View angle cooperative scheduling method and system of multi-pan-tilt camera

Abstract

The invention provides a visual angle collaborative scheduling method and system for a multi-pan-tilt camera, and relates to the technical field of pan-tilt collaboration, and the method comprises the steps: generating respective priority weight according to the dynamic attribute of each dynamic target; each PTZ camera calculates a bidding value for each dynamic target according to the dynamic state parameters, the dynamic target is allocated to the PTZ camera with the highest bidding value through a bidding allocation mechanism, and an allocation mapping table is generated; extracting the real-time position and priority weight of the dynamic target allocated to each PTZ camera according to the allocation mapping table, and generating an initial rotation path based on an improved RRT algorithm; performing rolling optimization on the kinematics model of the corresponding PTZ camera to generate a real-time rotation angle and a zooming parameter; and on the basis of the real-time rotation angle and the zooming parameter, state monitoring of the dynamic target is executed, a dynamic optimization strategy is triggered when the state does not meet a preset condition, and the tracking and monitoring capability of the multi-pan-tilt camera system on the dynamic target is improved through optimizing the scheduling strategy.

Description

Technical Field

[0001] The present invention relates to the field of pan-tilt collaboration technology, and in particular to a method and system for collaboratively scheduling viewing angles of multiple pan-tilt cameras. Background Art

[0002] With the rapid development of security surveillance, intelligent transportation, robotic navigation, and other fields, the demand for real-time, precise tracking and monitoring of dynamic targets is growing. Multi-panel camera systems, with their ability to observe targets from multiple angles and positions, have become a key technical means of meeting these demands. However, achieving efficient coordinated scheduling between multi-panel cameras to fully utilize their resources and improve the accuracy and real-time performance of target tracking remains a key technical challenge.

[0003] Among existing multi-PTZ camera target allocation technologies, common methods include distance-based allocation strategies, which simply determine which camera is responsible for tracking a dynamic target based on the spatial distance between the PTZ camera and the target. While computationally simple, this approach suffers from significant drawbacks. It fails to fully consider the dynamic properties of dynamic targets, such as target type, motion speed, and position sensitivity. Different types of targets have varying security sensitivities, such as the significant difference in security concerns between VIPs and ordinary objects. Targets moving at different speeds also have varying real-time tracking requirements: fast-moving targets require more immediate responses, while targets with high position sensitivity require more precise tracking. Furthermore, traditional methods fail to consider the dynamic state parameters of the PTZ camera itself, such as its current position, viewing angle, and current load status. A PTZ camera's current position and viewing angle determine its reachability and observation quality for different targets, while its current load status affects its ability to respond promptly to new tracking tasks. Therefore, this simple distance-based allocation strategy fails to achieve optimal allocation of multi-PTZ camera resources, resulting in low tracking efficiency and even potential target loss.

[0004] Therefore, it is necessary to provide a method and system for collaboratively scheduling the viewing angles of multi-pan-tilt cameras to solve the above technical problems. Summary of the Invention

[0005] To solve the above technical problems, the present invention provides a method and system for collaborative scheduling of the viewing angles of multi-pan-tilt cameras, which improves the tracking and monitoring capabilities of the multi-pan-tilt camera system for dynamic targets by optimizing the scheduling strategy.

[0006] The present invention provides a method for collaboratively scheduling viewing angles of multi-pan-tilt cameras, the method comprising the following steps:

[0007] Based on the preset dynamic target list, the priority weight of each dynamic target is generated according to the dynamic attributes of each dynamic target;

[0008] Each PTZ camera calculates a bid value for each dynamic target based on its own dynamic state parameters, and allocates the dynamic target to the PTZ camera with the highest bid value through a bidding allocation mechanism to generate an allocation mapping table;

[0009] Extracting the real-time position and priority weight of the dynamic target assigned to each PTZ camera according to the allocation mapping table, and generating an initial rotation path based on the improved RRT algorithm under obstacle constraints;

[0010] Based on the initial rotation path, a model predictive control method is used to perform rolling optimization on the kinematic model of the corresponding pan-tilt camera to generate real-time rotation angle and zoom parameters;

[0011] Based on the real-time rotation angle and zoom parameters, the state monitoring of the dynamic target is performed, and the dynamic optimization strategy is triggered when the state does not meet the preset conditions.

