Football match situation assessment method and system

Abstract

The invention relates to the technical field of match situation analysis, and discloses a football match situation assessment method and system.The football match situation assessment method comprises the steps that tracking data and event data in a football match are obtained, feature extraction processing is carried out, first feature data and second feature data are obtained, time alignment is carried out, and target fusion data are obtained; analyzing defense styles, attack stages and threat degrees of the match teams to obtain defense style information, attack stage information and threat degree information, determining football match analysis cues, and inputting the football match analysis cues into a big language model obtained based on big data training to obtain a match situation assessment result. According to the method, the tracking data and the event data are fused, so that multi-dimensional analysis of the competition situation is realized, the situation that only single-dimensional data can be analyzed in the prior art is avoided, and the defense style, the attack stage and the threat degree of a team are analyzed on the whole; and finally generating cue words and inputting the cue words into the large language model to obtain accurate football match situation analysis based on the global match.

Description

Technical Field

[0001] The present invention relates to the technical field of game situation analysis, and in particular to a football game situation evaluation method and system. Background Art

[0002] In football matches, situation assessment is a key link in understanding the progress of the game, conducting post-match analysis, and evaluating the performance of players and teams. There are two main types of situation assessment methods: The first type is player modeling based on theoretical dynamic models, which describes the player's movement behavior through mathematical models. This method lacks the ability to comprehensively analyze the overall situation of the game. The second type is to evaluate the specific actions of individual players, such as pass quality, shooting success rate, and interception effect, but this method can only provide analysis of individual performance and fails to take into account the overall tactical layout of the team, resulting in an incomplete understanding and analysis of the game situation.

[0003] Therefore, the existing technology for football game situation assessment ignores the importance of overall tactical layout and team collaboration in football games, and is unable to analyze the game situation from a global perspective. Summary of the invention

[0004] In view of this, the present invention provides a football match situation evaluation method and system to solve how to improve the comprehensiveness of football match situation analysis from a global perspective.

[0005] In a first aspect, the present invention provides a method for evaluating a football match situation, the method comprising:

[0006] Get tracking and event data for any moment in a football match;

[0007] Performing feature extraction processing on the tracking data and the event data to obtain first feature data after the tracking data processing and second feature data after the event data processing respectively;

[0008] Time-aligning the first feature data and the second feature data to obtain time-aligned target fusion data;

[0009] Analyzing the defensive style, offensive phase and threat level of the game team on the target fusion data to obtain defensive style information, offensive phase information and threat level information; wherein the target fusion data represents a data set obtained by aligning the first feature data and the second feature data on the same timeline;

[0010] Determine a football match analysis prompt word according to the target fusion data, the defensive style information, the offensive phase information, and the threat level information;

[0011] The football match analysis prompt words are input into a large language model obtained by training based on big data of real football matches to obtain a football match situation assessment result.

[0012] The embodiment of the present invention obtains tracking data and event data at any moment in a football match, performs feature extraction processing on the tracking data and event data, and obtains first feature data after tracking data processing and second feature data after event data processing, respectively, to ensure that there are enough valuable feature data for subsequent situation analysis; the first feature data and the second feature data are time-aligned to obtain target fusion data after time alignment, to ensure that data from different sources are consistent on the timeline, and to ensure that subsequent situation analysis can be performed from a global perspective. The target fusion data is analyzed for the defensive style, offensive phase and threat level of the game team to obtain defensive style information, offensive phase information and threat level information, and football match analysis prompt words are determined according to the target fusion data, defensive style information, offensive phase information and threat level information. Based on the above three analyses and combined with all the fusion data, prompt words are obtained, which can ensure that the generated prompt words have better description capabilities, so that the subsequent large language model analysis results are more in line with the actual game process, and the football match analysis prompt words are input into the large language model obtained by big data training based on real football matches to obtain football match situation evaluation results, and by using the trained large language model, a comprehensive football match situation analysis result is generated.

[0013] In an optional embodiment, the first feature data includes at least ball position information, team information, identification information of the defending team, identity information of the players, position information of the players, speed information of the players, coordinate information of the court, match information, remaining time of the match, identification information of the current ball-controlling player, and information of the current ball-controlling team;

[0014] The second characteristic data at least includes information about the team where the event occurs, identity information of players in the team where the event occurs, event type information, event result information, starting position information where the event occurs, and ending position information where the event occurs.

[0015] In the embodiment of the present invention, the first feature data includes static information of various players, teams, stadiums, and balls, and the second feature data includes dynamic information of events. In the subsequent situation analysis, there will be enough feature data with certain value, thereby improving the accuracy of the analysis and helping the large language model to better understand the game situation.

[0016] In an optional implementation, the target fusion data is subjected to a game team defense style analysis to obtain defense style information, specifically including:

[0017] Based on the player's position information, team information, stadium coordinate information, ball position information, current ball-controlling player identification information, current ball-controlling team information, event type information, event start position information, and event end position information, the offensive team and the defensive team are clustered to obtain the position information of each offensive line and each defensive line in the stadium;

[0018] According to the position information of each offensive line and the position information of each defensive line, the arrangement and combination information between the offensive line and the defensive line is determined;

[0019] The defensive style information of the game team is determined according to the permutation and combination information.

[0020] The embodiment of the present invention performs cluster analysis based on multi-dimensional data such as player position information, team information, and stadium coordinate information, and obtains the position information of each offensive line and defensive line. It can fully clarify the team's tactical arrangements in the game, and according to the arrangement and combination information between the offensive line and the defensive line, it can analyze and judge the team's tactical changes in real time, which is more in line with the actual game situation, and generate specific defensive style information, providing a data basis related to the defensive style for the determination of subsequent game prompts.

[0021] In an optional implementation, the target fusion data is analyzed for the offensive phase of the game team to obtain offensive phase information, specifically including:

[0022] Determine first relative position information between the player and the ball according to the player's position information, the court's coordinate information, and the ball's position information;

[0023] Determine the type of the offensive line controlling the ball according to the position information of each offensive line, the starting position information of the event, the ending position information of the event, and the first relative position information;

[0024] Determines offensive phase information based on the type of offensive line in control of the ball.

