A player intention recognition and shot tracking method based on dynamic region of interest

Abstract

The application discloses a kind of player intention recognition and shot tracking method based on dynamic region of interest, comprising the following steps: S1, scene adaptive partitioning: according to game real-time single, two, three, four-person scene mode, corresponding division initial ROI, boundary ROI and dynamic ROI three-level region of interest, complete the scene adaptation configuration of shot identification area, and the tracking stability is greatly improved: the application adopts the mechanism that dynamic ROI follows target player real-time update, discards traditional fixed area tracking mode, can adapt to player high-speed displacement scene, effectively reduce shot jitter, drift problem, keep shot tracking stable throughout, target switching precision is high: the application fuses player position area, occlusion state, threefold determination condition of multistage time threshold, constructs standardization person-changing intention recognition logic, can accurately distinguish player interpenetration run, temporary occlusion, long-term out-of-bounds and different scenes, from the root, eliminate shot mis-switching, frequent jumping problem.

Description

Technical Field

[0001] The invention relates to the field of computer vision applications, specifically a method for player intent recognition and camera tracking based on dynamic regions of interest. Background Technology

[0002] In scenarios such as game live streaming, game spectating, and intelligent editing, intelligent and stable camera tracking and accurate player intent recognition are core technologies for ensuring visual appeal and viewing experience. Currently, traditional game camera tracking solutions have many technical shortcomings and are difficult to adapt to the needs of complex multiplayer game scenarios.

[0003] Most existing technologies rely on manual camera control or simple fixed-area detection to achieve camera tracking, resulting in extremely low automation. They can only achieve basic passive tracking effects and lack the ability to actively recognize player intentions and intelligently switch players for tracking. Furthermore, traditional solutions often use single-target tracking algorithms, locking onto a single player with a fixed target bounding box and continuously tracking them. These algorithms lack scene recognition and intention judgment capabilities. When the tracked target is completely obscured by scene obstacles, or when the player quickly moves or leaves the frame, problems such as tracking drift, tracking failure, and screen stuttering are highly likely to occur. Moreover, they cannot intelligently switch to focusing on key new players within the scene.

[0004] Furthermore, traditional tracking solutions do not differentiate between the number of players in a game scene, nor do they establish a hierarchical region of interest division mechanism or multi-level time threshold judgment logic. In complex scenarios involving multiple players in chaotic battles, frequent player movement, and mutual occlusion, problems such as accidental camera switching, frequent jumps, and poor tracking continuity are prone to occur. They lack a complete player replacement logic that includes occlusion judgment, boundary triggering, and fallback for target loss, making it impossible to flexibly adapt to single-player and multi-player scenarios. This seriously affects the stability and professionalism of the game camera footage and greatly limits the practical application of intelligent camera tracking technology. Summary of the Invention

[0005] The purpose of this invention is to provide a method for player intent recognition and camera tracking based on dynamic regions of interest, in order to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the invention provides the following technical solution: a method for player intent recognition and camera tracking based on dynamic regions of interest, comprising the following steps:

[0007] S1. Scene Adaptation Partitioning: Based on the real-time single-player, two-player, three-player, and four-player scene modes of the game, the game divides the region of interest into three levels: initial ROI, boundary ROI, and dynamic ROI, and completes the scene adaptation configuration of the camera recognition area.

[0008] S2. Preset multi-level judgment thresholds: Pre-configure three types of time judgment parameters: boundary replacement threshold, occlusion replacement threshold, and loss fallback threshold, as the judgment benchmark for player intent recognition and camera switching;

[0009] S3. Initial target player selection: Within the initial ROI area, all player targets in the scene are identified through human detection algorithms, and the positions of each player are locked by multi-target tracking algorithms. The straight-line distance between the center of each player's detection box and the center of the screen is calculated, and the player with the closest distance is selected as the initial tracking target player.

[0010] S4. Real-time Update and Status Monitoring of Dynamic Regions: The center coordinates of the detection box of the target player in the current frame are obtained in real time, and the dynamic ROI area is updated synchronously with the player's movement. The dynamic judgment core area is formed by expanding outward from the center of the target player by a preset multiple. The area position, occlusion status, target loss status, and scene status of new players entering the dynamic ROI are monitored in real time.

[0011] S5, Multi-condition Intelligent Intent Recognition and Camera Switching: Based on real-time monitoring status and combined with multi-level time thresholds, it executes three intelligent switching logics: boundary replacement, occlusion replacement, and loss fallback, to complete the adaptive switching of the player's tracking target.

[0012] S6. Stabilized Lens Output Control: Based on the determined real-time tracking target, outputs a stabilized lens tracking image to achieve continuous lens control for short-term target loss, long-term target loss, and automatic target switching in multiple scenes.