[0012] Preferably, the dynamic attributes of each target include target type, movement speed and position sensitivity.

[0013] Preferably, the dynamic state parameters include current position, viewing angle range and current load state.

[0014] Preferably, the process of generating priority weights based on dynamic attributes includes:

[0015] A type weight is assigned according to the security sensitivity of the target type, and a priority weight is calculated in combination with the real-time motion speed and position sensitivity of the dynamic target. The calculation formula of the priority weight is:

[0016] P i =α·W+β·v+γ·S

[0017] Among them, P i is the priority weight of the i-th dynamic target, α, β, and γ are the preset normalization coefficients of type weight W, motion speed v, and position sensitivity S, respectively.

[0018] Preferably, each PTZ camera calculates a bid value for each dynamic target according to its own dynamic state parameter, and allocates the dynamic target to the PTZ camera with the highest bid value through a bidding allocation mechanism, thereby generating an allocation mapping table, including:

[0019] Each PTZ camera calculates a bid value for each dynamic target based on its current position, viewing angle range, and current load state in its dynamic state parameters, wherein the calculation formula for the bid value is:

[0020]

[0021] Among them, T k,irepresents the estimated time for the PTZ camera k to rotate to the dynamic target i, L k represents the current load state of camera k, λ represents the initial load balancing coefficient, B k,i represents the bid value of the dynamic target;

[0022] The highest bid value submitted by each camera is screened through central arbitration to generate an allocation mapping table containing target-camera binding relationships.

[0023] Preferably, the generation of the initial rotation path specifically includes:

[0024] Parsing the real-time position of the dynamic target corresponding to each pan-tilt camera and its associated priority weight from the allocation mapping table, wherein the priority weight is used to represent the scheduling priority of the dynamic target;

[0025] The real-time position of the dynamic target is used as the destination point of the path planning, and the current three-dimensional position of the pan-tilt camera is used as the starting point of the path planning, forming a start-end point constraint pair for the path planning;

[0026] Under the constraint of obstacle distribution in a known environment map, an improved RRT algorithm is used for path search, wherein the improved RRT algorithm uses the priority weight value of the dynamic target as a path optimization factor during the path generation process, and the priority weight is proportional to the planning priority;

[0027] Based on obstacle constraints and path optimization factors, an initial rotation path is generated starting from the starting position of the pan-tilt camera and passing through the target points of each dynamic target in sequence, wherein the initial rotation path includes a node sequence and a rotation time sequence.

[0028] Preferably, the rolling optimization process of the model predictive control method includes:

[0029] Constructing a reference trajectory within a rolling time domain window based on the node sequence of the initial rotation path and the corresponding rotation time sequence;

[0030] Taking target tracking error and gimbal energy consumption as optimization objectives, a multi-objective cost function is constructed;

[0031] Using the dynamic state parameters as constraints, physical constraints of the kinematic model of the pan-tilt camera are constructed;

[0032] In each control cycle, the optimal solution of the multi-objective cost function under the constraint conditions is solved to generate real-time rotation angle and zoom parameters.

[0033] Preferably, the triggering and execution process of the dynamic optimization strategy includes:

[0034] Based on the generated real-time rotation angle and zoom parameters, the following monitoring operations are performed:

[0035] Calculate the real-time tracking error and zoom error of the dynamic target. If any error exceeds the error threshold defined in the multi-objective cost function, call the model predictive control method to perform rolling horizon optimization adjustment.

[0036] When a dynamic target is detected outside the defined viewing angle range and / or the current load status exceeds a preset threshold, the bidding allocation mechanism is triggered to recalculate the bidding value and update the allocation mapping table, and the load balancing coefficient is dynamically adjusted based on the current load status.