[0025] The embodiment of the present invention can obtain a more accurate offensive situation analysis by determining the first relative position information between the player and the ball based on the player's position information, the coordinate information of the court and the position information of the ball. The type of offensive line that controls the ball is determined based on the position information of each offensive line, the start and end position information of the event and the first relative position information. The classification of offensive methods is clearer and has a basis. The offensive phase information is determined based on the type of offensive line that controls the ball, providing a data basis related to the offensive phase information for the subsequent generation of game prompt words.

[0026] In an optional implementation, the target fusion data is subjected to a threat level analysis of the game team to obtain threat level information, specifically including:

[0027] Divide the court area for threat analysis based on the court's coordinate information;

[0028] Determine second relative position information between the ball and the court according to the event type information, the position information of the ball and the court area;

[0029] The threat level information of the playing team is determined according to the second relative position information.

[0030] The embodiment of the present invention divides the threat analysis court area based on the coordinate information of the court, combines the event type information, the position information of the ball and the court area, and determines the second relative position information between the ball and the court, so as to determine the threat level information in the dynamic football game, and provide a data basis related to the threat level information for the subsequent generation of game prompt words.

[0031] In an optional implementation manner, determining the football match analysis prompt words according to the target fusion data, the defensive style information, the offensive phase information and the threat level information specifically includes:

[0032] A data set of first feature data and second feature data aligned on the same timeline and corresponding defensive style information, offensive phase information and threat level information are input into a preset prompt word template to obtain football match analysis prompt words; wherein the preset prompt word template represents a question template obtained by performing semantic feature analysis based on football match analysis data.

[0033] The embodiment of the present invention integrates the first feature data and the second feature data aligned on the same timeline, and combines the corresponding defensive style information, offensive phase information and threat level information. The generated prompt words have better descriptive ability, and the subsequent analysis results of the large language model are more in line with the actual game process, ensuring that the analysis results are not only comprehensive but also have higher accuracy.

[0034] In a second aspect, the present invention provides a football match situation assessment system, the system comprising:

[0035] A data acquisition module is used to obtain tracking data and event data at any moment in a football match;

[0036] A feature extraction module, used to perform feature extraction processing on the tracking data and the event data, and obtain first feature data after the tracking data is processed and second feature data after the event data is processed;

[0037] A time alignment module, used for performing time alignment on the first feature data and the second feature data to obtain time-aligned target fusion data;

[0038] an analysis module, configured to analyze the target fusion data in terms of the defensive style, offensive phase, and threat level of the game team, and obtain defensive style information, offensive phase information, and threat level information; wherein the target fusion data represents a data set obtained by aligning the first feature data and the second feature data on the same timeline;

[0039] A prompt word generation module, used for determining a football match analysis prompt word according to the target fusion data, the defensive style information, the offensive phase information and the threat level information;

[0040] The evaluation result module is used to input the football match analysis prompt words into a large language model obtained by training based on big data of real football matches to obtain a football match situation evaluation result.

[0041] In a third aspect, the present invention provides a computer device, comprising: a memory and a processor, the memory and the processor being communicatively connected to each other, the memory storing computer instructions, and the processor executing the football match situation assessment method of the first aspect or any corresponding embodiment thereof by executing the computer instructions.

[0042] In a fourth aspect, the present invention provides a computer-readable storage medium having computer instructions stored thereon, the computer instructions being used to enable a computer to execute the football match situation assessment method of the first aspect or any corresponding embodiment thereof.

[0043] In a fifth aspect, the present invention provides a computer program product, comprising computer instructions for causing a computer to execute the football match situation assessment method of the first aspect or any corresponding embodiment thereof. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0045] Figure 1 is a flowchart of a football match situation assessment method according to an embodiment of the present invention;

[0046] Figure 2 is a flow chart of another football match situation assessment method according to an embodiment of the present invention;

[0047] Figure 3 is a schematic diagram of defensive style arrangement and combination according to a football match situation assessment method according to an embodiment of the present invention;

[0048] Figure 4 is a schematic diagram of court area division according to a football match situation assessment method according to an embodiment of the present invention;

[0049] Figure 5 is a flow chart of another football match situation assessment method according to an embodiment of the present invention

[0050] Figure 6 is a structural block diagram of a football match situation assessment system according to an embodiment of the present invention;

[0051] Figure 7 It is a schematic diagram of the hardware structure of a computer device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0052] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0053] Football is a typical group intelligence confrontation sport, in which situation assessment plays an important role in the progress of the game, post-match analysis, and evaluation of players and teams. At the same time, in group games, relying solely on whether a player scores or wins the game can no longer meet the needs of player performance evaluation and tactical analysis.

[0054] One of the traditional methods for evaluating the situation of a football match is to use a theoretical dynamic model to model the players. Although this method has a certain theoretical basis, it only provides a basic general small model, which cannot be applied to different scene analyses and cannot provide comprehensive situation analysis. Another method is to evaluate the actions of players, especially those holding the ball, such as evaluating the quality of a pass, evaluating whether a shot can be successful, whether an interception can be successful, etc. This type of method can only evaluate the specific actions of a single player, ignoring the importance of the overall tactical layout and team collaboration in a football match, and cannot analyze the situation of the game from a global perspective.

[0055] The present invention provides a football match situation assessment method, which is applied to scenarios such as real-time football match analysis, post-match analysis of football matches, player assessment, team assessment, and football match tactical research. By integrating tracking data and event data, a multi-dimensional analysis of the game situation is achieved. Through time alignment processing, dynamic assessment can be performed at different stages of the game, avoiding the situation where the existing technology can only analyze data of a single dimension. The team's defensive style, offensive stage, and threat level are analyzed as a whole, providing a comprehensive analysis. The final generated prompt words are combined with a large language model to obtain a more accurate football match situation analysis based on the overall situation of the game.

