Augmented reality-based football tactical rehearsal auxiliary training method, system and device

Abstract

The present application relates to the field of AR football training, and particularly relates to the field of a football tactical exercise auxiliary training method, system and equipment based on augmented reality, and the football tactical exercise auxiliary training method based on augmented reality has the characteristics that the method comprises the following steps: acquiring real-time images of a football training ground, player dynamic data and ground environment parameters, and constructing a ground player joint working condition characteristic vector; based on the working condition characteristic vector, performing scene clustering analysis on historical tactical exercise data, and establishing a mapping relationship between training scene characteristics and tactical execution parameters. The method significantly improves the tactical information perception efficiency, reduces the understanding deviation, and overcomes the problems of lack of smoothing processing and unstable dynamic display in the prior art.

Description

Technical Field

[0001] This invention relates to the field of AR football training, and in particular to a method, system and device for assisting football tactical drills based on augmented reality. Background Technology

[0002] In modern football tactical training, augmented reality (AR) technology, with its ability to blend virtual and real elements, can overlay pre-set tactical markers such as running arrows, receiving areas, and passing routes onto the players' AR glasses in real time. This intuitively conveys tactical intentions and supports collaborative drills of complex offensive and defensive tactics. While current AR football tactical training solutions have achieved the initial display of overlaid tactical markers, existing technology still has significant shortcomings in real-world multi-player, highly dynamic tactical training scenarios, making it difficult to meet the high-intensity, high-dynamic, and highly collaborative training demands of football.

[0003] First, existing AR tactical display solutions typically rely solely on the fixed field of view of the AR glasses for visibility assessment, failing to consider the spatial angular relationship between the player's real-time gaze direction and the AR tactical markings, and also lacking a viewing depth comfort evaluation mechanism. During football tactical drills, players continuously run, frequently turn their heads, and their gaze directions dynamically change. Relying solely on a fixed field of view for visibility assessment can easily lead to key tactical markings such as running arrows and receiving areas being within the field of view but far from the player's center of attention or at an uncomfortable viewing depth. This results in low efficiency in tactical information perception, significant comprehension errors, and an inability to effectively support the accurate transmission of the coach's intentions.

[0004] Secondly, existing technologies generally ignore the impact of players' own body posture on tactical markers, failing to utilize body posture information for real-time occlusion assessment. In football training, players frequently perform actions such as turning, swinging their arms, and bending over, making their arms, torso, and other body parts highly susceptible to obstructing their line of sight. Existing solutions cannot dynamically identify such occlusion states, still classifying obscured AR markers as visible, resulting in distorted visibility assessments. Key passing routes and positioning cues are obscured by the body and cannot be obtained by players in a timely manner, severely impacting the accuracy of tactical execution.

[0005] Furthermore, existing technologies often employ independent adjustment strategies when deploying multiple AR tactical markers, lacking a collaborative optimization mechanism that addresses visual overlap between markers, size readability, and the preservation of the coach's original tactical layout. Complex football tactical drills require the simultaneous display of multiple markers such as running arrows, receiving areas, and passing routes. Existing solutions only adjust the parameters of individual markers, which easily leads to problems such as markers overlapping, core tactical information being obscured, and marker positions deviating significantly from the coach's preset tactical anchoring areas. This disrupts the integrity of the tactical layout, making it difficult for players to accurately understand the overall offensive and defensive logic.

[0006] Finally, existing technologies, when updating AR marker parameters frame by frame, often directly output the optimized results without a smooth transition mechanism. Football tactical drills are a continuous dynamic process, with players' spatial positions and head postures changing in real time. Existing solutions directly switch the display position, size, and transparency of the markers, which can easily lead to phenomena such as sudden changes in the position, size, and brightness of the AR markers in the field of vision. This generates strong visual interference, significantly increases the cognitive load on players, and affects training focus and tactical execution efficiency. Summary of the Invention

[0007] The purpose of this invention is to solve the problems mentioned in the background section.

[0008] To achieve the above objectives, the present invention provides the following technical solution:

[0009] An augmented reality-based method for assisting in football tactical drills, the method comprising:

[0010] Acquire real-time images of football training fields, player dynamic data, and field environmental parameters to construct a joint field and player working condition feature vector;

[0011] Based on the aforementioned working condition feature vector, scenario clustering analysis is performed on historical tactical exercise data to establish a mapping relationship between training scenario features and tactical execution parameters.

[0012] A football tactical training parameter optimization model is constructed. Under the constraints of field space, player mobility boundary and tactical logic consistency, the virtual tactical markings, running trajectory parameters and interactive prompt parameters are optimized in multiple objectives to obtain the optimal tactical auxiliary parameter vector.

[0013] Establish a scenario tactical parameter matching library, train scenario recognition based on the current field situation feature vector of the earth workers, and retrieve the corresponding optimal tactical auxiliary parameter vector from the matching library;

[0014] Based on the optimal tactical auxiliary parameter vector, real-time tactical visualization information that integrates virtual and real elements is generated to dynamically guide player positioning, offensive and defensive stances, and coordination, thereby completing football tactical collaborative drills.

[0015] Furthermore, the construction of the joint field and player condition feature vector includes:

[0016] The system simultaneously collects field images, player 3D coordinates, movement speed, posture angle, field size, turf flatness, and light intensity through a high-definition field camera unit, player positioning sensor unit, and environmental perception unit.

[0017] Multi-source data are time-aligned and spatially registered according to a fixed sampling frequency to form data on the field player's running status at the same moment;

[0018] Based on the operational status data, a condition feature vector is constructed to characterize the current training scenario and player status.

[0019] Furthermore, the scene clustering analysis includes:

[0020] The collected historical tactical exercise data was processed by outlier removal, coordinate normalization, and scene labeling.

[0021] Using the field staff condition feature vector as input, the K-means clustering algorithm is used to cluster historical data, and the number of scene clusters is determined by the elbow rule;

[0022] Multiple scenario clusters corresponding to typical football training scenarios are obtained, with each scenario cluster corresponding to positional attack, wing breakthrough, central penetration, zonal defense, or offensive-defensive transition scenarios.

[0023] For each scene cluster, calculate its central feature vector and distribution matrix to characterize the combined features of the field environment and player movement in that training scene;

[0024] Based on the distance relationship between the current working condition feature vector and each scene cluster, the degree of membership of the current training scene to each scene cluster is calculated.

[0025] Furthermore, the optimal tactical auxiliary parameter vector includes: virtual tactical line coordinates, running trajectory curve parameters, offensive and defensive positioning reference points, interactive prompt trigger threshold, virtual-real fusion transparency, tactical switching response time, player action matching weight, and error correction compensation coefficient.

[0026] Furthermore, the constraints of the parameter optimization model include: field space boundary constraints, player movement speed and acceleration constraints, tactical action continuity constraints, player physical exertion constraints, and virtual-real fusion display clarity constraints.

[0027] Furthermore, the multi-objective optimization objectives include: tactical execution accuracy, player positioning accuracy, smoothness of offensive and defensive coordination, interactive response speed during training, and standardization of player action execution.

[0028] Furthermore, the parameter optimization model is solved using the Grey Wolf optimization algorithm, including:

[0029] During the population initialization phase, an initial population is generated based on the distribution characteristics of tactical auxiliary parameters under different training scenarios.

[0030] During the iteration process, the global search for tactical auxiliary parameters is achieved by updating the individual positions of gray wolves;

[0031] During the search process, the constraints of the field space, the player's athletic ability, and tactical logic are combined to correct individuals that do not meet the conditions;

[0032] The optimal tactical auxiliary parameter vector that satisfies the multi-objective optimization requirements is obtained through iterative convergence.

[0033] Furthermore, the construction and use of the scenario tactical parameter matching library includes:

[0034] The optimal tactical auxiliary parameter vectors corresponding to each training scenario cluster are classified and stored to form a mapping relationship between scenario features and tactical parameters;

[0035] During football tactical drills, the membership degree of the current working condition feature vector to each scene cluster is calculated in real time.

[0036] Select the tactical auxiliary parameter vector corresponding to the scene cluster with the highest degree of membership as the control parameter of the current training scene;

[0037] The control parameters are input into the augmented reality rendering unit to generate and dynamically adjust virtual tactical information, guiding players to complete tactical actions.

[0038] An augmented reality-based football tactical drill auxiliary training system includes a data acquisition module for acquiring real-time images of the football training field, player dynamic data, and field environmental parameters, and constructing a joint field and player working condition feature vector.