[0037] The present invention also provides a multi-pan-tilt camera viewing angle collaborative scheduling system, which is used to execute the multi-pan-tilt camera viewing angle collaborative scheduling method. The system includes:

[0038] A priority weight generation module is used to generate respective priority weights according to the dynamic attributes of each dynamic target based on a preset dynamic target list;

[0039] The bidding allocation mechanism module is used for each PTZ camera to calculate the bidding value for each dynamic target according to its own dynamic state parameters, and allocate the dynamic target to the PTZ camera with the highest bidding value through the bidding allocation mechanism to generate an allocation mapping table;

[0040] An initial path planning module is used to extract the real-time position and priority weight of the dynamic target assigned to each PTZ camera according to the allocation mapping table, and generate an initial rotation path based on the improved RRT algorithm under obstacle constraints;

[0041] A rolling optimization control module is used to perform rolling optimization on the kinematic model of the corresponding PTZ camera based on the initial rotation path using a model predictive control method to generate real-time rotation angles and zoom parameters;

[0042] The dynamic optimization trigger module is used to perform status monitoring of dynamic targets based on real-time rotation angle and zoom parameters, and trigger the dynamic optimization strategy when the status does not meet the preset conditions.

[0043] Compared with related technologies, the method and system for collaborative scheduling of viewing angles of multi-pan-tilt cameras provided by the present invention have the following beneficial effects:

[0044] In target allocation, the present invention integrates dynamic target attributes and pan-tilt camera state parameters to determine priority weights and bidding values, accurately allocates targets, improves allocation rationality and efficiency, and fully utilizes resources. Path planning uses an improved RRT algorithm, taking into account obstacle constraints and target priority weights, to generate a more optimal path, ensure the tracking of important targets, and make the path smoother, reduce energy consumption, and improve tracking stability. Motion control is based on model predictive control, with target tracking error and energy consumption as optimization targets. Combined with state parameter constraints, the optimal solution is solved in real time to generate control parameters, achieving a balance between tracking effect and energy consumption. The dynamic optimization strategy can monitor and respond to changes in tracking error, target position, and load status in real time, and adjust control parameters or reallocate targets in a timely manner, thereby enhancing the system's adaptability and robustness to complex environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 A flowchart of a method for collaboratively scheduling viewing angles of multi-pan-tilt cameras provided by the present invention;

[0046] Figure 2 This is a module structure diagram of a multi-pan-tilt camera viewing angle collaborative scheduling system provided by the present invention. DETAILED DESCRIPTION

[0047] The present invention will be further described in detail below with reference to the accompanying drawings and examples. It will be understood that the specific embodiments described herein are intended only to illustrate the present invention and are not intended to limit the present invention. It should also be noted that, for ease of description, the accompanying drawings only illustrate portions relevant to the present invention, not all of the structures. Furthermore, the embodiments of the present invention and the features of the embodiments may be combined with one another unless there is a conflict.

[0048] It should also be noted that, for ease of description, only the part relevant to the present invention, rather than all of the content, is shown in the accompanying drawings. Before discussing exemplary embodiments in more detail, it should be mentioned that some exemplary embodiments are described as processing or methods depicted as flow charts. Although flow charts describe various operations (or steps) as sequential processing, many operations therein can be implemented in parallel, concurrently or simultaneously. In addition, the order of various operations can be rearranged. When its operation is completed, the processing can be terminated, but can also have additional steps not included in the accompanying drawings. The processing can correspond to methods, functions, procedures, subroutines, subprograms, etc.

[0049] Example 1

[0050] The present invention provides a method for coordinating the viewing angles of multi-panel cameras. Figure 1 As shown, the method includes the following steps:

[0051] S1: Based on a preset dynamic target list, generate respective priority weights according to the dynamic attributes of each dynamic target.

[0052] The dynamic attributes of each target include target type, movement speed and position sensitivity.

[0053] Specifically, in step S1, the process of generating priority weights based on dynamic attributes includes:

[0054] A type weight is assigned according to the security sensitivity of the target type, and a priority weight is calculated in combination with the real-time motion speed and position sensitivity of the dynamic target. The calculation formula of the priority weight is:

[0055] P i =α·W+β·v+γ·S

[0056] Among them, P i is the priority weight of the i-th dynamic target, α, β, and γ are the preset normalization coefficients of type weight W, motion speed v, and position sensitivity S, respectively.