[0056] According to an embodiment of the present invention, an embodiment of a football match situation assessment method is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0057] In this embodiment, a football match situation assessment method is provided, which can be used in the above-mentioned computer. Figure 1 is a flow chart of a football match situation assessment method according to an embodiment of the present invention. Figure 1 As shown, the process includes the following steps:

[0058] Step S101, obtaining tracking data and event data at any moment in a football match.

[0059] In this embodiment, the tracking data may include all tracking data of players and balls in the court area, where the court area is the entire area on the football field where players can run freely. Players may be virtual characters in a ball game simulation program, players in a real ball game, intelligent robots in a ball game, or intelligent entities in the field of artificial intelligence. The types of players are not specifically limited here. The tracking data may be collected through wearable devices (such as sensors on sports vests or shoes worn by players), or may be extracted from game videos through computer vision technology. The built-in GPS or other positioning technology of the wearable device can ensure that the precise position, speed, acceleration and other parameters of each player are captured in real time.

[0060] It is understandable that when the players are in a virtual environment, the tracking data of the players and the ball can be tracked and extracted in the virtual environment through a program; when the players are in a real ball game environment, the players and the ball can be pre-configured or wear positioning devices and motion capture devices, and the tracking data of each player and the ball can be captured in real time through the positioning devices and motion capture devices. The tracking data at any time in the game is the original data extracted from any time in any game. The time interval for data tracking can be pre-set, for example, data is extracted every 1s or 1ms; the data extraction time interval can also be defined by pre-setting the number of tracking data. For example, if a game is 90 minutes, the number of tracking data can be set to 5000, and then data is extracted every 1s. The embodiment of the present invention does not specifically limit this.

[0061] In this embodiment, the event data is a description of key events in the game, including passes, shots, fouls, etc. The acquisition of event data can be achieved through manual annotation (for example, a live commentator or data analyst records the event in real time), or through automated game analysis software, wherein the automated game analysis software can automatically identify and mark events in the game based on the player actions and ball movement patterns in the game video.

[0062] For example, in the 15th minute of the game, player A passes the ball from the midfield to player B. The tracking data at this moment will record the real-time position, speed and other information of player A and player B; the event data is based on the automated match analysis software to mark the start and end time of the above pass, the identity of player A and player B, and the starting and ending positions of the pass. Only by obtaining sufficiently accurate and comprehensive data can we provide data support for subsequent feature extraction, time alignment and situation analysis, which is the basis for completing the situation analysis of the entire football game.

[0063] Step S102 , performing feature extraction processing on the tracking data and the event data to obtain first feature data after the tracking data processing and second feature data after the event data processing, respectively.

[0064] Specifically, for tracking data, data completion needs to be performed through interpolation, and for event data, different marking methods need to be unified to obtain a consistent description method.

[0065] For example, the tracking data records the specific position and speed of player A at 15 minutes and 12 seconds of the game, and the processed first feature data will retain the identity, position and speed information of player A at this moment. The event data records a pass made by player A at that moment, and the processed second feature data will retain the start and end time of the pass, the identity of player A, and the result of the pass (whether it was successfully passed to the target player).

[0066] This embodiment takes into account that the original features of the tracking data contain many data features, and it is difficult to understand the complex time series data features by directly inputting them into the large language model. Some of these features have a greater relevance in the evaluation and prediction of the game situation, such as the position and speed of the ball and the player, while other features have a relatively smaller relevance in the evaluation and prediction of the game situation, such as the player's role. At the same time, considering that some features with a greater relevance to the evaluation and prediction of the game situation cannot be directly obtained from the original features of the tracking data, it is necessary to perform feature engineering on the original features of the tracking data, and the feature engineering process may specifically include advanced feature extraction and feature screening. For example, features such as player roles that have a relatively smaller relevance in the evaluation and prediction of the game situation can be screened out, and features with a relatively larger relevance can be retained.

[0067] Step S103: Time-align the first feature data and the second feature data to obtain time-aligned target fusion data.

[0068] Specifically, a time alignment algorithm is needed to time align the tracking data and event data of different granularities and annotation methods in the same game, and obtain the tracking data and event data at the same time. Time alignment ensures that the analysis model can combine the tracking data and event data at the same time. Only when the timeline is consistent can the situation analysis be comprehensive and accurate.

[0069] For example, Player A passes the ball at 15 minutes and 12 seconds into the game, and the event data marks the time of the pass; while the tracking data records the player's position information 10 times per second. The result of time alignment will ensure that at 15 minutes and 12 seconds, there is both Player A's pass event data and tracking data of his position and speed.

[0070] Step S104, analyzing the target fusion data in terms of the defensive style, offensive phase and threat level of the game team to obtain defensive style information, offensive phase information and threat level information; wherein the target fusion data represents a data set obtained by aligning the first feature data and the second feature data on the same timeline.

[0071] Specifically, the defensive style needs to be clustered through the player position analysis. Using clustering algorithms such as K-means, the position of each player is clustered to form the offensive and defensive lines of the offensive and defensive teams. The relative positions of the offensive and defensive lines determine the team's defensive style.

[0072] The attack phase is analyzed based on the relative position of the ball and the attacking players. The attack process is divided into three phases: build-up attack, passing attack, and threatening attack. If the ball is in a farther position and the players are building the attack, it is the build-up attack phase; when the ball is passed to an area closer to the opponent's goal, it enters the threatening attack phase.

[0073] Threat level analysis is based on the position of the ball on the court, dividing the court into multiple zones (such as the 16 zones defined by Athletic), each with a different threat level.

[0074] Step S105, determining football match analysis prompt words according to the target fusion data, the defensive style information, the offensive phase information and the threat level information.

[0075] It should be noted that the prompt word design is a description of the tracking data, event data and game situation analysis results to guide the large language model to further evaluate the game situation. The prompt word passes the key information of the game to the large language model to ensure that the model can perform accurate analysis based on this information. The prompt word generation process can be automated, for example, through a preset template, specific information is inserted into the template to obtain the prompt word.