[0039] The scene recognition module is used to cluster historical tactical exercise data into scenes, establish a mapping relationship between scene features and tactical parameters, and identify the current training scene.

[0040] The parameter optimization module is used to build a tactical parameter optimization model, perform multi-objective optimization under multiple constraints, and generate the optimal tactical auxiliary parameter vector.

[0041] The parameter matching module is used to build a scenario tactical parameter matching library and retrieve the optimal tactical auxiliary parameters that are suitable for the current scenario.

[0042] The virtual-real interaction module is used to generate virtual-real tactical visualization information based on optimal parameters, and guide players' running, offensive and defensive positioning and coordination in real time to complete tactical drills.

[0043] An electronic device,

[0044] It includes a processor and a memory, the memory storing a computer program, and the processor executing the computer program to implement an augmented reality-based football tactical drill-assisted training method.

[0045] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0046] (1) This invention incorporates the player's real-time gaze direction and depth comfort into the perception evaluation, constructing a visual sensitivity model that integrates angle deviation and depth deviation. This allows the visibility judgment of AR tactical markers to no longer be limited to a fixed field of view, but to dynamically match the player's current attention center with the comfortable viewing distance. In the high-intensity running and frequent head-turning tactical drills of football, key tactical information such as running arrows, receiving areas, and passing routes can continuously fall within the player's visually sensitive area, significantly improving the efficiency of tactical information perception, reducing comprehension bias, and ensuring that the coach's intentions are received quickly and accurately. This effectively solves the problem of existing technologies relying solely on fixed field of view judgments and having poor perception matching.

[0047] (2) This invention introduces a body occlusion judgment mechanism based on human posture information. By constructing a limb capsule and intersecting it with the line of sight, the occlusion state is quantified into an occlusion coefficient, which is then multiplied by visual sensitivity to obtain the perception effectiveness, strictly distinguishing between "theoretically visible" and "actually visible". In football tactical drills, players frequently turn, swing their arms, and bend over, and their arms and torso can easily obstruct their vision. This invention can identify occlusion in real time and reduce the perception weight of occluded markers, avoiding misjudging AR tactical markers within the field of vision but obstructed by the body as perceptible. This ensures that key running prompts and passing route information are effectively presented only when they are actually visible, significantly improving the accuracy of tactical execution and overcoming the shortcomings of existing technologies that ignore body occlusion and have distorted visibility evaluation.

[0048] (3) This invention addresses the need for collaborative display of multiple identifiers and multiple players by designing a particle swarm collaborative layout optimization method that incorporates a gaze guidance mechanism. The fitness function simultaneously aggregates comprehensive perceptual benefits, size readability, visual overlap penalties, tactical deviation penalties, and perceptual infeasibility penalties, and adds a local perceptual repair step. In complex football tactical drills, multiple tactical identifiers such as running arrows, receiving areas, and passing routes need to be displayed simultaneously. This invention can achieve multi-identifier collaborative optimization, actively moving towards the player's gaze center, suppressing identifier stacking and coverage, while strictly maintaining the coach's original tactical layout. This prevents core tactical information from being submerged or deviating from the tactical anchoring area, ensuring that the overall offensive and defensive logic is clear and easy to understand. This effectively solves the problems of independent adjustment of multiple identifiers, chaotic layout, and deviation from tactical intent in existing technologies.

[0049] (4) In the online rendering stage of this invention, a smooth transition strategy between adjacent frames is adopted to exponentially smooth the position, opacity, and size of the optimized output markers, significantly reducing marker jitter and abrupt changes between consecutive frames. Football tactical drills are a highly dynamic and continuous process, with players' spatial positions and head postures changing in real time. This invention enables AR tactical markers to remain stable and change smoothly in the player's field of vision, avoiding visual interference such as position jumps, sudden changes in size, and sudden changes in brightness, reducing the cognitive load on players, improving training focus and the quality of tactical information reception, and overcoming the shortcomings of existing technologies such as lack of smooth processing and unstable dynamic display. Attached Figure Description

[0050] Figure 1 This is a flowchart of the method of the present invention;

[0051] Figure 2 This is a training-limited, multi-source state data acquisition diagram for the present invention;

[0052] Figure 3 This is a modeling diagram of visual sensitivity and depth comfort in this invention;

[0053] Figure 4 This is a diagram illustrating the effectiveness of body occlusion judgment and perception in this invention.

[0054] Figure 5 This is a schematic diagram of the gaze guidance vector generation of the present invention;

[0055] Figure 6 This is a rendering optimization and online smoothing output diagram for the present invention. Detailed Implementation

[0056] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0057] This invention provides a technical solution:

[0058] S1, AR Tactical Training Multi-Source Status Data Acquisition and Display Task Definition

[0059] In AR football tactical training, coaches use tablet terminals to deploy AR tactical markers such as running directions, receiving areas, passing routes, and attention prompts. These AR tactical markers need to be simultaneously overlaid onto the AR glasses worn by multiple players. This invention first collects information on player poses, gaze directions, body postures, and the AR tactical marker tasks deployed by the coach. This information forms the input foundation for subsequent perception effectiveness calculations and collaborative layout optimization. The specific steps are as follows:

[0060] S101, Multi-source status data acquisition at the training site

[0061] 1> Let the total number of players participating in AR football tactical training and wearing AR glasses be . , dimensionless, the first One player satisfies Let the total number of AR tactical markers that need to be displayed and optimized be... , dimensionless, the first An AR tactical identifier satisfies .

[0062] 2> The coach sets the original anchored world coordinates for each AR tactical marker via a tablet terminal. , This indicates the pre-specified display position in the world field coordinate system according to the coach's tactical intentions, used to represent the first... The initial spatial location where the AR tactical signage was intended to appear; It is a dimension of The column vector is in meters.

[0063] In the subsequent perception evaluation and layout optimization process, the first The candidate anchor world coordinates actually used in the calculation of each AR tactical identifier are as follows: , Indicates the first The current temporary spatial location of each AR tactical marker; before layout optimization is implemented. .

[0064] As an example, the coach can drag and drop a "forward arrow" icon pointing towards the opponent's goal onto the world field coordinate system on a tablet terminal. Location, then the identifier Meters; its candidate anchor world coordinates before optimization begins. It is also equal to rice.

[0065] 3> is the first Each AR tactical identifier identifies the associated player set. , Used to indicate that the first [item] requires actual attention. The range of players identified by AR tactical markers is dimensionless.

[0066] If the first An AR tactical icon is used to indicate when a winger is making a run towards the byline, and then the associated player set... It may include only that winger; if the first An AR tactical icon is used to indicate when three players in the midfield and forward lines are about to complete a triangular passing sequence; the associated player set... It can include all 3 players simultaneously; through associated player sets Subsequent layout optimizations do not require all players to simultaneously view all AR tactical markers, thereby reducing the interference of irrelevant markers on the vision layout.

[0067] 4> Obtain the first Spatial position of the center of each player's eyes in the world field coordinate system , Used to characterize the The approximate imaging center or binocular observation center of each player's AR glasses. It is a dimension of The column vector is in meters.

[0068] 5> Obtain the transformation from the world site coordinate system to the first... Rotation matrix of player's vision coordinate system , Used to convert the orientation and positional relationships in the world field coordinate system into a view representation with the player's current head orientation, it is a dimension-10 ... The rotation matrix.

[0069] In practical implementation, if the AR glasses directly output a pose matrix transformed from the player's field of view coordinate system to the world field coordinate system, then the rotation matrix can be obtained by transposing the rotated part of the pose matrix. .

[0070] 6> Get the first The current gaze direction angle vector of each player ,Right now ;in, Used to characterize the The angular position of a player’s current center of attention in the player’s field of vision coordinate system, in degrees; Indicates the first The horizontal angle of a player's current gaze direction relative to the center of their field of vision, in degrees; Indicates the first The vertical angle of a player's current gaze direction relative to the center of their field of vision, in degrees; This represents the vector transpose operation.

[0071] It should be noted that when a player's eyes turn horizontally to the right relative to the front of their head, Take a positive value; when rotating horizontally to the left, Take a negative value; when the player's eyes move vertically upwards relative to the front of their head. Take a positive value; when rotating vertically downwards, Take the negative value. For example, if the player's eyes are looking straight ahead, then... If the player's eyes turn to the upper left and rotate horizontally... Vertical rotation ,but .

[0072] The output is directly from the eye-tracking module of the AR glasses; when the confidence level of eye tracking is lower than the preset threshold, the angle vector corresponding to the direction directly in front of the player's head can be used as the alternative gaze direction angle vector.