[0057] In this embodiment, first, based on the preset dynamic target list, a priority weight is generated for each dynamic target. The specific process is as follows:

[0058] Target types are quantified based on their security sensitivity. For example, targets are divided into three security levels: high, medium, and low, each with a corresponding weight. Highly sensitive targets such as people are given a higher weight, medium-sensitive targets such as vehicles are given a medium weight, and low-sensitivity targets such as environmental disturbances are given a lower weight.

[0059] After the real-time motion speed of the dynamic target is obtained through the tracking algorithm, it is normalized and mapped to a preset numerical range to eliminate dimensional differences. For example, the actual speed is scaled to between 0 and 1 according to the scene requirements.

[0060] Position sensitivity is dynamically adjusted based on the criticality of the target's monitored area. For example, core areas are assigned the highest sensitivity, buffer areas the second highest, and edge areas the lowest. Type weight, normalized speed, and position sensitivity are weighted and summed using preset normalization coefficients. The sum of these coefficients must meet normalization requirements to ensure the appropriateness of the priority weights. For example, through experimental calibration, the weights of each coefficient are adjusted so that the influence of target type on the weight is slightly greater than that of speed and position sensitivity. The final calculated priority weight is updated in real time and serves as a core input parameter for subsequent bidding allocation and path optimization.

[0061] S2: Each PTZ camera calculates a bid value for each dynamic target based on its own dynamic state parameters, and allocates the dynamic target to the PTZ camera with the highest bid value through a bidding allocation mechanism to generate an allocation mapping table.

[0062] The dynamic state parameters include the current position, the viewing angle range and the current load state.

[0063] Specifically, step S2 includes the following steps:

[0064] S21: Each PTZ camera calculates a bid value for each dynamic target based on its current position, viewing angle range, and current load state in its dynamic state parameters, wherein the calculation formula of the bid value is:

[0065]

[0066] Among them, T k,i represents the estimated time for the PTZ camera k to rotate to the dynamic target i, L k represents the current load state of camera k, λ represents the initial load balancing coefficient, B k,i Indicates the bid value of a dynamic target.

[0067] In this embodiment, the bid value calculation formula uses quantitative parameters to comprehensively evaluate the PTZ camera's ability to track dynamic targets and the rationality of system resource allocation. Priority weight (determined by target type, motion speed, and position sensitivity) serves as the core driver, with higher values indicating a higher priority for target tracking. The estimated rotation time is calculated based on the geometric relationship between the camera's current position and the target's real-time position, as well as the PTZ's mechanical properties (such as maximum rotation speed). This reflects the camera's spatial coverage efficiency of the target; shorter times indicate faster tracking responses.

[0068] The current load status is quantified using the number of tasks or resource usage as an indicator to measure the real-time working pressure of the camera and avoid performance degradation due to overload.

[0069] The load balancing coefficient is a preset parameter used to adjust the weight of the load's impact on the bid value. For example, the higher the coefficient value, the more the system tends to assign tasks to low-load devices to maintain global balance.

[0070] The numerator of the bid value calculation formula prioritizes allocation rights for high-priority targets, while the denominator achieves a balance between efficiency and stability through the combined effects of rotation time and load status. Specifically, rotation time encourages the camera to choose the shortest path to track the target, while load status suppresses the bid value of high-load devices, triggering task diversion. This mechanism enables dynamic targets to dynamically adjust their allocation strategy based on real-time status (such as target movement and changes in regional sensitivity) and system resources (such as fluctuations in camera load).

[0071] S22: The highest bid value submitted by each camera is screened through central arbitration to generate an allocation mapping table including target-camera binding relationships.

[0072] In this embodiment, the central arbitration unit receives bid value data submitted by all PTZ cameras, aggregates it by dynamic target, and stores it. Each target's bid value set contains the bid results of all cameras for that target. The central arbitration unit then compares each target's bid value set one by one, selecting the camera with the highest bid value for each target.