[0076] For example, for the overall game overview, the prompt words are designed as follows:

[0077] "Based on the data at the current moment of the game and the data at previous moments, as well as the above data (target fusion data, defensive style information, offensive phase information and threat level information), please tell me the time of the current game, red cards, yellow cards and substitutions, and from an overall perspective, tell me the stage of the current game and the analysis reasons for the current stage of the game."

[0078] In view of the macro situation of football games, the prompt words are designed as follows:

[0079] "Based on the data provided (target fusion data, defensive style information, offensive phase information, and threat level information), I hope you can help me analyze the macro situation of the game from the following three aspects.

[0080] First, from the perspective of formation, please help me analyze the formations of the home team and the visiting team in the current game, as well as the reasons for this analysis and the advantages and disadvantages of the current formation.

[0081] Second, the positions and speeds of all players are known in the data. From the perspective of the players' position distribution and speed, please help me analyze the current macro situation of the home team and the reasons for reaching this conclusion.

[0082] Third, the known data includes the position and speed of the ball. From the perspective of the ball's speed, position and speed, please help me analyze the current situation of the game and the reasons for reaching this conclusion.

[0083] Fourth, from the perspective of the relative positions of the ball and the players, please help me analyze the current situation of the home team and the visiting team, the game strategies they have adopted, and the reasons for reaching this conclusion.

[0084] Finally, based on the above analysis, a summary of the macro situation of the game is made. "

[0085] Step S106, inputting the football match analysis prompt words into a large language model trained based on big data of real football matches to obtain a football match situation assessment result.

[0086] It should be noted that in the embodiment of the present invention, the large language model (such as GPT-3.5-turbo) is trained through a large amount of historical football game data and has the ability to analyze the game situation. When the prompt word is input into the large language model, the model will generate a comprehensive analysis of the current game situation based on the prompt word information context. The football game situation assessment results include at least a summary of the game overview, a macro situation analysis (formation, player position, ball position, relative position of the ball and player), a local situation analysis (weak areas and advantageous areas), etc.

[0087] For example, for the summary of the game, the output of the large language model is as follows:

[0088] "Based on the provided tracking data, we can make the following analysis: Game time: The current frame number is 621. If we use the total number of frames in the first half, 28081, as a reference, we can estimate it this way: Assuming the first half is about 45 minutes long, there are about 28081 / 45 ~ 624 frames per minute. Therefore, frame 621 roughly corresponds to the end of the first minute of the first half.

[0089] Game phase: Since the game is at the beginning of the first half, we can assume that the game is still in the early stages. This period is usually characterized by the two teams trying to control the tempo of the game, establish a rhythm on the court, and find weaknesses in the opposing defense.

[0090] The positional data on the pitch shows that the game is in the early stages of the first half, there is a potential attack in the right area, and the players are just beginning to deploy their tactical deployment.

[0091] 1. Analysis of player position distribution: The home team goalkeeper and several defensive players are in their own half, which indicates that the home team may be looking for an opportunity to launch an attack from a defensive position. In particular, the goalkeeper's position is close to the goal, suggesting that the home team may be in a defensive state. 2. Analysis of the ball's position: The ball is in a high position close to the right edge, which may mean that a wing attack is in progress or a long pass across the midfield. 3. Analysis of the relative position of the ball and the players: The forward player's position is closer to the opponent's half and closer to the ball, and may be preparing to receive a cross or use the high space of the opponent's back line to launch an attack.

[0092] Analysis from the perspective of formation: Judging from the positions of the players, the home team may have arranged a formation that emphasizes midfield control, such as 4-4-2 or 4-2-3-1, with players spread out in the wide front 4 and midfield to control the ball and launch attacks, while the central defenders remain compact to block the opponent's attacks.

[0093] The embodiment of the present invention obtains tracking data and event data at any moment in a football match, performs feature extraction processing on the tracking data and event data, and obtains first feature data after tracking data processing and second feature data after event data processing, respectively, to ensure that there are enough valuable feature data for subsequent situation analysis; the first feature data and the second feature data are time-aligned to obtain target fusion data after time alignment, to ensure that data from different sources are consistent on the timeline, and to ensure that subsequent situation analysis can be performed from a global perspective. The target fusion data is analyzed for the defensive style, offensive phase and threat level of the game team to obtain defensive style information, offensive phase information and threat level information, and football match analysis prompt words are determined according to the target fusion data, defensive style information, offensive phase information and threat level information. Based on the above three analyses and combined with all the fusion data, prompt words are obtained, which can ensure that the generated prompt words have better description capabilities, so that the subsequent large language model analysis results are more in line with the actual game process, and the football match analysis prompt words are input into the large language model obtained by big data training based on real football matches to obtain football match situation evaluation results, and by using the trained large language model, a comprehensive football match situation analysis result is generated.

[0094] In this embodiment, a football match situation assessment method is provided, which can be used in the above-mentioned computer. Figure 2 is a flow chart of a football match situation assessment method according to an embodiment of the present invention. Figure 2 As shown, the process includes the following steps:

[0095] Step S201, obtain tracking data and event data at any time in the football game. Figure 1 Step S101 of the illustrated embodiment will not be described in detail here.

[0096] Step S202: Perform feature extraction on the tracking data and event data to obtain first feature data after the tracking data is processed and second feature data after the event data is processed. Figure 1 Step S102 of the illustrated embodiment will not be described in detail here.

[0097] Step S203: Time-align the first feature data and the second feature data to obtain time-aligned target fusion data.

[0098] In an optional embodiment, the first feature data includes at least ball position information, team information, identification information of the defending team, identity information of the players, position information of the players, speed information of the players, coordinate information of the court, match information, remaining time of the match, identification information of the current ball-controlling player, and information of the current ball-controlling team;

[0099] The second characteristic data at least includes information about the team where the event occurs, identity information of players in the team where the event occurs, event type information, event result information, starting position information where the event occurs, and ending position information where the event occurs.

[0100] In the embodiment of the present invention, the first feature data includes static information of various players, teams, stadiums, and balls, and the second feature data includes dynamic information of events. In the subsequent situation analysis, there will be enough feature data with certain value, thereby improving the accuracy of the analysis and helping the large language model to better understand the game situation.

