The player location dataset may include location values indicating the locations of a plurality of players. Sports sessions may include sessions for various events related to sports, such as match sessions and training sessions. A plurality of position values for a plurality of time points may include coordinates within a field indicating the positions of each of the plurality of players over time.

The processor 130 may obtain a plurality of position values in various ways. As an example, the processor 130 may receive a plurality of position values from an external device through the communication interface 110. The external device may be an external server that stores the location value obtained based on the location information obtained by the player tracking device (T). As another example, the processor 130 may receive the player's location information from the player tracking device (T). The processor 130 may obtain a location value based on the received location information.

Processor 130 may generate a team location dataset from the player location dataset. A team location dataset may reflect the placement of players belonging to a team on the stadium. Team location datasets may be permutation invariant.

The processor 130 may obtain speed data of multiple players. Speed data of a plurality of players may be acquired by a player tracking device (T) worn by each of the plurality of players. For example, speed data of a plurality of players may be obtained by a GPS sensor included in the player tracking device (T).

The processor 130 may identify a target event based on speed data of a plurality of players. For example, if the player's speed remains within a preset speed range for more than a threshold time in the first time period, the processor 130 may determine that a target event has occurred in the first time period. Alternatively, if the player's speed is greater than or equal to the threshold speed in the first time interval, the processor 130 may determine that the target event occurred in the first time interval.

The processor 130 may obtain information about the target event. Information about the target event may include time information corresponding to the target event and information about the player who performed the target event. For example, processor 130 may obtain a time stamp at which the target event is identified. Processor 130 may identify a player identifier corresponding to the identified target event.

The processor 130 may identify a plurality of location datasets corresponding to the target event from the plurality of location datasets based on time information corresponding to the target event. The processor 130 may identify a plurality of location data sets corresponding to the target event based on the time stamp corresponding to the target event section.

The processor 130 may obtain the type of target event based on a plurality of location data sets. Processor 130 may generate a plurality of team location datasets from a plurality of location datasets. The processor 130 may obtain the type of target event by inputting a plurality of team location datasets and the location value of the player who performed the target event into a neural network model.

The processor 130 may define a target event section. For example, the target event section is a first time section in which the instantaneous speed of the player who performed the target event is maintained within a preset range for more than a critical time, and a first time section from the time when the target event is intended by the player. It may include a second time section to the starting point, and a third time section from the end point of the first time section to a preset time after the end point of the first time section. At this time, the processor 130 may adjust the threshold time based on the distance traveled by the player who performed the target event during the first time period. For example, the greater the distance the player moves, the processor 130 can lower the threshold time. Meanwhile, the processor 130 may determine when the target event is intended based on the acceleration of the player who performed the target event. For example, the processor 130 may determine the time when the target player's acceleration becomes greater than a preset value as the time when the intent for the target event occurs.

The processor 130 may obtain the type of target event based on ball ownership in the target event section. For example, when the first team to which the target player belongs has possession of the ball in the target event section, the processor 130 uses a first neural network model learned to obtain the type of target event corresponding to the attack situation, The type of target event may be obtained based on the player's location value, the first team location dataset, and the second team location dataset. When the second team has possession of the ball in the target event section, the processor 130 uses a second neural network model learned to obtain the type of target event corresponding to the defense situation, and determines the position value of the target player and the second team. The type of target event can be obtained based on the location dataset.

Processor 130 may acquire ball possession based on a plurality of team location datasets and ball location datasets. For example, the processor 130 may acquire ball ownership by inputting a plurality of team location datasets and ball location datasets into a neural network model in segments of a predetermined time length.

The processor 130 may provide a video feed corresponding to the target event section to the user. Processor 130 may extract a target scene corresponding to a target event from a video feed for a sports session. The processor 130 may transmit a command to the user terminal to display the target scene. Additionally, the processor 130 may transmit a link for executing the target scene to the user terminal. The target scene may be provided to the user along with the type of target event determined by the processor 130. The user may input a command to modify the type of target event determined by the processor 130. Based on the command entered by the user, the processor 130 may update the type of target event.