[0073] If multiple cameras bid the same highest bid for the same target, the camera with the lower load will be selected based on its current load status. If the load status is the same, the camera with the closer distance to the target will be selected as the final binding object based on the distance between the camera and the target.

[0074] After screening, the central arbitration system integrates each dynamic target with its corresponding optimal camera ID, bid value, and priority weight into a structured table, generating an allocation mapping table for target-camera binding relationships. This mapping table is updated in real time and records the target's unique identifier, camera's unique identifier, calculated bid value, and priority weight. If the system detects load fluctuations or environmental changes, the central arbitration system dynamically adjusts the load balancing factor and re-triggers the bid allocation process to ensure global resource balance. This allocation mapping table serves as input for subsequent path planning and real-time control, and is transmitted to the path planning module via a standardized interface, seamlessly integrating task allocation with dynamic scheduling.

[0075] S3: extracting the real-time position and priority weight of the dynamic target assigned to each PTZ camera according to the allocation mapping table, and generating an initial rotation path based on the improved RRT algorithm under obstacle constraints.

[0076] Specifically, in step S3, the generation of the initial rotation path specifically includes:

[0077] S31: parsing the real-time position of the dynamic target corresponding to each PTZ camera and its associated priority weight from the allocation mapping table, wherein the priority weight is used to represent the scheduling priority of the dynamic target.

[0078] In this embodiment, the real-time position coordinates and priority weights of the dynamic targets corresponding to each PTZ camera are parsed from the allocation mapping table. The specific operations are:

[0079] The target unique identifier and its associated camera unique identifier in the mapping table are read, and the position data (coordinates in the three-dimensional Cartesian coordinate system) of all dynamic targets currently bound to the camera are extracted. The priority weight values generated in step S1 are also obtained simultaneously. Priority weights are stored as floating-point numbers with a normalized range (e.g., 0 to 1). High-weighted targets (e.g., human targets) are identified as priority targets. The parsed data is passed to the path planning module in a structured format, for example, using JSON format to encapsulate the target location, priority weight, and camera identifier.

[0080] S32: The real-time position of the dynamic target is used as the end point of the path planning, and the current three-dimensional position of the pan-tilt camera is used as the starting point of the path planning, thereby forming a start-end point constraint pair of the path planning.

[0081] In this embodiment, based on the resolved target position and the real-time position of the PTZ camera, a start-point-end point constraint pair for path planning is constructed. The current position of the PTZ camera is acquired in real time by a built-in positioning sensor (including but not limited to GPS or visual SLAM) and converted into a coordinate point in a three-dimensional coordinate system as the starting point of the path. The real-time position of the dynamic target is continuously updated by the target tracking algorithm (Kalman filter tracker) as the end point of the path. Each camera-target pair generates an independent path planning task. For example, if a camera needs to track three dynamic targets, three independent start-point-end point constraint pairs are generated. Environmental map data (such as obstacle distribution, passable area) is provided by a preloaded raster map or a three-dimensional point cloud model to ensure that the spatial constraints of the path search process are complete.

[0082] S33: Under the constraint condition of obstacle distribution in a known environment map, an improved RRT algorithm is used to perform path search, wherein the improved RRT algorithm uses the priority weight value of the dynamic target as a path optimization factor during the path generation process, and the priority weight is proportional to the planning priority.

[0083] In this embodiment, an improved RRT algorithm is used for path search. The core improvement is to incorporate priority weights into the node expansion strategy. The specific process is as follows:

[0084] Taking the camera's current position as the root node, when randomly sampling the expansion direction, the priority is tilted towards the area where the target with high priority weight is located. For example, the sampling probability of the target direction with high priority weight is increased to 70%, and the probability of the target direction with low priority weight is 30%.

[0085] During node expansion, candidate path segments are checked for collision risks with obstacles in real time. If a path segment intersects an obstacle area, the node is removed. Furthermore, the view coverage of path nodes must meet the camera's viewing range constraints. For example, the node position must ensure that the target is within the camera's 60° horizontal and 45° vertical viewing angles. Through iterative expansion, a collision-free path connecting the start and end points is generated, and the path node sequence is recorded.