[0073] 7> Get the first The player's human posture information is used to construct the body occlusion body in the subsequent process.

[0074] Human posture information can be obtained from at least one of the following: field camera, multi-view visual positioning device, wearable inertial sensor or AR glasses outward camera; human posture information includes the three-dimensional coordinates of key points such as head, shoulder, elbow, wrist, hip, knee and so on; subsequently, a simplified limb occlusion body can be constructed based on the skeletal connection relationship from shoulder to elbow, elbow to wrist and so on, to determine whether AR tactical markings are occluded by the player's own arm or torso.

[0075] 8> Establish a world field coordinate system based on the fixed location of the football field.

[0076] The center circle of the field can be set as the origin of the coordinate system, and the length, width and vertical directions of the field can be set as three fixed coordinate axes. After adopting a unified world field coordinate system, the coach tablet terminal, AR glasses positioning module, AR tactical sign layout optimization module and AR rendering module can share a consistent spatial reference benchmark, avoiding coordinate definition conflicts between the tactical anchor position of AR tactical signs and the actual rendering position.

[0077] The analysis of multi-source state data collected during training consists of two sub-figures, left and right. The left figure shows the AR tactical markings set up by the coach and the player positions, demonstrating the distribution of the AR tactical markings and the center positions of the players' eyes on the flat field. The horizontal axis represents the length of the field (in meters), and the vertical axis represents the width of the field (in meters). The origin of the coordinate system corresponds to the center circle of the field, reflecting the relative relationship between the original anchored world coordinates and the players' spatial positions. The experiment shows that the initial positions of different markings and different players vary, and some markings may be far from the center of the players' field of vision, requiring layout adjustments through subsequent perception evaluation. The right figure shows the angular deviation of AR icons in a player's field of vision. Five AR tactical icons were converted to the player's field of vision coordinate system, and the distribution of their horizontal and vertical deviation angles was displayed. The horizontal axis is the horizontal deviation angle (in degrees), the vertical axis is the vertical deviation angle (in degrees), the scatter color is the depth distance of each icon (in meters), and the green dashed box is the effective display field of vision of the player's AR glasses. The experiment shows that some icons may have exceeded the boundary of the effective field of vision or are located in an uncomfortable depth area, so they cannot be effectively perceived and need to be corrected by subsequent layout optimization.

[0078] S102, AR Tactical Marking Layout Variables and Permissible Movement Zone Definition

[0079] 1> For the first An AR tactical marker establishes a zone that allows movement. , Used to restrict the During the optimization process, each AR tactical marker is relative to its original anchored world coordinates. The movable range is limited to avoid optimizing results from deviating from the coach's original tactical intentions.

[0080] Permitted movement area It can be defined as being based on With the center of the ball, A spherical region with radius , where, Indicates the first The maximum spatial offset radius of an AR tactical marker, in meters; for AR tactical markers representing running path points. A distance of 0.3 to 0.5 meters is acceptable; for AR tactical markers used for area alerts, A length of 0.5 meters to 1.0 meters is acceptable.

[0081] 2> is the first Each AR tactical identifier setting displays an opacity parameter. , Used to control the The visual salience of an AR tactical symbol in AR glasses can be categorized into values ​​ranging from [value range missing]. ;when When AR tactical markings are displayed most clearly; when At that time, the AR tactical markings were displayed in a semi-transparent manner.

[0082] 3> is the first Each AR tactical identifier sets the display size scaling factor. , Used to control the The scaling factor of each AR tactical identifier relative to the default template size can take values ​​ranging from [value range missing]. ;when When AR tactical markings are displayed, they maintain the default display size; when... At that time, the AR tactical markings were magnified; when At that time, the AR tactical markings were reduced in size.

[0083] 4> Before starting layout optimization, the first step can be... Candidate anchored world coordinates for AR tactical markers Initialize to This will display the opacity parameter. Initializing to 1.0 will display the scaling factor. Initialize to 1.0; this initialization method allows the system to make limited, adaptive layout adjustments based on the coach's original layout and the players' current field of vision.

[0084] S2. Construction of AR tactical identification effectiveness based on dynamic field of view perception

[0085] Whether AR tactical markings are suitable for display depends not only on whether they are within the display area of ​​the AR glasses, but also on whether they are close to the player's current center of attention, whether they are at a relatively comfortable viewing depth, and whether they are obstructed by the player's own body. Simply using a fixed field of view as the basis for visibility judgment cannot reflect the real perception situation in AR football tactical training.

[0086] This invention transforms AR tactical markers into the player's field of vision coordinate system, and further combines field of vision constraints, gaze deviation, depth comfort, and body occlusion status to construct a perception effectiveness score, which is used to evaluate the player's overall perception quality of AR tactical markers. The specific steps are as follows:

[0087] S201, AR tactical marker player vision coordinate conversion

[0088] This step transforms the AR tactical markers from a unified global field coordinate system to each player's local field of vision coordinate system. This provides a coordinate basis for subsequent judgments on whether the markers are within the effective field of vision, how far they are from the center of gaze, and whether the depth is comfortable. The specific steps are as follows:

[0089] 1> For the first Each AR tactical identifier is anchored to world coordinates. Convert to the The player's vision coordinate system is used to obtain the vision coordinate vector. , can be represented as:

[0090] ;

[0091] in, Indicates the first The AR tactical identifier in the first Spatial position in a player's field of vision coordinate system It is a dimension of The column vector is in meters.

[0092] Will Expand as ;in, Indicates the first The AR tactical identifier is relative to the first The horizontal offset of the player's field of vision center is positive to the right, and the unit is meters; Indicates the first The AR tactical identifier is relative to the first The vertical offset of the player's visual field center, with upward being positive, is expressed in meters; Indicates the first The AR tactical sign along the first The depth distance in the direction directly in front of each player's face, in meters; This represents the vector transpose operation.

[0093] 2> Based on the field of view coordinate vector lateral component Longitudinal component and depth components Calculate the horizontal deviation angle and vertical deviation angle .

[0094] In specific implementation, when At that time, the judgment of the first The AR tactical sign is located at the... The area behind a player's face is not considered a valid display candidate location; therefore, the visual sensitivity for the combination of that player and the AR tactical icon is directly set to 0. At that time, continue calculating the first... The AR tactical identifier is relative to the first Horizontal deviation angle of the player's visual center and vertical deviation angle The calculation method is expressed as follows:

[0095] ;

[0096] ;

[0097] in, Used to characterize the The AR tactical identifier is relative to the first The horizontal angle deviation directly in front of each player's line of sight is expressed in degrees. Used to characterize the The AR tactical identifier is relative to the first The vertical angular deviation directly in front of a player's line of sight, in degrees; This represents the two-parameter arctangent function, used to calculate the azimuth angle of a coordinate point relative to the origin. The return value range is... .

[0098] Will and Combined into an angle deviation vector ,Right now ;in, Used to characterize the The AR tactical identifier in the first The position of each player in the spatial field of vision; This represents the vector transpose operation.

[0099] 3> Set the horizontal effective field of view half-angle according to the effective display field of view of the AR glasses. and vertical effective field of view half angle .

[0100] Specifically, This indicates the range of horizontal angles from the center of the field of view to the left / right boundary of the display area, in degrees, and can be between 20° and 30°. This indicates the vertical angle range from the center of the field of view to the upper / lower boundary of the display area, in degrees, ranging from 15° to 25°.

[0101] In specific implementation, when or At that time, the judgment of the first The AR tactical identifier exceeds the first The effective display range of the player's AR glasses is set to 0, and the subsequent visual sensitivity is directly set to 0.

[0102] As an example, let the horizontal effective field of view half-angle be... Vertical effective field of view half angle If the horizontal deviation angle of a certain AR tactical marker relative to a certain player is calculated... Vertical deviation angle ,because If the indicator is outside the effective horizontal display range of the player's AR glasses, then the visual sensitivity is determined to be... It is set directly to 0 and will not participate in subsequent calculations.

[0103] S202, Visual sensitivity calculation that integrates gaze deviation and depth comfort

[0104] For AR tactical markers located within the effective field of view, this step quantifies their angular proximity and depth matching with the player's current gaze center. Combined with real-time gaze deviation and gaze depth, the player's visual sensitivity to the AR tactical markers is constructed. The specific steps are as follows:

[0105] 1> When the first The AR tactical sign is located at the... When a player is effectively displayed within the field of view, calculate the angular deviation of that player relative to their current gaze direction. The calculation method is expressed as follows:

[0106] ;

[0107] in, Used to characterize the The AR tactical identifier and the first The angular distance between the current centers of attention of each player, in degrees; This represents the L2 norm operation of vectors.