Meanwhile, functions related to artificial intelligence according to the present disclosure are operated through the processor 130 and memory 120. The processor 130 may be comprised of one or multiple processors. At this time, one or more processors may be a general-purpose processor such as a CPU, AP, or DSP (Digital Signal Processor), a graphics-specific processor such as a GPU or VPU (Vision Processing Unit), or an artificial intelligence-specific processor such as an NPU. One or more processors control input data to be processed according to predefined operation rules or artificial intelligence models stored in the memory 120. Alternatively, when one or more processors are dedicated artificial intelligence processors, the artificial intelligence dedicated processors may be designed with a hardware structure specialized for processing a specific artificial intelligence model.

Predefined operation rules or artificial intelligence models are characterized by being created through learning. Here, being created through learning means that the basic artificial intelligence model is learned using a large number of learning data by a learning algorithm, thereby creating a predefined operation rule or artificial intelligence model set to perform the desired characteristics (or purpose). It means burden. This learning may be accomplished in the device itself that performs the artificial intelligence according to the present disclosure, or may be accomplished through a separate server and/or system. Examples of learning algorithms include supervised learning, unsupervised learning, semi-supervised learning, or reinforcement learning, but are not limited to the examples described above.

Artificial intelligence models can be created through learning. An artificial intelligence model may be composed of multiple neural network layers. Each of the plurality of neural network layers has a plurality of weight values, and neural network calculation is performed through calculation between the calculation result of the previous layer and the plurality of weights. Multiple weights of multiple neural network layers can be optimized by the learning results of the artificial intelligence model. For example, a plurality of weights may be updated so that loss or cost values obtained from the artificial intelligence model are reduced or minimized during the learning process.

Artificial neural networks may include deep neural networks (DNN), such as Convolutional Neural Network (CNN), Deep Neural Network (DNN), Recurrent Neural Network (RNN), Generative Adversarial Network (GAN), Examples include Restricted Boltzmann Machine (RBM), Deep Belief Network (DBN), Bidirectional Recurrent Deep Neural Network (BRDNN), or Deep Q-Networks, but are not limited to the examples described above.

The various embodiments described above may be implemented in a recording medium that can be read by a computer or similar device using software, hardware, or a combination thereof. In some cases, embodiments described herein may be implemented in a processor itself. When implemented in software, embodiments such as procedures and functions described in this specification may be implemented as separate software modules. Each of the software modules may perform one or more functions and operations described herein.

Computer instructions for performing processing operations according to various embodiments of the present disclosure described above may be stored in a non-transitory computer-readable medium. Computer instructions stored in such a non-transitory computer-readable medium, when executed by a processor, can cause a specific device to perform processing operations according to the various embodiments described above.

A non-transitory computer-readable medium refers to a medium that stores data semi-permanently and can be read by a device, rather than a medium that stores data for a short period of time, such as registers, caches, and memories. Specific examples of non-transitory computer-readable media may include CD, DVD, hard disk, Blu-ray disk, USB, memory card, ROM, etc.

A storage medium that can be read by a device may be provided in the form of a non-transitory storage medium. Here, 'non-transitory storage medium' only means that it is a tangible device and does not contain signals (e.g. electromagnetic waves). This term refers to cases where data is semi-permanently stored in a storage medium and temporary storage media. It does not distinguish between cases where it is stored as . For example, a 'non-transitory storage medium' may include a buffer where data is temporarily stored.

Methods according to various embodiments disclosed in this document may be provided and included in a computer program product. Computer program products are commodities and can be traded between sellers and buyers. The computer program product may be distributed in the form of a machine-readable storage medium (e.g. compact disc read only memory (CD-ROM)) or through an application store (e.g. Play StoreTM) or on two user devices (e.g. It can be distributed (e.g. downloaded or uploaded) directly between smartphones) or online. In the case of online distribution, at least a portion of the computer program product (e.g., a downloadable app) is stored on a machine-readable storage medium, such as the memory of a manufacturer's server, an application store's server, or a relay server. It can be temporarily stored or created temporarily.

In the above, preferred embodiments of the present disclosure have been shown and described, but the present disclosure is not limited to the specific embodiments described above, and may be used in the technical field pertaining to the disclosure without departing from the gist of the disclosure as claimed in the claims. Of course, various modifications can be made by those with the knowledge, and these modifications should not be understood individually from the technical ideas or perspectives of the present disclosure.