[0086] S34: Based on the obstacle constraints and the path optimization factor, an initial rotation path is generated starting from the starting position of the pan-tilt camera and sequentially passing through the target points of each dynamic target, wherein the initial rotation path includes a node sequence and a rotation time sequence.

[0087] In this embodiment, an initial rotation path and a rotation time sequence are generated based on a path node sequence. The path node sequence is smoothed (using B-spline interpolation) to eliminate redundant inflection points, forming a continuously executable rotation trajectory. The rotation time sequence is calculated based on the mechanical parameters of the gimbal: the Euclidean distance from the current node to the next node is divided by the maximum rotation speed of the gimbal to obtain the rotation time of this path segment. For example, if the node spacing is 2 meters and the maximum rotation speed of the gimbal is 0.5 meters per second, the rotation time of this segment is 4 seconds.

[0088] The resulting initial rotation path, consisting of a sequence of node coordinates, rotation timestamps, and priority weights, is passed to the model predictive control module via a standardized interface. Furthermore, the path must be verified for global view coverage, ensuring that all high-priority targets are covered at least once during path execution. Otherwise, path replanning is triggered.

[0089] S4: Based on the initial rotation path, a model predictive control method is used to perform rolling optimization on the kinematic model of the corresponding pan-tilt camera to generate a real-time rotation angle and zoom parameters.

[0090] Specifically, in step S4, the rolling optimization process of the model predictive control method includes:

[0091] S41: Constructing a reference trajectory in a rolling time domain window based on the node sequence of the initial rotation path and the corresponding rotation time sequence.

[0092] In this embodiment, the timestamps of the path node sequence are aligned with the control cycle, and a continuous time-space reference trajectory is generated by interpolation. For example, if the node sequence of the initial path includes position A (time 0 seconds), position B (time 3 seconds), and position C (time 6 seconds), then the rolling time domain window (such as a window length of 5 seconds) intercepts the trajectory segment of the next 5 seconds within each control cycle (such as 0.1 seconds), and refines the trajectory points through cubic spline interpolation to ensure a smooth transition of the angle and zoom parameters. Each time point of the reference trajectory contains the expected value of the target angle and zoom parameter, which serves as the tracking benchmark for model predictive control.

[0093] S42: Taking target tracking error and gimbal energy consumption as optimization objectives, construct a multi-objective cost function J;

[0094]

[0095] Among them, θ n and z n Represent the real-time rotation angle and real-time zoom parameter of the PTZ camera, θ ref,n and z ref,n denote the angle and zoom parameters of the nth node in the generated reference trajectory, E(T k,n ) represents the energy consumption corresponding to the rotation time T required for the pan-tilt camera k to rotate to the nth node in the reference trajectory, which is calculated using the pan-tilt power model. ω1 and ω2 are preset weight coefficients.

[0096] In this embodiment, a multi-objective cost function is constructed with target tracking error and gimbal energy consumption as optimization targets. The tracking error is quantified as the sum of squares of the deviations between the real-time angle and the reference angle, and the sum of squares of the deviations between the real-time zoom parameter and the reference zoom parameter. The gimbal energy consumption is calculated by multiplying the rotation time by the power model. For example, if the rotation time of a certain path is 4 seconds and the power is 50 watts, the energy consumption is 200 joules. The two objectives are balanced by preset weight coefficients. For example, the tracking error weight is set to 0.7 and the energy consumption weight is set to 0.3 to prioritize tracking accuracy. The weight coefficient is calibrated experimentally or adjusted online adaptively to meet the needs of different scenarios.

[0097] S43: Using the dynamic state parameters as constraints, constructing physical limitations of the kinematic model of the pan-tilt camera.

[0098] In this embodiment, dynamic state parameters are used as constraints to construct the physical limitations of the pan-tilt kinematic model. The angle range constraints are -90° to +90° for the pan-tilt horizontal rotation angle and 0° to 45° for the vertical angle; the zoom range constraints are 1 to 20 times the optical zoom factor; the rotation speed is limited to a maximum of 30° / second in the horizontal direction and 15° / second in the vertical direction. At the same time, the load state is limited to the number of assigned tasks not exceeding the maximum capacity of the camera (such as tracking 5 targets simultaneously). The constraints are embedded in the optimization model in the form of inequalities to ensure that the generated control instructions meet the mechanical properties of the equipment.