[0101] Step S204, analyzing the target fusion data in terms of the defensive style, offensive phase and threat level of the game team to obtain defensive style information, offensive phase information and threat level information; wherein the target fusion data represents a data set obtained by aligning the first feature data and the second feature data on the same timeline.

[0102] Specifically, the above step S204 includes:

[0103] Step S2041, analyzing the defensive style of the game team on the target fusion data to obtain defensive style information.

[0104] Specifically, the above step S2041 includes:

[0105] A1, based on the player's position information, team information, stadium coordinate information, ball position information, current ball-controlling player identity information, current ball-controlling team information, event type information, event start position information and event end position information, the offensive team and the defensive team are clustered and analyzed to obtain the position information of each offensive line and each defensive line in the stadium.

[0106] Specifically, the real-time position information of each player is obtained, including their X and Y coordinates on the field, the position information of the ball, the information of the player and team in possession of the ball, the starting and ending position information in the event data, and the players of the offensive and defensive teams are grouped according to their position information using a clustering algorithm to obtain the spatial distribution of the players. The players of the offensive team are divided into three offensive lines: forward line, midfield line and back line. The players of the defensive team are divided into three defensive lines: back line, midfield line and forward line. Finally, the offensive and defensive lines are divided, and the average position of each offensive and defensive line is calculated.

[0107] A2, determining the permutation and combination information between the offensive line and the defensive line according to the position information of each offensive line and the position information of each defensive line.

[0108] Specifically, according to the relative positions of the offensive line and the defensive line, there are various combinations of the three offensive lines of the offensive side and the three defensive lines of the defensive side. For example, if all three offensive lines are ahead of the defensive line, it means that the offensive side has a strong offensive posture; if all three defensive lines are ahead of the offensive line, it means that the defensive side has adopted a solid defensive strategy; and if the offensive line and the defensive line partially cross, it means that the two sides may be in the midfield competition stage.

[0109] A3, determining the defensive style information of the game team according to the permutation and combination information.

[0110] Specifically, there is no fuzzy classification. According to the arrangement relationship between the 6 formation lines (3 for the attacking side and 3 for the defending side), through the combination relationship There are 20 situations in total. Specifically, both the attacking side and the defending side are arranged in a line of defender-midfielder-forward. Therefore, you only need to select 3 of the 6 lines as the attacking side, and the attack line will be automatically arranged, and the remaining 3 will be automatically arranged as the defense line. When the attacking line and the defense line happen to be in the same position, you can set a fixed rule, such as arranging the position closer to your own goal.

[0111] For example, see Figure 3, cluster all the members of the attacking team and the defending team separately into three attacking lines (a, b, c) and defensive lines (d, e, f), and divide the defensive style into 20 selection cases according to the permutations and combinations of the distribution of the six lines, where 0 represents the attacking team and 1 represents the defending team. Each permutation and combination corresponds to a mark (for example, 001011, the positions of 0 are 1, 2, and 4 respectively, so the sum of the positions of 0 is 1+2+4=7, and the sum of the positions of 1 is 2+5+6=14). From left to right is the direction from attack to defense of the goal, the permutation and combination 000111 corresponds to low defense, and the permutation and combination 111000 corresponds to high pressure.

[0112] The embodiment of the present invention performs cluster analysis based on multi-dimensional data such as player position information, team information, and stadium coordinate information, and obtains the position information of each offensive line and defensive line. It can fully clarify the team's tactical arrangements in the game, and according to the arrangement and combination information between the offensive line and the defensive line, it can analyze and judge the team's tactical changes in real time, which is more in line with the actual game situation, and generate specific defensive style information, providing a data basis related to the defensive style for the determination of subsequent game prompts.

[0113] Step S2042, analyzing the offensive phase of the game team on the target fusion data to obtain offensive phase information.

[0114] Specifically, the above step S2042 includes:

[0115] B1. Determine first relative position information between the player and the ball according to the player's position information, the coordinate information of the court, and the position information of the ball.

[0116] Specifically, the actual distance between the player and the ball can be calculated using a distance formula between two points through a geometric formula.

[0117] B2. Determine the type of the offensive line that controls the ball according to the position information of each offensive line, the starting position information of the event, the ending position information of the event, and the first relative position information.

[0118] Specifically, the position information of each offensive line has been determined in step A1, which is represented by the average position of all players on the line. The starting position and the ending position of the pass are obtained from the event data. For example, at the current moment, the ball is controlled by player A, and the event that occurs is a pass. Then, based on the starting position and the ending position of the pass, the position information of each offensive line, and the first relative position, it is determined on which offensive line the ball is located during the passing process.

[0119] B3, determining the offensive phase information according to the type of offensive line controlling the ball.

[0120] For details, please refer to Figure 3 , according to the ball being controlled by the first (a offensive line), the second (b offensive line) and the third (c offensive line) offensive lines respectively, it is divided into the organizing offensive phase, the passing offensive phase and the threatening offensive phase.

[0121] The embodiment of the present invention can obtain a more accurate offensive situation analysis by determining the first relative position information between the player and the ball based on the player's position information, the coordinate information of the court and the position information of the ball. The type of offensive line that controls the ball is determined based on the position information of each offensive line, the start and end position information of the event and the first relative position information. The classification of offensive methods is clearer and has a basis. The offensive phase information is determined based on the type of offensive line that controls the ball, providing a data basis related to the offensive phase information for the subsequent generation of game prompt words.

[0122] Step S2043, analyzing the threat level of the game team on the target fusion data to obtain threat level information.

[0123] Specifically, the above step S2043 includes:

[0124] C1, divide the court area for threat analysis based on the coordinate information of the court.

[0125] Specifically, Figure 4 -(a), the court can be divided into 15 areas according to five horizontal areas and three vertical areas; Figure 4 -(b) As shown in Figure 1, the court can also be divided into several small areas vertically and horizontally according to the 16-area division method proposed by Athletic. For example, the court can be divided into three attacking zones vertically (one's own half, midfield, and the opponent's half), and each zone can be further divided into multiple small areas; Figure 4 - As shown in (c), it is also possible to follow the 20 areas of Manchester City coach Guardiola. At the same time, a unique identifier is assigned to each small area. For example, numbers 1 to 16 can be used to represent different areas.