[0108] 2> According to the first The depth of the currently viewed target is estimated by the current gaze ray of each player.

[0109] Specifically, taking the first The center of each player's eyes Starting from the ray origin, and using the spatial direction vector corresponding to the gaze direction as the ray direction, intersect it with the football field plane or the 3D model of the training scene to obtain the world coordinates of the gaze point, then transform it to the ... The player's field of vision coordinate system is used, and its depth component is taken as the current gaze depth. , Used to characterize the The depth of the spatial area that a player is currently focusing on from his face, in meters.

[0110] As an example, if a player is looking at an imaginary receiving point 5 meters in front of them, and the coordinates of that point are calculated by intersecting the court plane with the ray, then after transforming to the field of view coordinate system, its depth component is 4.8 meters. rice.

[0111] In practical implementation, if a stable gaze intersection cannot be obtained in the current frame, the most recent effective gaze depth can be used as the reference point. If an effective gaze depth cannot be obtained for multiple consecutive frames, then the device-comfortable display depth will be used. As an alternative depth This indicates the most comfortable viewing distance for AR glasses or human eyes for extended periods, expressed in meters, with 3.0 meters being a suitable value.

[0112] 3> To make the target depth more compatible with the device's display comfort range, adjustments can be made to... Perform interval truncation to obtain depth reference values. , This refers to the target depth used to evaluate the depth comfort of AR tactical marking, and the unit is meters.

[0113] The comfort depth range can be set to rice; when When rice is cooked, Set to 1.0 meter; when When rice is cooked, Set to 6.0 meters; otherwise, let .

[0114] 4> Based on the angular deviation Depth distance of the k-th AR tactical marker and depth reference value Construct the first The player against the first Visual sensitivity of AR tactical markings , The basic perceptual intensity, evaluated from the perspectives of visual attention and depth comfort, is calculated as follows:

[0115] ;

[0116] in, Used to characterize the The player against the first The basic visual perception intensity of an AR tactical identifier, dimensionless, ranges from [value range missing]. When the first An AR tactical sign is outside the effective display field of view, or located in the [missing information]. When a player's face is behind, it can make ; This represents the angle-sensitive bandwidth, used to control the rate at which visual sensitivity decreases as the gaze angle deviates further. The unit is °, and 8° can be used. This represents the depth-sensitive bandwidth, used to control the rate at which visual sensitivity decreases as depth deviation increases. The unit is meters, and it can be taken as 1.2 meters to 1.8 meters. This represents the natural exponential function.

[0117] It should be noted that the closer the AR tactical markers are to the player's real-time gaze center ( Smaller) and closer to the player's current depth of gaze ( near The higher the visual sensitivity, the better; the first index term As a gaze consistency factor, the closer the AR tactical marker is to the gaze center, the closer its weight is to 1; the farther it deviates, the faster its weight decays. Controlling the decay rate; the second exponential term As a depth comfort factor, the closer the depth of the AR tactical marker matches the player's current gaze depth, the closer its weight is to 1; the greater the depth difference, the more its weight decreases. Control the decay rate.

[0118] For example, if , If the first The AR tactical identifier and the first The angle deviation of the player's current gaze center ,depth Depth reference value Then visual sensitivity At a relatively high level; if the angle deviation of another candidate position is... ,depth Depth reference value Then visual sensitivity A significant decrease.

[0119] The analysis of visual sensitivity and depth comfort modeling consists of two subplots, left and right. The left plot is a surface plot of the visual sensitivity function, showing how visual sensitivity changes with angular and depth deviations; the horizontal axis represents angular deviation (in degrees), the vertical axis represents depth deviation (in meters), and the horizontal axis represents visual sensitivity (dimensionless). Experiments show that the smaller the angular deviation between the marker and the center of gaze, and the closer the depth is to the target depth, the higher the visual sensitivity; conversely, visual sensitivity rapidly decreases. The right plot is a matching plot of AR marker depth and target depth, showing the correspondence between the actual depth of an AR marker in a player's field of vision and the depth of the player's current target gaze; the horizontal axis represents AR marker depth (in meters), the vertical axis represents target depth (in meters), the scatter color represents the depth comfort factor (dimensionless), and the black dashed line represents the ideal matching line; experiments show that when the marker depth is close to the depth the player is currently gazing at, the depth comfort factor is close to 1, and the marker is more easily perceived comfortably; conversely, the depth comfort factor rapidly decreases.

[0120] S203, Construction of Body Occlusion Detection and Perceptual Effectiveness

[0121] This step uses the player's body posture information to determine whether the AR tactical markings are obscured by the player's own body, and combines the obscuration result with visual sensitivity to form the final perception effectiveness, in order to distinguish between "theoretically visible" and "actually visible". The specific steps are as follows:

[0122] 1> According to the first The player's human posture information is used to construct a simplified body occlusion volume, which is used to approximate the main body parts that may obstruct the view.

[0123] The torso can be represented by an ellipsoid or a cuboid; the upper arm and forearm can be represented by a capsule; the capsule is composed of line segments formed by key points at both ends of the corresponding limb and a preset radius; the radius of the upper arm and forearm capsule can be between 0.05 meters and 0.10 meters, and the size of the torso can be adaptively set according to the player's body proportions.

[0124] 2> Determine the occlusion status based on the intersection test between the line of sight ray and the simplified body occlusion object.

[0125] Specifically, taking the first The center of each player's eyes As the starting point of the ray, with the first Candidate anchored world coordinates for AR tactical markers Construct a line-of-sight ray to represent the endpoint of the ray, and determine the location of the line-of-sight ray upon arrival. Whether it previously intersected with any simplified body-covering object.

[0126] In the specific implementation, the first The AR tactical identifier is relative to the first The occlusion coefficient of each player is denoted as , Used to characterize the The AR tactical identifier is relative to the first The occlusion state of a player is dimensionless and takes only 0 or 1. If the line of sight intersects with any simplified body occlusion object, the occlusion coefficient is 0; if no intersection occurs, the occlusion coefficient is 1.

[0127] 3> Visual sensitivity With occlusion coefficient Multiply to obtain perceived validity. ,Right now ;in, Used to characterize the The player against the first The overall perceived quality of an AR tactical signage, including visibility, gaze adaptation, depth comfort, and unobstructedness, is dimensionless and ranges from [value range missing]. ; The larger the value, the more suitable the AR tactical icon is to be shown to that player; The smaller the value, the more difficult it is for the player to effectively perceive the AR tactical marker.

[0128] As an example, if an arrow-shaped AR tactical marker is located near the center of a winger's field of vision, with high visual sensitivity, but the winger raises their hand to signal, and the forearm capsule intersects with the line of sight, then the occlusion coefficient... Set to 0, final perceived effectiveness It is also set to 0, which prevents the system from mistakenly assuming that "located in the center of the field of view" means "actually visible".

[0129] 4> Set the perception effectiveness threshold and check the perception compliance of associated players.

[0130] Specifically, set a perception effectiveness threshold. , Used to determine the first The AR tactical identifier for the first Do the associated players meet the minimum perceptible requirements? A value of 0.1 to 0.25 is acceptable.

[0131] In specific implementation, for the first Each associated player with an AR tactical icon Check whether all conditions are met. If all associated players meet this condition, then the current candidate's anchored world coordinates are determined. This can be considered perceptibly acceptable; if any associated player does not meet this condition, the current candidate is anchored to world coordinates. Adjustments are needed in subsequent collaborative layout optimizations.

[0132] As an example, setting The effectiveness of the indicator in a certain area for the perception of associated winger players. If the value is greater than the threshold, then the current layout is acceptable for that player; however, if the perceived effectiveness is greater than the threshold for the associated forward player... If the value is less than the threshold, the current layout does not meet the requirements, and the position of the identifier needs to be optimized and adjusted.

[0133] It should be noted that the occlusion judgment enables the present invention to distinguish between the two types of situations: "theoretically within the field of vision" and "actually visible," which is especially applicable to the dynamic environment in football training where players frequently swing their arms, turn around, and make directional gestures.