[0099] S44: In each control cycle, solving the optimal solution of the multi-objective cost function under the constraint conditions, and generating real-time rotation angle and zoom parameters.

[0100] In this embodiment, in each control cycle, a quadratic programming algorithm is used to solve the optimal solution of the multi-objective cost function under the constraints. The specific process is as follows:

[0101] Based on the current gimbal state (angle, zoom, payload) and future window data of the reference trajectory, the future state sequence is predicted, and the control variables (angle increment, zoom increment) are iteratively adjusted to minimize the cost function. The solution output is the real-time rotation angle and zoom parameters for the next control cycle, for example, the horizontal angle is adjusted to +25° and the zoom factor is adjusted to 8x. The optimization results are executed by the underlying controller and fed back to the status monitoring module to form a closed-loop control. If the solution fails (e.g., no feasible solution is found), the control parameters of the previous cycle are retained and an alarm is triggered, waiting for path replanning.

[0102] S5: Based on the real-time rotation angle and zoom parameters, the state monitoring of the dynamic target is performed, and the dynamic optimization strategy is triggered when the state does not meet the preset conditions.

[0103] Specifically, in step S5, the triggering and execution process of the dynamic optimization strategy includes:

[0104] Based on the generated real-time rotation angle and zoom parameters, the following monitoring operations are performed:

[0105] S51: Calculate the real-time tracking error and zoom error of the dynamic target. If any error exceeds the error threshold defined in the multi-objective cost function, call the model predictive control method to perform rolling time domain optimization adjustment.

[0106] In this embodiment, the tracking error and zoom error of the dynamic target are calculated in real time. The tracking error is calculated by the absolute difference between the current gimbal rotation angle and the expected angle at the corresponding time point in the reference trajectory, and the zoom error is determined by the absolute difference between the real-time zoom parameter and the reference zoom parameter. The error threshold is defined by the tolerance range preset in the multi-objective cost function. For example, the tracking angle error threshold is set to 5 degrees, and the zoom error threshold is set to 0.3 times. If any error exceeds the threshold, the model predictive control module is called to perform rolling time domain optimization adjustment: based on the current gimbal state and the updated reference trajectory, the optimal solution of the multi-objective cost function is re-solved, the corrected rotation angle and zoom parameter are generated, and the adjustment instruction is executed through the underlying controller. For example, if the angle error reaches 6 degrees, the optimization calculation is triggered, and the angle increment of the next control cycle is adjusted to reduce the deviation.

[0107] S52: When the dynamic target is detected to be beyond the defined viewing angle range and / or the current load state exceeds the preset threshold, the bidding allocation mechanism is triggered to recalculate the bidding value and update the allocation mapping table, and the load balancing coefficient is dynamically adjusted based on the current load state.

[0108] In this embodiment, dynamic detection of whether a target exceeds the PTZ viewing angle is performed based on the target's azimuth (horizontal) and pitch (vertical) angles in the camera coordinate system. If the target's azimuth exceeds -90° to +90° or its pitch exceeds 0° to 45°, the viewing angle is considered out of bounds. For example, a target azimuth of +95° triggers reallocation.

[0109] Simultaneously, the camera load status is monitored in real time. If the number of assigned tasks exceeds 80% of the maximum capacity (for example, when tracking a maximum of five targets, a trigger is triggered if four tasks have been assigned), the load is determined to be overloaded. Once triggered, the central arbitration unit re-executes the bidding allocation process: all cameras recalculate their bids based on the updated target positions and load status. The calculation formula is the priority weight divided by the sum of the rotation time and the load balancing coefficient multiplied by the current load. The load balancing coefficient is dynamically adjusted based on the current load. For example, when the load exceeds 80%, the coefficient is increased from 0.5 to 0.8 to enhance load suppression. The updated allocation map is synchronized to the path planner via an interface, triggering the PTZ to replan the path. For example, after target A migrates from camera 1 to camera 2, camera 2 immediately generates a collision-free path based on the new position and updates the time-space node sequence of the reference trajectory. After the load drops below 60%, the load balancing coefficient is gradually restored to its initial value, for example, decreasing at a rate of 0.1 per second to prevent frequent parameter fluctuations.