[0126] C2, determining second relative position information between the ball and the court according to the event type information, the position information of the ball and the court area.

[0127] For example, the ball is currently in the opponent's half (number 10) and the event is a shot, then the relative distance between the ball and the goal area (such as number 16 area) will affect the threat assessment. If the ball is close to the penalty area, the threat is higher; if the ball is far away from the penalty area, the threat is lower.

[0128] C3: Determine threat level information of the playing team according to the second relative position information.

[0129] The embodiment of the present invention divides the threat analysis court area based on the coordinate information of the court, combines the event type information, the position information of the ball and the court area, and determines the second relative position information between the ball and the court, so as to determine the threat level information in the dynamic football game, and provide a data basis related to the threat level information for the subsequent generation of game prompt words.

[0130] Specifically, the threat level is divided into several levels, namely low threat, medium threat and high threat.

[0131] For example, player C initiates a pass from the center line, the position of the ball is (50, 30), the event type is a pass, and the end point of the ball is in the opponent's penalty area (number 10); at this time, the calculated distance between the ball and the opponent's goal is higher than the preset value, and the pass enters the threat area (number 10).

[0132] Step S205, determining football match analysis prompt words according to the target fusion data, defense style information, attack phase information and threat level information. Figure 1 Step S105 of the illustrated embodiment will not be described in detail here.

[0133] Step S206, input the football match analysis prompt words into the large language model trained based on the big data of real football matches to obtain the football match situation assessment result. Figure 1 Step S106 of the illustrated embodiment will not be described in detail here.

[0134] In this embodiment, a method for evaluating football match situation is provided, and the process includes the following steps:

[0135] Step S501, obtain tracking data and event data at any time in the football game. Figure 1 Step S101 of the illustrated embodiment will not be described in detail here.

[0136] Step S502: Perform feature extraction on the tracking data and event data to obtain first feature data after the tracking data is processed and second feature data after the event data is processed. Figure 1 Step S102 of the illustrated embodiment will not be described in detail here.

[0137] Step S503: Time-align the first feature data and the second feature data to obtain time-aligned target fusion data. Figure 1 Step S103 of the illustrated embodiment will not be described in detail here.

[0138] Step S504: Analyze the defensive style, offensive phase and threat level of the game team on the target fusion data to obtain defensive style information, offensive phase information and threat level information; wherein the target fusion data represents a data set obtained by aligning the first feature data and the second feature data on the same timeline. For details, please refer to Figure 1 Step S104 of the illustrated embodiment will not be described in detail here.

[0139] Step S505, determining football match analysis prompt words according to the target fusion data, the defensive style information, the offensive phase information and the threat level information.

[0140] Specifically, step S505 includes: inputting a data set of first feature data and second feature data aligned on the same timeline and corresponding defensive style information, offensive phase information and threat level information into a preset prompt word template to obtain football match analysis prompt words; wherein the preset prompt word template represents a question template obtained by performing semantic feature analysis based on football match analysis data.

[0141] It is understandable that due to the large number of dimensions and types of raw data, directly using this data for analysis will increase the processing burden of the large language model and may lead to deviations in the analysis results. Therefore, by using prompt word templates, the raw competition data is converted into structured and semantically clear descriptions, which improves the description ability and ensures the accuracy of the model analysis results.

[0142] It should be noted that the prompt word template is a formatted description framework that embeds key information in the game (defensive style, offensive phase, threat level, target fusion data) into a specific position in the template. Each template can be used to generate different types of questions or descriptions.

[0143] For example, the content of a prompt word template is as follows:

[0144] "The game data is as follows: At time A, team B1 is at position b, team C1 is at position c, the current attacker is in area X, the defender adopts defensive strategy Y, and the threat level is Z. Does the operation of the attacker's player s cause pressure on the defender higher than Z?" Each of the above letter-marked positions is a placeholder and can be replaced by real data.

[0145] Step S506: Input the football match analysis prompt words into the large language model trained based on the big data of real football matches to obtain the football match situation assessment result. Figure 1 Step S106 of the illustrated embodiment will not be described in detail here.

[0146] The embodiment of the present invention integrates the first feature data and the second feature data aligned on the same timeline, and combines the corresponding defensive style information, offensive phase information and threat level information. The generated prompt words have better descriptive ability, and the subsequent analysis results of the large language model are more in line with the actual game process, ensuring that the analysis results are not only comprehensive but also have higher accuracy.

[0147] The present invention provides a football match situation assessment method, the overall process of which can be referred to Figure 5 , obtain the original features of the tracking data at any moment in the game, obtain the original features of the event data at any moment in the game, extract and process the original features of the tracking data to obtain the first feature data after the tracking data is processed, extract and process the original features of the event data to obtain the second feature data after the tracking data is processed, design a small model for game situation analysis from three perspectives: defensive style, offensive stage and threat level, design prompt words for the feature data of the tracking data and event data, and the small model for situation analysis, input the feature data of the tracking data and event data, the small model for situation analysis and the prompt words into the large language model for game situation evaluation to obtain the football game situation evaluation result, and the evaluation result includes a summary of the game overview, a macro situation analysis (formation, player position, ball position and relative position of the ball and player), a local situation analysis (weak areas and advantageous areas), etc.

[0148] Compared with the existing technology that only evaluates the players in contact with the ball, this method efficiently utilizes the tracking data and event data of all players and the ball in the court area, and realizes comprehensive situation evaluation and interpretability evaluation from a global perspective. Compared with the existing small model situation evaluation that can only evaluate one type of data or one situation, it realizes the time alignment of tracking data and event data, as well as the comprehensive evaluation of multi-level and multi-angle game situations. At the same time, the situation evaluation results of the game can assist players, coaches and other personnel in tactical analysis and post-match review.