[0134] The analysis of body occlusion judgment and perception effectiveness diagram consists of two sub-graphs, left and right. The left graph is a schematic diagram of the intersection detection between the body occlusion object and the line of sight ray, showing the player's simplified body occlusion object and the line of sight ray pointing from the center of the eyes to the AR tactical mark; the horizontal axis is the forward distance (in meters), and the vertical axis is the height (in meters). The red dashed line represents the line of sight ray, and the orange "×" represents the intersection of the line of sight ray and the forearm capsule, at which point the occlusion coefficient is judged to be 0; the experiment shows that even if the mark is within the field of vision, the player's own body may obstruct the line of sight, resulting in a perception effectiveness of 0. The right figure is a bar chart showing the perceived effectiveness of each AR tactical icon for associated players, comparing the perceived effectiveness of four AR tactical icons for three associated players; the horizontal axis represents AR tactical icons (dimensionless), the vertical axis represents perceived effectiveness (dimensionless), and the red dashed line represents the perceived effectiveness threshold (taken as 0.2), showing the perceived differences of different icons on different players; the experiment shows that the perceived effectiveness of some icons for specific players is lower than the threshold (e.g., icon 2 for player C, icon 3 for player B), thus triggering the perceived infeasibility penalty in the layout optimization.

[0135] S3. Optimization of AR tactical signage collaborative layout based on field-of-view perception constraints

[0136] After obtaining the perception effectiveness of each AR tactical icon for the associated player, it is necessary to further consider the visual overlap, display size, display opacity, and degree of preservation of the original tactical layout between multiple AR tactical icons. If only individual AR tactical icons are adjusted independently, it is easy for multiple AR tactical icons to overlap in the field of vision of some players, or to destroy the overall tactical structure in order to improve the visibility of a single icon.

[0137] This invention employs a particle swarm cooperative layout optimization method that incorporates a gaze-guided mechanism to jointly search for candidate anchor world coordinates, display opacity parameters, and display size scaling factors of AR tactical markers. The specific steps are as follows:

[0138] S301. Particle Encoding and Cooperative Optimization Variable Definition

[0139] This step jointly encodes the spatial coordinates, display opacity, and display size scaling factor of each AR tactical marker into a high-dimensional vector of a particle, enabling the particle swarm algorithm to simultaneously search for the optimal combination of layout parameters for all markers. The specific steps are as follows:

[0140] 1> Let the total number of particles be... , No. Each particle satisfies , No. Each particle corresponds to a complete AR tactical marker layout scheme; total number of particles The value can be between 20 and 40. The initial particle velocity is taken as a random vector with a small amplitude to increase the diversity of the initial search.

[0141] 2> For the first The particle and the first Each AR tactical marker is set with particle candidate anchoring world coordinates. Particle display opacity parameter and particle display size scaling factor .

[0142] Specifically, Indicates the first The first particle gives the first Candidate anchor world coordinates for AR tactical markers It is a dimension of , a column vector, in meters; Indicates the first The first particle gives the first The display opacity parameter of an AR tactical sign is dimensionless. Indicates the first The first particle gives the first The display size scaling factor for each AR tactical sign is dimensionless.

[0143] In practical implementation, particle candidates are anchored to world coordinates. Must be located in the first The permitted movement area corresponding to each AR tactical marker. Internal; Particle display opacity parameter Limited to Within range; particle display size scaling factor Limited to Within the range.

[0144] As an example, for shared In a scene with AR tactical markers, in the layout scheme of the third particle, the position of the first marker... Meters, indicating opacity Display size scaling factor .

[0145] 3> The first The complete encoded vector of each particle Represented as ;

[0146] in, Indicates the first In the particle scheme, all variables to be optimized are concatenated into a vector according to their identifier indices. Each AR tactical identifier contains a display opacity parameter, three spatial coordinate components, and a display size scaling factor. It is a dimension of Column vectors; This represents the vector transpose operation.

[0147] In practical implementation, when evaluating the first... When considering the layout scheme of individual particles, the candidate anchored world coordinates in S2 Values And calculate the corresponding visual sensitivity based on the spatial location. and perceived effectiveness .

[0148] It should be noted that at least one reference particle is retained during particle swarm initialization; each AR tactical identifier in the reference particle uses the original anchored world coordinates. The display opacity parameter is set to 1.0, and the display size scaling factor is set to 1.0; this baseline particle is used to ensure that the optimization process can always be directly compared with the coach's original layout. For the remaining particles, movement is allowed within the permitted area. Internal random sampling of particle candidate anchoring world coordinates And random sampling within the specified range and .

[0149] S302, Construction of a fitness function that integrates perceptual quality, readability, visual overlap, and tactical preservation.

[0150] This step defines a unified fitness function to quantitatively evaluate the merits of the complete layout scheme encoded by each particle. This function aggregates multiple objectives such as perceptual effectiveness, size readability, visual overlap, and tactical deviation into a single scalar value. The specific steps are as follows:

[0151] 1> For the first Each AR tactical identifier in each particle calls S2 to calculate all associated players. Corresponding perceptual validity .

[0152] Furthermore, according to the first Display size scaling factor in individual particles Calling the projection matrix that matches the actual rendering of the AR glasses, the first... The display template of the AR tactical symbol is projected onto the first... The player's field of vision plane is used to obtain the projection bounding box, which is used to approximate the area occupied by the AR tactical mark in the player's field of vision plane.

[0153] 2> Calculate the shortest side pixel length of the projection bounding box , Used to characterize the The AR tactical identifier in the first The smallest recognizable size in a player's field of vision, in pixels.

[0154] First, based on the projection matrix of the AR glasses, The scaled identifier template vertices are transformed to clip space; then, through perspective division and viewport transformation, they are mapped to a screen coordinate system in pixels; finally, its 2D bounding box on the screen is calculated, and the pixel length of the shortest side of this bounding box is... .

[0155] As an example, an arrow icon with a default size of 100 pixels will change when the scaling factor... Furthermore, when the projection distance is relatively far, its bounding box on the viewing plane may be compressed into a strip with a height of 10 pixels and a width of 30 pixels. It is 10 pixels.

[0156] 3>According to Build size readability coefficient , This term describes the degree to which the recognition capability of AR tactical markings decreases due to their small projection size. It is dimensionless and its value ranges from [value range missing]. .

[0157] Minimum readable length can be set. Pixels, saturated readable length Pixel; when When pixels are used, let ;when When pixels are used, let Follow Linear increase, i.e. ;when When pixels are used, let .

[0158] 4> Set attention weight for the players associated with each AR tactical icon. , Used to indicate the first The related player for the first The importance of AR tactical markings and their satisfaction .

[0159] In practice, when an AR tactical icon is equally important to multiple players, the attention weight of each associated player can be made equal; when an AR tactical icon mainly serves a specific core player, a higher attention weight can be assigned to that player through a pre-set method.

[0160] As an example, a marker could be used to indicate that wingers and centers should make overlapping runs, with the wingers being the primary executors and the centers the supporters. In this case, a weight could be assigned to the wingers. The weight of the center .

[0161] 5> The number is calculated based on attention weight, display opacity, size readability coefficient, and perceived effectiveness. The comprehensive perceptual benefits of AR tactical identifiers in the particle scheme The calculation method is expressed as follows:

[0162] ;

[0163] in, Used for comprehensive characterization of the first The AR tactical identifier in the first Information transmission capability in individual particle layout schemes; The larger the value, the clearer, more visible, and easier for the relevant players to perceive the AR tactical marker.

[0164] It should be noted that by using a weighted summation method, the perceived benefits are aggregated to a certain identifier. For all of its target players The overall information delivery efficiency; This highlights the differences in importance among different players. and The final benefits were adjusted from the perspectives of display clarity and projection size, respectively. This is the foundation of perception; using a multiplicative form implies that if any link in the chain fails (for example, being occluded), it will cause problems. The entire benefit item will be reset to zero, thus forcing all associated players to be unobstructed and in a good state of perception, ensuring that tactical intentions can be fully received.

[0165] 6> Calculate the degree of visual overlap between different AR tactical symbols.

[0166] Specifically, for the first Calculate the intersection-union ratio (IoU) of any two AR tactical sign projection bounding boxes within the player's field of vision plane. , Used to characterize the The AR tactical identifier and the first The AR tactical identifier in the first The percentage overlap of projected images within each player's field of vision; among which... and Indicates the index of two different AR tactical identifiers; if the two projection bounding boxes do not overlap, then If the overlapping area is large, then The corresponding increase.

[0167] For example, if the projected areas of markers A and B overlap by 30% in the field of vision of a midfielder, then... If there is no overlap at all, then .