[0013] As a preferred embodiment of the invention, the division of the three-level regions of interest in S1 is as follows: the initial ROI is the core area of ​​the camera's default focus, used for initial target screening and fallback target matching; the boundary ROI is the outer transition area of ​​the initial ROI, used to trigger the player switching judgment mechanism; the dynamic ROI is dynamically generated with the target player being tracked in real time as the center, and moves in real time with the target player, serving as the core area for real-time player switching judgment.

[0014] As a preferred embodiment of the invention, the various thresholds in S2 are configurable parameters. The boundary replacement threshold and the occlusion replacement threshold are both set to 1s, and the loss fallback threshold is set to 3s. The parameter values ​​can be adaptively adjusted according to the game screen resolution, operating scenario, and business requirements.

[0015] As a preferred embodiment of the invention, the generation range of the dynamic ROI in S4 is: based on the center coordinates of the target player detection box, it expands outward by 1.5 times to form a closed area, and the expansion multiple of the dynamic ROI supports custom configuration to adapt to different screen resolution scenes.

[0016] As a preferred embodiment of the invention, the boundary player switching logic in S5 specifically includes: when the target player moves into the boundary ROI area, the boundary player switching timer is started; if the timer duration reaches the boundary player switching threshold and there is a new player in the dynamic ROI, the tracking target is switched to the new player in the dynamic ROI; if there is no new player in the dynamic ROI, the target is switched to the player closest to the center of the screen in the initial ROI.

[0017] As a preferred embodiment of the invention, the occlusion switching logic in S5 specifically includes: when a new player enters the dynamic ROI area and occludes the currently tracked target player, and the duration of the occlusion reaches the occlusion switching threshold, the tracking target is switched immediately; for short-term occlusions formed by instantaneous movement, the switching operation is not triggered if the threshold duration is not reached, thus avoiding accidental switching.

[0018] As a preferred embodiment of the invention, the loss fallback logic in S5 specifically includes: when the currently tracked target player walks out of the screen or is completely obscured, causing the target to be lost, a loss timer is started; if the loss duration does not reach the loss fallback threshold, the original camera tracking state is maintained, and the target is waited to reappear; if the loss duration reaches the loss fallback threshold, the system automatically switches to the nearest player within the initial ROI, or performs a camera center reset operation.

[0019] As a preferred embodiment of the invention, the camera output control logic of S6 includes: in the normal target tracking state, keeping the camera stably following the target player's movement; when the target is lost for a short time, locking the camera view to avoid screen flickering; after the target is lost for a long time and the fallback logic is triggered, automatically switching the target or resetting the camera to ensure that the camera view is continuous and uninterrupted.

[0020] As a preferred embodiment of the invention, the method is adaptable to game scenarios with a full range of player numbers, including single-player, two-player, three-player, and four-player games. Throughout the game, the algorithm automatically identifies player intentions and switches tracking targets, eliminating the need for manual adjustment of the camera angle.

[0021] Compared with existing technologies, the beneficial effects of the invention are:

[0022] 1. Significantly improved tracking stability: This invention adopts a mechanism that dynamically updates the ROI to follow the target player in real time, abandoning the traditional fixed area tracking mode. It can adapt to high-speed player movement scenarios, effectively reducing camera shake and drift issues, and maintaining stable camera tracking throughout the entire process.

[0023] 2. High accuracy of target switching: This invention integrates three judgment conditions: player position area, occlusion status, and multi-level time thresholds, to construct a standardized logic for recognizing player switching intentions. It can accurately distinguish different scenarios such as players running around, brief occlusion, and long-term out-of-bounds movement, thus eliminating the problems of accidental camera switching and frequent jumps from the root.

[0024] 3. Full coverage of scene adaptation: This invention can perfectly adapt to single-player, two-player, three-player, and four-player game scenarios, match exclusive ROI division rules for different scenarios, adapt to various game battle modes, and has extremely strong versatility.

[0025] 4. Excellent algorithm robustness: Corresponding fallback strategies are set up for various abnormal conditions such as player occlusion, going out of bounds, temporary loss, and long-term disconnection to avoid problems such as interruption of camera footage, black screen, and wrong cut, and to ensure continuous and stable output of camera footage.

[0026] 5. Highly flexible and configurable: ROI expansion factor, various time thresholds, and regional division parameters can all be customized and adjusted. It can adapt to different resolutions such as 1080P and 4K, and can be optimized according to different game business scenarios, making it highly practical. Attached Figure Description

[0027] Figure 1 A flowchart illustrating a method for player intent recognition and camera tracking based on dynamic regions of interest (ROIs) for the present invention.