Claims

In a method of obtaining the type of target event within a sports session,

Obtaining a plurality of location datasets each corresponding to a plurality of time points within the sports session, each location dataset including location values of players at the corresponding time points;

Obtaining motion data of the plurality of players;

Identifying the target event based on the motion data, and obtaining information about the target event - the information includes time information corresponding to the target event and information about the player who performed the target event. step;

Situation information reflecting a game situation during the target event based on time information corresponding to the target event - the situation information includes a first location dataset reflecting the location of the player who performed the target event and the player Obtaining a second location dataset that reflects the placement or movement of at least one other player other than the player on the playing field; and

Comprising the step of obtaining the type of the target event based on the situation information.

method.
According to claim 1,

A plurality of team location datasets corresponding to the target event from the plurality of location datasets - the plurality of team location datasets are the arrangement status of players of the first team to which the player who performed the target event belongs in the stadium It further includes generating a plurality of first team location datasets reflecting a plurality of first team location datasets, and a plurality of second team location datasets reflecting the placement status of players of a second team in the stadium,

The step of obtaining the type of the target event is,

Obtaining the type of the target event by inputting at least one of the plurality of first team location datasets or the plurality of second team location datasets and the location values of the player who performed the target event into a neural network model.

method.
According to clause 2,

Further comprising: obtaining the position of a ball used in the sports session,

The step of obtaining the type of the target event is,

Obtaining the type of the target event by additionally inputting the position of the ball in the time interval corresponding to the target event into the neural network model

method.
According to clause 2,

The step of obtaining the type of the target event is,

Obtaining the type of the target event by additionally inputting the role information of the player who performed the target event into the neural network model

method.
According to claim 1,

The time information corresponding to the target event is,

A first time period in which the instantaneous speed of the player who performed the target event is maintained within a preset range for more than a critical time,

a second time interval from the point at which the target event is intended by the player to the starting point of the first time interval, and

Comprising a third time interval from the end point of the first time interval to a preset time after the end point of the first time interval.

method.
According to clause 5,

The critical time is,

Adjusted based on the distance moved by the player during the first time period

method.
According to clause 5,

When the target event is intended,

Determined based on the player's acceleration

method.
A computer-readable recording medium on which a program for executing the method of claim 1 on a computer is recorded.
An electronic device for obtaining a type of target event within a sports session, comprising:

a memory storing at least one instruction; and

Including a processor;

The processor executes the at least one instruction,

Obtaining a plurality of location datasets, each of which corresponds to a plurality of time points within the sports session, each location dataset including location values of players at the corresponding time points,

Obtaining motion data of the plurality of players,

Identify the target event based on the motion data, and obtain information about the target event - the information includes time information corresponding to the target event and information about the player who performed the target event. do,

Situation information reflecting a game situation during the target event based on time information corresponding to the target event - the situation information includes a first location dataset reflecting the location of the player who performed the target event and the player Obtaining a second location dataset that reflects the placement or movement on the playing field of at least one other player other than

Obtaining the type of the target event based on the situation information

Electronic devices.
According to clause 9,

The processor,

A plurality of team location datasets corresponding to the target event from the plurality of location datasets - the plurality of team location datasets are the arrangement status of players of the first team to which the player who performed the target event belongs in the stadium generating a plurality of first team location datasets reflecting a plurality of first team location datasets, and a plurality of second team location datasets reflecting the placement status of players of a second team in the stadium;

Obtaining the type of the target event by inputting at least one of the plurality of first team location datasets or the plurality of second team location datasets and the location values of the player who performed the target event into a neural network model.

Electronic devices.
According to claim 10,

The processor,

Obtaining the position of the ball used in the sports session,

Obtaining the type of the target event by additionally inputting the position of the ball in the time interval corresponding to the target event into the neural network model

Electronic devices.
According to claim 10,

The processor,

Obtaining the type of the target event by additionally inputting the role information of the player who performed the target event into the neural network model

Electronic devices.
According to clause 9,

The time information corresponding to the target event is,

A first time period in which the instantaneous speed of the player who performed the target event is maintained within a preset range for more than a threshold time,

a second time interval from when the target event is intended by the player to the starting point of the first time interval, and

Comprising a third time interval from the end point of the first time interval to a preset time after the end point of the first time interval.

Electronic devices.
According to claim 13,

The processor,

Adjusting the threshold time based on the distance moved by the player during the first time period

Electronic devices.
According to claim 13,

The processor,

Determining when the target event is intended based on the player's acceleration

Electronic devices.