[0168] 7> The visual overlap penalty is calculated based on the degree of visual overlap and the display opacity parameter. , Used to quantify the total visual interference caused by overlapping signage projections in the entire layout scheme.

[0169] In practice, the calculation is performed by accumulating the product of the overlapping area of ​​the projections of any two different icons in the field of vision of all associated players and their opacity. The opacity parameter is multiplied by the intersection-union ratio to show that the visual severity of the overlap interference depends not only on the area overlap ratio, but also on the degree of opacity of the two icons themselves. The calculation method is expressed as follows:

[0170] ;

[0171] in, The smaller the value, the less visual overlap and interference there is in the layout scheme, and the clearer and more orderly the image. Indicates the first The first particle gives the first The display opacity parameter for each AR tactical identifier; Indicates the first The first particle gives the first The display opacity parameter for each AR tactical identifier.

[0172] For example, if two AR tactical icons overlap significantly in a player's field of vision, and the display opacity parameters of both AR tactical icons are close to 1.0, then the visual overlap penalty is large; if the opacity parameter of one of the AR tactical icons is reduced to 0.4, and its size still meets the size readability requirements, then the overlap penalty decreases.

[0173] 8> Construct a tactical deviation penalty term based on the normalized offset between the candidate anchored world coordinates and the original anchored world coordinates. , To limit the optimized layout from deviating excessively from the coach's original tactical configuration, the calculation method is expressed as follows:

[0174] ;

[0175] in, The smaller the value, the closer the current particle layout is to the coach's original settings; This represents the square of the second norm of a vector.

[0176] 9> Construct a perceptual infeasibility penalty term based on the difference between the perceived validity and the perceived validity threshold. , This is used to quantify the severity of all cases that do not meet the minimum perception requirements under the current layout scheme. The calculation method is expressed as follows:

[0177] ;

[0178] in, The larger the value, the more severe the perception deficiency problem in the current particle layout scheme; when the perception effectiveness... Less than the threshold When calculating the difference between the two, if the perceived validity is... All are greater than the threshold If the penalty term for this combination is 0, then the penalty term for this combination is 0. This represents the function that takes the maximum value.

[0179] In practical implementation, if any AR tactical identifier has an associated player that does not meet the requirements... When the corresponding insufficiency difference is accumulated, it serves as a penalty for perceived infeasibility; the larger the penalty for perceived infeasibility, the more or more serious the perceived insufficiency problem exists in the particle scheme.

[0180] As an example, suppose Under a certain particle scheme, the identifier is 1 pair of players 1. For player 2 ;Identifier 2 for player 3 The punishment for perceived infeasibility is .

[0181] 10> Based on the comprehensive perceptual benefits, visual overlap penalty, tactical deviation penalty, and perceptual infeasibility penalty, construct the first... Fitness function of individual particles , Used for comprehensive evaluation of the The overall merits of the complete layout scheme represented by each particle are calculated as follows:

[0182] ;

[0183] in, The smaller the value, the higher the value. The better the layout scheme for each particle; This represents the weight of the overall perceived benefit, which can be taken as... ; The weight representing the visual overlap penalty can be chosen. ; This indicates that the tactical deviation from the penalty weight is acceptable. ; This represents the perceived infeasibility penalty weight, which is acceptable. .

[0184] It should be noted that this fitness function transforms a multi-objective minimization problem (maximizing perceptual gains, minimizing visual overlap, minimizing tactical deviations, and eliminating perceptual infeasibility) into a single-objective minimization problem through a weighted sum. By seeking to minimize the fitness function value, it improves the overall perceptual gains of the identifier while minimizing negative issues such as overlapping interference, deviations from coach tactics, and the emergence of perceptual blind spots. Furthermore, the penalty weights can adjust the relative importance of different optimization objectives.

[0185] S303, Introducing gaze-guided particle velocity update

[0186] Standard particle swarm optimization primarily relies on the historical best position of particles and the global historical best position for searching, without directly utilizing real-time player gaze information. To make AR tactical icons more naturally close to the relevant player's current attention area, this invention introduces a gaze guidance term during particle update, giving the spatial layout update an attention-drawing effect. The specific steps are as follows:

[0187] 1> Let the first The particle in the first The position of the particle in the generation is The particle velocity is , The current iteration number of the particle swarm optimization is dimensionless; the th iteration number is... The historical best position of each particle is The global historical best position of the entire particle swarm is .

[0188] 2> Regarding the first For each player, the world coordinates of the gaze point are obtained based on the intersection of the gaze ray and the court plane or 3D training scene model. , Used to characterize the The player's current focus is on the spatial location of the field in the world coordinate system. It is a dimension of The column vector is in meters.

[0189] In practice, when the gaze ray does not have a stable intersection point, or the intersection point is outside the reasonable training area, the gaze point of the player in the current frame can be ignored and not included in the gaze guidance calculation.

[0190] 3> Regarding the first Each AR tactical icon is assigned based on the associated player's attention weight. Calculate the world coordinates of the gaze center The calculation method is expressed as follows:

[0191] ;

[0192] in, Used to characterize the The weighted center of attention for the player group associated with AR tactical icons. It is a dimension of The column vector is in meters.

[0193] 4> Regarding the first The first of the particles An AR tactical identifier, based on the gaze center of the associated group. With the current candidate position of the particle The difference between the constructs position gaze guide vector This causes particles to tend to move towards the center of the player's gaze, thus ensuring the marker falls within the player's area of ​​focus. The calculation method is as follows:

[0194] ;

[0195] in, Used to indicate the first The first of the particles Each AR tactical icon guides the direction in which the player's attention is drawn. It is a dimension of , a column vector, in meters; This indicates the gaze guidance step size, used to control the spatial update amplitude of each gaze traction, and is measured in meters, ranging from 0.05 meters to 0.15 meters. This represents a very small positive number, used to avoid a denominator of 0. Desirable ; This represents the L2 norm operation of vectors.

[0196] The diagram illustrating the generation of gaze guidance vectors demonstrates how to generate gaze guidance vectors using the gaze points of associated players. The horizontal axis represents the horizontal coordinates of the world field (in meters), and the vertical axis represents the vertical coordinates of the world field (in meters). Taking a certain marker as an example, the blue "×" indicates the gaze point of the associated player, the blue solid line represents its gaze direction, and the purple star represents the group gaze center obtained by weighting according to attention weights. The purple dashed line pointing from the original position of the marker to the gaze center is the gaze guidance direction. Experiments show that gaze guidance items can effectively guide the marker to the player's current attention concentration area.

[0197] 5> All Combined into a complete particle gaze guidance vector , Dimensions and Same, is the dimension of A column vector that is only anchored to world coordinates for each corresponding candidate identifier. The three spatial component dimensions take values ​​of Each component corresponds to a specific display opacity parameter. and display size scaling factor The value of 0 in the dimension can be represented as .

[0198] 6> Focus guidance item As an independent velocity component, the velocity update rule of the particle swarm is directly added, so that the spatial search of the particles is not only driven by the historical best solution, but also attracted by the player's current gaze center, thus ensuring that the optimized marker position falls within the player's attention area. The velocity update rule is expressed as:

[0199] ;

[0200] in, Indicates the first The particle in the first The velocity vector of the generation; This represents the inertia weight, used to balance the continuity of the original search direction and the convergence speed; it can be set to 0.72. This represents the individual cognitive coefficient, used to control the degree to which a particle moves closer to its historical best position; it can be set to 1.5. This represents the group cognition coefficient, used to control the degree to which particles move closer to the global historical best position; it can be set to 1.5. and Indicates the range of values ​​as The random variable is used to enhance search diversity.

[0201] It should be noted that the gaze guidance term, as a flexible traction direction in particle swarm search, works in conjunction with the individual historical best position and the group historical best position. Based on this, it can improve the response speed of the layout to the player's current attention state without causing the AR tactical marker to drift away from the original tactical area.

[0202] 7> The temporary layout scheme is obtained by updating the calculation based on the particle velocity, that is... ;in, Indicates the first A temporary layout scheme for each particle after completing velocity updates but before constraint repair is implemented.

[0203] 8> To avoid imbalance in the update amplitude of different types of parameters, particle velocities are classified and limited.

[0204] Specifically, the single update magnitude corresponding to the world coordinates of the particle candidate anchor can be limited to within 0.2 meters; the single update amplitude corresponding to the display opacity parameter can be limited to within 0.1; and the single update amplitude corresponding to the display size scaling factor can be limited to within 0.2.