[0028] Figure 2 A schematic diagram of the ROI region concept for a player intent recognition and camera tracking method based on dynamic regions of interest (ROIs) in this invention;

[0029] Figure 3 A schematic diagram illustrating the region definition configuration for a player intent recognition and camera tracking method based on dynamic regions of interest (ROIs).

[0030] Figure 4 This is a flowchart illustrating the initialization process for a target player in a method for player intent recognition and camera tracking based on dynamic regions of interest. Detailed Implementation

[0031] The technical solutions in the embodiments of the invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the invention, and not all embodiments.

[0032] In the description of the invention, it should be noted that the terms "upper," "lower," "inner," "outer," "front end," "rear end," "both ends," "one end," and "the other end," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are used only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on the invention. Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0033] In the description of the invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installed," "equipped with," "connected," etc., should be interpreted broadly. For example, "connection" can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium; it can be a connection within two components. Those skilled in the art can understand the specific meaning of the above terms in the invention based on the specific circumstances.

[0034] Please see Figure 1-4 The invention provides an embodiment of a method for player intent recognition and camera tracking based on dynamic regions of interest, comprising the following steps;

[0035] Step 1: Scenario-based ROI region division: For a four-player game scenario, complete the initial configuration of three-level ROI regions, define the fixed area in the center of the screen as the initial ROI, and the preset range around the initial ROI as the boundary ROI; set the dynamic ROI expansion multiple to 1.5 times, and generate a dynamic judgment area in real time based on the center of the target player detection box.

[0036] Step 2, Multi-level threshold preset configuration: Fixed configuration boundary replacement threshold 1s, occlusion replacement threshold 1s, and loss fallback threshold 3s, which serve as the core judgment parameters for this scenario.

[0037] Step 3: Initial target player selection: Perform human detection on the initial game frame to identify all player targets in the frame. Use a multi-target tracking algorithm to lock the position of each player's detection box, calculate the distance between the center of each player's box and the center of the frame, and select the player with the smallest distance as the initial core target for tracking.

[0038] Step 4, Dynamic Updates and Real-time Monitoring: During game operation, the center coordinates of the currently tracked target player are read frame by frame, and the position and range of the dynamic ROI are refreshed in real time to ensure that the dynamic ROI always covers the core activity area of ​​the target player; at the same time, it is monitored in real time whether the target player enters the boundary ROI, whether he is blocked by other players, whether he completely walks out of the screen and is lost, and whether a new player enters the dynamic ROI area.

[0039] Step 5: Multi-scenario intelligent personnel replacement judgment and execution:

[0040] When a target player continuously moves into the boundary ROI area and stays there for a duration of 1 second, if there is a new player target within the dynamic ROI, the camera will immediately switch the target to the new player; if there is no new player, the camera will switch to the player closest to the center of the screen within the initial ROI.

[0041] When a new player enters the dynamic ROI area and occludes the currently tracked target, and the occlusion lasts for a full 1 second, which is the threshold for switching players, it is determined to be an intentional player switching and the target switch is executed immediately. If the occlusion is only a momentary interlude and lasts for less than 1 second, it is determined to be unintentional occlusion and no player switching is triggered, effectively avoiding accidental switching.

[0042] When the currently tracked target player walks out of the screen or is completely obscured, causing the target to be lost, the system starts a timer. If the target is not detected to reappear within 3 seconds, the loss fallback logic is triggered, and the system automatically switches to the nearest player within the initial ROI. If there is no valid player, the camera center is reset.

[0043] Step 6: Stabilize Camera Output: Throughout the process, the control logic of preserving the camera in short-term loss and switching targets in long-term loss is followed. When the target is moving normally, it follows smoothly. When the target is briefly obscured and lost, the camera view is maintained. In case of abnormal conditions, the fallback logic is triggered and the new target is automatically adapted to continuously output stable, coherent and focused game camera footage for core players.

[0044] It will be apparent to those skilled in the art that the invention is not limited to the details of the exemplary embodiments described above, and that it can be implemented in other specific forms without departing from the spirit or essential characteristics of the invention. Therefore, the embodiments should be considered illustrative and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