[0149] In the present embodiment, a football match situation assessment system is also provided, which is used to implement the above-mentioned embodiment and preferred implementation mode, and the description thereof will not be repeated. As used below, the term "module" can implement a combination of software and / or hardware of a predetermined function. Although the system described in the following embodiments is preferably implemented with software, the implementation of hardware, or a combination of software and hardware is also possible and contemplated.

[0150] This embodiment provides a football match situation assessment system. Figure 6 As shown, including:

[0151] Data acquisition module 601, used to acquire tracking data and event data at any moment in a football match;

[0152] A feature extraction module 602 is used to perform feature extraction processing on the tracking data and the event data to obtain first feature data after the tracking data is processed and second feature data after the event data is processed;

[0153] A time alignment module 603 is used to perform time alignment on the first feature data and the second feature data to obtain target fusion data after time alignment;

[0154] The analysis module 604 is used to analyze the defensive style, offensive phase and threat level of the game team on the target fusion data to obtain defensive style information, offensive phase information and threat level information; wherein the target fusion data represents a data set obtained by aligning the first feature data and the second feature data on the same time line;

[0155] A prompt word generation module 605, for determining a football match analysis prompt word according to the target fusion data, the defensive style information, the offensive phase information and the threat level information;

[0156] The evaluation result module 606 is used to input the football match analysis prompt words into a large language model trained based on big data of real football matches to obtain a football match situation evaluation result.

[0157] In an optional embodiment, the first feature data includes at least ball position information, team information, identification information of the defending team, identity information of the players, position information of the players, speed information of the players, coordinate information of the court, match information, remaining time of the match, identification information of the current ball-controlling player, and information of the current ball-controlling team;

[0158] The second characteristic data at least includes information about the team where the event occurs, identity information of players in the team where the event occurs, event type information, event result information, starting position information where the event occurs, and ending position information where the event occurs.

[0159] In an optional implementation, the analysis module 604 specifically includes:

[0160] A cluster analysis unit, for performing cluster analysis on the attacking team and the defending team based on the player's position information, the team information, the coordinate information of the court, the ball's position information, the current ball-controlling player's identity information, the current ball-controlling team's information, the event type information, the event's starting position information, and the event's ending position information, to obtain the position information of each attack line and each defense line in the court;

[0161] A permutation and combination unit, used to determine the permutation and combination information between the offensive line and the defensive line according to the position information of each offensive line and the position information of each defensive line;

[0162] The defensive style unit is used to determine the defensive style information of the game team according to the permutation and combination information.

[0163] In an optional implementation, the analysis module 604 specifically includes:

[0164] A relative position unit, used to determine first relative position information between the player and the ball according to the player's position information, the coordinate information of the court, and the position information of the ball;

[0165] An offensive line type unit, configured to determine the type of the offensive line controlling the ball according to the position information of each offensive line, the starting position information of the event, the ending position information of the event, and the first relative position information;

[0166] An offensive phase unit is used to determine offensive phase information based on the type of offensive line that controls the ball.

[0167] In an optional implementation, the analysis module 604 specifically includes:

[0168] A threat area analysis unit, used for dividing a court area for threat analysis based on the coordinate information of the court;

[0169] A second relative position unit, used for determining second relative position information between the ball and the court according to the event type information, the position information of the ball and the court area;

[0170] A threat level unit is used to determine the threat level information of the playing team according to the second relative position information.

[0171] In an optional implementation, the prompt word generating module 605 specifically includes:

[0172] The analysis prompt word generation unit is used to input the data set of the first feature data and the second feature data aligned on the same timeline and the corresponding defensive style information, offensive phase information and threat level information into a preset prompt word template to obtain football match analysis prompt words; wherein the preset prompt word template represents a question template obtained by performing semantic feature analysis based on the football match analysis data.

[0173] The further functional description of each of the above modules and units is the same as that of the above corresponding embodiments and will not be repeated here.

[0174] The football match situation assessment system in this embodiment is presented in the form of a functional unit, where the unit refers to an ASIC (Application Specific Integrated Circuit) circuit, a processor and memory that executes one or more software or fixed programs, and / or other devices that can provide the above functions.

[0175] The embodiment of the present invention also provides a computer device having the above Figure 6 The football match situation assessment system shown.

[0176] See also Figure 7 , Figure 7 is a schematic diagram of the structure of a computer device provided by an optional embodiment of the present invention, such as Figure 7 As shown, the computer device includes: one or more processors 10, a memory 20, and interfaces for connecting various components, including high-speed interfaces and low-speed interfaces. Various components are connected to each other using different buses for communication, and can be installed on a common mainboard or installed in other ways as needed. The processor can process the instructions executed in the computer device, including instructions stored in or on the memory to display the graphical information of the GUI on an external input / output device (such as, a display device coupled to the interface). In some optional embodiments, if necessary, multiple processors and / or multiple buses can be used together with multiple memories and multiple memories. Similarly, multiple computer devices can be connected, and each device provides some necessary operations (for example, as a server array, a group of blade servers, or a multi-processor system). Figure 7 A processor 10 is taken as an example.

[0177] The processor 10 may be a central processing unit, a network processor or a combination thereof. The processor 10 may further include a hardware chip. The hardware chip may be a dedicated integrated circuit, a programmable logic device or a combination thereof. The programmable logic device may be a complex programmable logic device, a field programmable gate array, a general purpose array logic or any combination thereof.

[0178] The memory 20 stores instructions executable by at least one processor 10, so that at least one processor 10 executes the method shown in the above embodiment.

[0179] The memory 20 may include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function; the data storage area may store data created according to the use of the computer device, etc. In addition, the memory 20 may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some optional embodiments, the memory 20 may optionally include a memory remotely arranged relative to the processor 10, and these remote memories may be connected to the computer device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0180] The memory 20 may include a volatile memory, such as a random access memory; the memory may also include a non-volatile memory, such as a flash memory, a hard disk or a solid state drive; the memory 20 may also include a combination of the above types of memory.

[0181] The computer device further comprises a communication interface 30 for the computer device to communicate with other devices or a communication network.