[0205] S304. Termination of boundary repair, perception repair, and optimization of temporary layout scheme.

[0206] This step involves reasonably correcting the temporary solutions generated by the particle swarm optimization algorithm that may violate hard constraints, and introducing rule-based local fine-tuning to quickly fix perception failures such as local occlusion. This is a crucial step in ensuring the robustness and usability of the final solution. The specific steps are as follows:

[0207] 1> Perform hard constraint boundary repair on the temporary layout scheme.

[0208] Specifically, let's first discuss the temporary layout plan. The display opacity parameter in the data is used for boundary correction. When a certain display opacity parameter exceeds the specified limit... When that happens, truncate it to the corresponding boundary value; then adjust the temporary layout scheme. Boundary correction is performed on the display size scaling factor. When a certain display size scaling factor exceeds... When that happens, truncate it to the corresponding boundary value; then adjust the temporary layout scheme. In the process of anchoring particle candidates to world coordinates for allowed movement region repair, if the first The particle candidate for the AR tactical identifier is anchored to world coordinates outside the allowed movement area. Then along the original anchored world coordinates Point to the direction of this temporary location and project it back. boundary.

[0209] For example, if a particle's opacity parameter is updated to 1.3, it will be truncated to... If, after a particle is updated, the z-coordinate of an identifier in the temporary scheme is 0.7 meters, exceeding its... Center, radius The sphere of meters allows for the boundary of the region; during repair, it will be from... The vector pointing to this temporary location is normalized and multiplied by Add back This allows the position to be projected onto the sphere.

[0210] 2> Reassess the perceived effectiveness based on the revised solution.

[0211] Specifically, the perception effectiveness is recalculated for each AR tactical marker after boundary repair is completed. If the first Each AR tactical icon satisfies the requirements for all associated players. If the current particle candidate's world coordinates are not met, then the local perception repair is performed.

[0212] 3> Generate candidate locations for micro-movements in local sensory repair.

[0213] In the specific implementation, determine the first The player with the lowest current perception effectiveness identified by the AR tactical marker is designated as the target player to be repaired. Each AR tactical identifier, with the current particle candidate anchored at world coordinates, generates multiple micro-movement candidate positions within a local plane approximately perpendicular to the direction of the target player's line of sight.

[0214] Eight candidate micro-movement locations can be generated, distributed at equal angles. The micro-movement radius can be between 0.03 meters and 0.08 meters. For each candidate micro-movement location, it is necessary to recheck whether it is within the allowed movement area.

[0215] As an example, suppose the third AR tactical marker's perception effectiveness towards the second associated player (target repair player) is insufficient, and its current candidate anchor world coordinates are... The system uses this position as the center and generates eight candidate positions at 45° intervals within a local plane approximately perpendicular to the player's line of sight, with a radius of 0.05 meters. One of these candidate positions may be... All eight candidate locations, including this one, were checked to ensure they were all within the permitted movement area indicated by the sign. The boundary.

[0216] 4> Perform local sensing and optimal repair based on candidate micro-movement locations.

[0217] In the specific implementation, the perception effectiveness, projection overlap change, and tactical deviation change are calculated for each micro-movement candidate position, and priority is given to selecting those that satisfy all related players. The candidate positions for micro-movement are determined. If there are multiple candidate positions that meet the conditions, the candidate position that maximizes the minimum perceptual effectiveness and minimizes the newly added visual overlap is selected. If there are no candidate positions that fully meet the conditions, the candidate position that maximizes the improvement in perceptual effectiveness is selected.

[0218] As an example, the perceived validity of marker 3 for defensive midfielders is 0.08, which is below the threshold. The main reason is that it is too far from the player's center of gaze; the local repair generates 8 candidate positions on a circle with a radius of 0.05 meters around its current position for testing; it is found that after moving slightly to the right by 0.04 meters, the perception effectiveness increases to 0.15 and does not cause new overlap, so the slight movement position is accepted.

[0219] 5> Perform historical and global optimal updates for particles.

[0220] Specifically, the particle scheme after boundary repair and perception repair is recorded as the official particle position. And recalculate the fitness function. , Indicates the first The particle in the first The final fitness value after complete repair; if Better than the first The historical best fitness of a particle is then updated to the historical best position of the particle. ;like If the fitness is better than the global historical best, then update the global historical best position. .

[0221] 6> Set optimization termination conditions and output the optimal layout scheme.

[0222] Specifically, set the maximum number of iterations. , This represents the maximum number of iterations allowed by the algorithm, which can be 50; it sets the threshold for the globally optimal stagnation number of iterations. , This represents the maximum number of generations allowed to stagnate when the improvement in global optimal fitness is less than a threshold; a value of 10 can be used. The fitness improvement threshold is set. , This represents the minimum fitness improvement threshold used to determine whether optimization has stalled; it can be taken as... .

[0223] In practical implementation, when the number of iterations reaches... Or the globally optimal fitness is in a continuous The improvement within each generation was less than When the time is reached, terminate the optimization and output the global historical best position. It also decodes the optimized anchoring world coordinates, optimized display opacity parameters, and optimized display size scaling factors used for each AR tactical marker.

[0224] For example, after 45 iterations, the global optimal fitness decreased from the initial 0.52 to 0.18, and the improvement was less than 0.18 in the last 12 consecutive generations. If the termination condition is met, the system will terminate the optimization process. The system will then decode the globally optimal particle and obtain final parameters such as "passing arrow identifier: anchored world coordinates (10.15, 5.08, 1.62), display opacity 0.95, display size scaling factor 1.05".

[0225] It should be noted that this invention adds a local perception repair mechanism after the particle swarm search is completed, rather than simply passively accepting the insufficient perception state based on the global optimization result. This enables rapid remediation of dynamic problems such as local occlusion, boundary compression, and short-term field of view deviation, thereby improving the stability and practical usability of the optimization results.

[0226] S4, AR tactical marking online adaptive rendering application process

[0227] After completing the perception effectiveness modeling and collaborative layout optimization, this invention deploys the method into an AR football tactical training system, enabling the system to continuously update the AR tactical marker layout based on the real-time status of the training site. The specific steps are as follows:

[0228] 1> During each layout update cycle, collect the first... The world coordinates of the center of each player's eyes Rotation matrix gaze direction angle vector Human posture information and AR tactical marker tasks currently issued by the coach.

[0229] The layout update cycle can be set from 100 milliseconds to 500 milliseconds, and the specific value can be set according to the computing power of the AR glasses and the real-time requirements of the training system.

[0230] 2> Read the original anchored world coordinates of each AR tactical marker. Related player collection and permitted movement areas .

[0231] Execute S2 within the current update cycle to calculate the visual sensitivity corresponding to the location of each candidate AR tactical marker. and perceived effectiveness .

[0232] 3> Using the state data collected in the current cycle as input, perform collaborative layout optimization in S3, and output the optimized anchoring world coordinates, optimized display opacity parameters, and optimized display size scaling factor for each AR tactical marker.

[0233] 4> Optimize the anchoring of world coordinates for smooth transition of adjacent period outputs based on the smoothing coefficient.

[0234] Specifically, let the first The anchored world coordinates of each AR tactical marker obtained in the current cycle optimization are: The actual world coordinates for the previous rendering cycle were: The current period actually renders the anchored world coordinates. The calculation method is expressed as follows:

[0235] ;

[0236] in, This represents the display smoothness coefficient, used to control the balance between layout update response speed and visual stability, and can be set from 0.3 to 0.6.

[0237] It should be noted that optimizing the anchoring of world coordinates for adjacent output cycles to achieve a smooth transition can effectively improve the display stability between consecutive frames, avoid sudden jumps in AR tactical icons, and thus reduce cognitive interference and unnecessary attention distraction caused by icon jitter for players.

[0238] 5> For optimizing the display opacity parameter and optimizing the display size scaling factor, the same or similar smoothing methods are also used to make the AR tactical markings present more natural changes in visibility and size in consecutive frames.

[0239] For example, if the smoothing coefficient is 0.5, the actual rendering transparency of a certain identifier in the previous cycle was 0.8, and the transparency of the optimized output in the current cycle is 0.6, then the actual rendering transparency in the current cycle is... The identifier will smoothly transition from 0.8 to 0.7 in consecutive frames, rather than jumping to 0.6, thus achieving a more natural visual change.