Claims

1. A method for player intent recognition and camera tracking based on dynamic regions of interest, characterized in that, Includes the following steps: S1. Scene Adaptation Partitioning: Based on the real-time single-player, two-player, three-player, and four-player scene modes of the game, the game divides the region of interest into three levels: initial ROI, boundary ROI, and dynamic ROI, and completes the scene adaptation configuration of the camera recognition area. S2. Preset multi-level judgment thresholds: Pre-configure three types of time judgment parameters: boundary replacement threshold, occlusion replacement threshold, and loss fallback threshold, as the judgment benchmark for player intent recognition and camera switching; S3. Initial target player selection: Within the initial ROI area, all player targets in the scene are identified through human detection algorithms, and the positions of each player are locked by multi-target tracking algorithms. The straight-line distance between the center of each player's detection box and the center of the screen is calculated, and the player with the closest distance is selected as the initial tracking target player. S4. Dynamic Region Real-time Update and Status Monitoring: The center coordinates of the detection box of the target player in the current frame are obtained in real time, and the dynamic ROI area is updated synchronously with the player's movement. The dynamic judgment core area is formed by expanding outward from the center of the target player by a preset multiple. Real-time monitoring of the target player's location, occlusion status, target loss status, and scene status when a new player enters the dynamic ROI; S5, Multi-condition Intelligent Intent Recognition and Camera Switching: Based on real-time monitoring status and combined with multi-level time thresholds, it executes three intelligent switching logics: boundary replacement, occlusion replacement, and loss fallback, to complete the adaptive switching of the player's tracking target. S6. Stabilized Lens Output Control: Based on the determined real-time tracking target, outputs a stabilized lens tracking image to achieve continuous lens control for short-term target loss, long-term target loss, and automatic target switching in multiple scenes.

2. The method for player intent recognition and camera tracking based on dynamic regions of interest according to claim 1, characterized in that, The division of the three levels of Region of Interest (ROI) in S1 is as follows: the initial ROI is the core area of ​​the camera's default focus, used for initial target screening and fallback target matching; the boundary ROI is the outer transition area of ​​the initial ROI, used to trigger the player switching judgment mechanism; the dynamic ROI is dynamically generated with the target player being tracked in real time as the center, and moves in real time with the target player, serving as the core area for real-time player switching judgment.

3. The method for player intent recognition and camera tracking based on dynamic regions of interest according to claim 1, characterized in that, In S2, the various thresholds are configurable parameters. The boundary replacement threshold and the occlusion replacement threshold are both set to 1s, and the loss fallback threshold is set to 3s. The parameter values ​​can be adaptively adjusted according to the game screen resolution, running scenario and business requirements.

4. The method for player intent recognition and camera tracking based on dynamic regions of interest according to claim 1, characterized in that, The generation range of the dynamic ROI in S4 is: based on the center coordinates of the target player detection box, it expands outward by 1.5 times to form a closed area, and the expansion multiple of the dynamic ROI supports custom configuration to adapt to different screen resolution scenes.

5. The method for player intent recognition and camera tracking based on dynamic regions of interest according to claim 1, characterized in that, The boundary player swapping logic in S5 specifically includes: when the target player moves into the boundary ROI area, the boundary player swapping timer is started; if the timer duration reaches the boundary player swapping threshold and there is a new player in the dynamic ROI, the tracking target is switched to the new player in the dynamic ROI; if there is no new player in the dynamic ROI, the target is switched to the player closest to the center of the screen in the initial ROI.

6. The method for player intent recognition and camera tracking based on dynamic regions of interest according to claim 1, characterized in that, The occlusion-based player switching logic in S5 specifically includes: when a new player enters the dynamic ROI area and occludes the currently tracked target player, and the duration of the occlusion reaches the occlusion-based player switching threshold, the tracking target switching is immediately executed; for short-term occlusions caused by momentary interspersed movement, the player switching operation is not triggered if the threshold duration is not reached, thus avoiding accidental switching.

7. The method for player intent recognition and camera tracking based on dynamic regions of interest according to claim 1, characterized in that, The loss fallback logic in S5 specifically includes: when the currently tracked target player walks out of the screen or is completely obscured, causing the target to be lost, a loss timer is started; if the loss duration does not reach the loss fallback threshold, the original camera tracking state is maintained, and the target is waited to reappear; if the loss duration reaches the loss fallback threshold, the system automatically switches to the nearest player within the initial ROI, or performs a camera center reset operation.

8. The method for player intent recognition and camera tracking based on dynamic regions of interest according to claim 1, characterized in that, The camera output control logic of the S6 includes: under normal target tracking conditions, keeping the camera stably following the target player's movement; when the target is briefly lost, locking the camera view to avoid screen jitter; when the target is lost for a long time and the fallback logic is triggered, automatically switching the target or resetting the camera to ensure continuous and uninterrupted camera footage.

9. The method for player intent recognition and camera tracking based on dynamic regions of interest according to claim 1, characterized in that, The method is suitable for single-player, two-player, three-player, and four-player game scenarios. It automatically identifies player intentions and switches tracking targets throughout the game through algorithms, without the need for manual adjustment of the camera angle.

Images