[0182] The embodiment of the present invention also provides a computer-readable storage medium. The method according to the embodiment of the present invention can be implemented in hardware, firmware, or can be implemented as a computer code that can be recorded in a storage medium, or can be implemented as a computer code that is originally stored in a remote storage medium or a non-temporary machine-readable storage medium and will be stored in a local storage medium through a network download, so that the method described herein can be stored in such software processing on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. Among them, the storage medium can be a magnetic disk, an optical disk, a read-only storage memory, a random access memory, a flash memory, a hard disk or a solid-state hard disk, etc.; further, the storage medium can also include a combination of the above types of memories. It can be understood that a computer, a processor, a microprocessor controller, or programmable hardware includes a storage component that can store or receive software or computer code. When the software or computer code is accessed and executed by a computer, a processor, or hardware, the method shown in the above embodiment is implemented.

[0183] A part of the present invention may be applied as a computer program product, such as a computer program instruction, which, when executed by a computer, can call or provide the method and / or technical solution according to the present invention through the operation of the computer. Those skilled in the art should understand that the existence of the computer program instruction in a computer-readable medium includes, but is not limited to, a source file, an executable file, an installation package file, etc., and accordingly, the way in which the computer program instruction is executed by the computer includes, but is not limited to: the computer directly executes the instruction, or the computer compiles the instruction and then executes the corresponding compiled program, or the computer reads and executes the instruction, or the computer reads and installs the instruction and then executes the corresponding installed program. Here, the computer-readable medium may be any available computer-readable storage medium or communication medium accessible to the computer.

[0184] Although the embodiments of the present invention have been described in conjunction with the accompanying drawings, those skilled in the art may make various modifications and variations without departing from the spirit and scope of the present invention, and such modifications and variations are all within the defined scope.

Claims

1. A football match situation assessment method, characterized in that: The method comprises: Get tracking and event data for any moment in a football match; Performing feature extraction processing on the tracking data and the event data to obtain first feature data after the tracking data processing and second feature data after the event data processing respectively; Time-aligning the first feature data and the second feature data to obtain time-aligned target fusion data; Analyzing the defensive style, offensive phase and threat level of the game team on the target fusion data to obtain defensive style information, offensive phase information and threat level information; wherein the target fusion data represents a data set obtained by aligning the first feature data and the second feature data on the same timeline; Determine a football match analysis prompt word according to the target fusion data, the defensive style information, the offensive phase information, and the threat level information; The football match analysis prompt words are input into a large language model trained based on big data of real football matches to obtain a football match situation assessment result.

2. The method according to claim 1, characterized in that The first characteristic data at least includes ball position information, team information, defense team identification information, player identification information, player position information, player speed information, court coordinate information, match number information, match remaining time information, current ball-controlling player identification information, and current ball-controlling team information; The second characteristic data at least includes information about the team where the event occurs, identity information of players in the team where the event occurs, event type information, event result information, starting position information where the event occurs, and ending position information where the event occurs.

3. The method according to claim 2, characterized in that The target fusion data is subjected to a defensive style analysis of the game team to obtain defensive style information, specifically including: Based on the player's position information, team information, stadium coordinate information, ball position information, current ball-controlling player identification information, current ball-controlling team information, event type information, event start position information, and event end position information, the offensive team and the defensive team are clustered to obtain the position information of each offensive line and each defensive line in the stadium; According to the position information of each offensive line and the position information of each defensive line, the arrangement and combination information between the offensive line and the defensive line is determined; The defensive style information of the game team is determined according to the permutation and combination information.

4. The method according to claim 3, characterized in that The target fusion data is analyzed for the offensive phase of the game team to obtain offensive phase information, specifically including: Determine first relative position information between the player and the ball according to the player's position information, the court's coordinate information, and the ball's position information; Determine the type of the offensive line controlling the ball according to the position information of each offensive line, the starting position information of the event, the ending position information of the event, and the first relative position information; Determines offensive phase information based on the type of offensive line in control of the ball.

5. The method according to claim 2, characterized in that: The target fusion data is subjected to a threat level analysis of the game team to obtain threat level information, specifically including: Divide the court area for threat analysis based on the court's coordinate information; Determine second relative position information between the ball and the court according to the event type information, the position information of the ball and the court area; The threat level information of the playing team is determined according to the second relative position information.

6. The method according to any one of claims 1 to 5, characterized in that: The determining of football match analysis prompt words according to the target fusion data, the defensive style information, the offensive phase information and the threat level information specifically includes: A data set of first feature data and second feature data aligned on the same timeline and corresponding defensive style information, offensive phase information and threat level information are input into a preset prompt word template to obtain football match analysis prompt words; wherein the preset prompt word template represents a question template obtained by performing semantic feature analysis based on football match analysis data.

7. A football match situation assessment system, characterized in that: The system comprises: A data acquisition module is used to obtain tracking data and event data at any moment in a football match; A feature extraction module, used to perform feature extraction processing on the tracking data and the event data, and obtain first feature data after the tracking data is processed and second feature data after the event data is processed; A time alignment module, used for performing time alignment on the first feature data and the second feature data to obtain time-aligned target fusion data; an analysis module, configured to analyze the target fusion data in terms of the defensive style, offensive phase, and threat level of the game team, and obtain defensive style information, offensive phase information, and threat level information; wherein the target fusion data represents a data set obtained by aligning the first feature data and the second feature data on the same timeline; A prompt word generation module, used for determining a football match analysis prompt word according to the target fusion data, the defensive style information, the offensive phase information and the threat level information; The evaluation result module is used to input the football match analysis prompt words into a large language model obtained by training based on big data of real football matches to obtain a football match situation evaluation result.

8. A computer device, characterized in that: include: A memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the football match situation assessment method according to any one of claims 1 to 6 by executing the computer instructions.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a computer to execute the football match situation assessment method according to any one of claims 1-6.

10. A computer program product, characterized in that The method comprises computer instructions, wherein the computer instructions are used to enable a computer to execute the football match situation assessment method according to any one of claims 1 to 6.

Images