[0240] The analysis of rendering optimization and online smoothing output consists of two sub-plots, left and right. The left plot shows an example of overlapping bounding boxes in the player's field of vision, displaying the bounding boxes and overlapping areas of two AR tactical markers (A and B) on a player's field of vision. The horizontal axis represents the horizontal direction of the field of vision (in pixels), and the vertical axis represents the vertical direction (in pixels). Blue and orange rectangles represent the projection areas of the two markers, respectively, while the purple area represents the overlapping portion. Experiments show that when the projection areas of the two markers overlap significantly, it causes visual interference. The overlap penalty term in the fitness function will drive the optimization algorithm to adjust the position, opacity, or size to reduce overlap. The right plot shows the smooth transition effect curve between adjacent frames, displaying the optimized output position and the actual rendered position of an AR tactical marker after smoothing within several consecutive frames. The horizontal axis represents the frame number (dimensionless), and the vertical axis represents the coordinates along the length of the field (in meters). Red dotted lines represent the optimized output position, and blue solid squares represent the smoothed rendering position. Experiments show that smoothing can effectively filter out high-frequency jitter and abrupt changes in the optimized output, reducing visual interference caused by the movement of AR markers in consecutive frames.

[0241] 6> The smoothed render is anchored to world coordinates, and the display opacity parameter and display size scaling factor are sent to the AR rendering module to complete the adaptive overlay display of AR tactical markings in the AR glasses of the relevant players.

[0242] It should be noted that this invention, through the complete process of "dynamic vision perception modeling - multi-variable collaborative layout optimization - online smooth rendering output", enables AR tactical icons to adapt to complex training states such as player movement, head turning, gaze changes, and partial occlusion. This not only improves the real-time perceptibility of tactical information, but also reduces visual interference when multiple icons are superimposed, and tries to keep the coach's original tactical intentions intact.

[0243] As an example, in a fast break drill, the coach placed a "pass" arrow marker near the center circle, targeting winger A and full-back B. The system initially displayed the marker in the coach's designated position. As the drill progressed, winger A suddenly sprinted towards the sideline and turned to look for the ball behind him. At this moment, the system detected a sharp increase in the marker's deviation angle from winger A, resulting in decreased visual sensitivity. Simultaneously, as full-back B returned to defend, his raised arm capsule intersected with the marker's line of sight, reducing the obstruction coefficient to 0 and the perception effectiveness to 0. The optimization algorithm immediately activated, searching for a new position within the allowed movement area. A candidate point was found 0.3 meters in front and 0.2 meters above the original position, a point that could simultaneously avoid full-back B's obstruction and fall within the center of winger A's field of vision. Ultimately, the "pass" marker smoothly moved to the new position, slightly increasing its size and decreasing its opacity to adapt to the new background, successfully maintaining the effective transmission of tactical information amidst dynamic changes.

[0244] Finally, it should be noted that the above content is only used to illustrate the technical solution of the present invention, and is not intended to limit the scope of protection of the present invention. Simple modifications or equivalent substitutions made by those skilled in the art to the technical solution of the present invention do not depart from the essence and scope of the technical solution of the present invention.

Claims

1. A method for assisting in football tactical drills based on augmented reality, characterized in that, The method includes: Acquire real-time images of football training fields, player dynamic data, and field environmental parameters to construct a joint field and player working condition feature vector; Based on the aforementioned working condition feature vector, scenario clustering analysis is performed on historical tactical exercise data to establish a mapping relationship between training scenario features and tactical execution parameters. A football tactical training parameter optimization model is constructed. Under the constraints of field space, player mobility boundary and tactical logic consistency, the virtual tactical markings, running trajectory parameters and interactive prompt parameters are optimized in multiple objectives to obtain the optimal tactical auxiliary parameter vector. Establish a scenario tactical parameter matching library, train scenario recognition based on the current field situation feature vector of the earth workers, and retrieve the corresponding optimal tactical auxiliary parameter vector from the matching library; Based on the optimal tactical auxiliary parameter vector, real-time tactical visualization information that integrates virtual and real elements is generated to dynamically guide player positioning, offensive and defensive stances, and coordination, thereby completing football tactical collaborative drills.

2. The augmented reality-based football tactical drill auxiliary training method according to claim 1, characterized in that, The constructed joint field player performance feature vector includes: Through the high-definition camera unit, player positioning sensor unit and environmental perception unit, the system simultaneously collects field images, player three-dimensional coordinates, movement speed, posture angle, field size, turf flatness and light intensity. Multi-source data are time-aligned and spatially registered according to a fixed sampling frequency to form data on the field player's running status at the same moment; Based on the operational status data, a condition feature vector is constructed to characterize the current training scenario and player status.

3. The augmented reality-based football tactical drill auxiliary training method according to claim 1, characterized in that, The scene clustering analysis includes: The collected historical tactical exercise data was processed by outlier removal, coordinate normalization, and scene labeling. Using the field staff condition feature vector as input, the K-means clustering algorithm is used to cluster historical data, and the number of scene clusters is determined by the elbow rule; Multiple scenario clusters corresponding to typical football training scenarios are obtained, with each scenario cluster corresponding to positional attack, wing breakthrough, central penetration, zonal defense, or offensive-defensive transition scenarios. For each scene cluster, calculate its central feature vector and distribution matrix to characterize the combined features of the field environment and player movement in that training scene; Based on the distance relationship between the current working condition feature vector and each scene cluster, the degree of membership of the current training scene to each scene cluster is calculated.

4. The augmented reality-based football tactical drill auxiliary training method according to claim 1, characterized in that, The optimal tactical auxiliary parameter vector includes: virtual tactical line coordinates, running trajectory curve parameters, offensive and defensive positioning reference points, interactive prompt trigger threshold, virtual-real fusion transparency, tactical switching response time, player action matching weight, and error correction compensation coefficient.

5. The augmented reality-based football tactical drill auxiliary training method according to claim 1, characterized in that, The constraints of the parameter optimization model include: field space boundary constraints, player movement speed and acceleration constraints, tactical action continuity constraints, player physical exertion constraints, and virtual-real fusion display clarity constraints.

6. The augmented reality-based football tactical drill auxiliary training method according to claim 1, characterized in that, The multi-objective optimization objectives include: tactical execution accuracy, player positioning accuracy, smoothness of offensive and defensive coordination, interactive response speed during training, and standardization of player action execution.

7. The augmented reality-based football tactical drill auxiliary training method according to claim 1, characterized in that, The parameter optimization model is solved using the Grey Wolf optimization algorithm, including: During the population initialization phase, an initial population is generated based on the distribution characteristics of tactical auxiliary parameters under different training scenarios. During the iteration process, the global search for tactical auxiliary parameters is achieved by updating the individual positions of gray wolves; During the search process, the constraints of the field space, the player's athletic ability, and tactical logic are combined to correct individuals that do not meet the conditions; The optimal tactical auxiliary parameter vector that satisfies the multi-objective optimization requirements is obtained through iterative convergence.

8. A method for assisting in football tactical drills based on augmented reality, characterized in that, The construction and use of the scenario tactical parameter matching library includes: The optimal tactical auxiliary parameter vectors corresponding to each training scenario cluster are classified and stored to form a mapping relationship between scenario features and tactical parameters; During football tactical drills, the membership degree of the current working condition feature vector to each scene cluster is calculated in real time. Select the tactical auxiliary parameter vector corresponding to the scene cluster with the highest membership degree as the control parameter of the current training scene; The control parameters are input into the augmented reality rendering unit to generate and dynamically adjust virtual tactical information, guiding players to complete tactical actions.

9. A football tactical drill auxiliary training system based on augmented reality, characterized in that, It includes a data acquisition module, which is used to acquire real-time images of football training fields, player dynamic data and field environmental parameters, and construct a joint field and player working condition feature vector. The scene recognition module is used to cluster historical tactical exercise data into scenes, establish a mapping relationship between scene features and tactical parameters, and identify the current training scene. The parameter optimization module is used to build a tactical parameter optimization model, perform multi-objective optimization under multiple constraints, and generate the optimal tactical auxiliary parameter vector. The parameter matching module is used to build a scenario tactical parameter matching library and retrieve the optimal tactical auxiliary parameters that are suitable for the current scenario. The virtual-real interaction module is used to generate virtual-real tactical visualization information based on optimal parameters, and guide players' running, offensive and defensive positioning and coordination in real time to complete tactical drills.

10. An electronic device, characterized in that, It includes a processor and a memory, the memory storing a computer program, and the processor executing the computer program to implement the augmented reality-based football tactical drill auxiliary training method according to any one of claims 1 to 8.

Images