[0110] Example 2

[0111] The present invention also provides a multi-panel camera viewing angle collaborative scheduling system for executing the multi-panel camera viewing angle collaborative scheduling method, referring to Figure 2 As shown, the system includes:

[0112] The priority weight generating module 100 is configured to generate respective priority weights according to the dynamic attributes of each dynamic target based on a preset dynamic target list.

[0113] The bidding allocation mechanism module 200 is used for each PTZ camera to calculate the bidding value for each dynamic target according to its own dynamic state parameters, and to allocate the dynamic target to the PTZ camera with the highest bidding value through the bidding allocation mechanism to generate an allocation mapping table.

[0114] The initial path planning module 300 is used to extract the real-time position and priority weight of the dynamic target assigned to each pan-tilt camera according to the allocation mapping table, and generate an initial rotation path based on the improved RRT algorithm under obstacle constraints.

[0115] The rolling optimization control module 400 is used to perform rolling optimization on the kinematic model of the corresponding pan-tilt camera based on the initial rotation path using a model predictive control method to generate real-time rotation angles and zoom parameters.

[0116] The dynamic optimization triggering module 500 is used to perform state monitoring of the dynamic target based on the real-time rotation angle and zoom parameters, and trigger the dynamic optimization strategy when the state does not meet the preset conditions.

[0117] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems) and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0118] Those skilled in the art will appreciate that all or part of the steps in the various methods of the above embodiments can be completed by instructing related hardware through a program, and the program can be stored in a computer-readable storage medium, including a read-only memory (ROM), a random access memory (RAM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), a one-time programmable read-only memory (OTPROM), an electronically erasable programmable read-only memory (EEPROM), a compact disc read-only memory (CD-ROM) or other optical disc storage, magnetic disk storage, magnetic tape storage, or any other computer-readable medium that can be used to carry or store data.

[0119] It should also be noted that the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, commodity, or apparatus that includes a series of elements includes not only those elements but also other elements not explicitly listed, or includes elements inherent to such process, method, commodity, or apparatus. In the absence of further limitations, an element defined by the phrase "comprises a..." does not exclude the presence of other identical elements in the process, method, commodity, or apparatus that includes the element.

Claims

1. A method for coordinated viewing angle scheduling of multi-pan-tilt cameras, characterized in that: The method comprises the following steps: Based on the preset dynamic target list, the priority weight of each dynamic target is generated according to the dynamic attributes of each dynamic target; Each PTZ camera calculates a bid value for each dynamic target based on its own dynamic state parameters, and allocates the dynamic target to the PTZ camera with the highest bid value through a bidding allocation mechanism to generate an allocation mapping table; Extracting the real-time position and priority weight of the dynamic target assigned to each PTZ camera according to the allocation mapping table, and generating an initial rotation path based on the improved RRT algorithm under obstacle constraints; Based on the initial rotation path, a model predictive control method is used to perform rolling optimization on the kinematic model of the corresponding pan-tilt camera to generate real-time rotation angle and zoom parameters; Based on the real-time rotation angle and zoom parameters, the state monitoring of the dynamic target is performed, and the dynamic optimization strategy is triggered when the state does not meet the preset conditions.

2. The method for coordinated viewing angle scheduling of a multi-pan-tilt camera according to claim 1, characterized in that: The dynamic attributes of each target include target type, movement speed and position sensitivity.

3. The method for coordinated viewing angle scheduling of a multi-pan-tilt camera according to claim 2, characterized in that: The dynamic state parameters include current position, viewing angle range and current load state.

4. The method for coordinated viewing angle scheduling of a multi-pan-tilt camera according to claim 3, characterized in that: The process of generating priority weights based on dynamic attributes includes: A type weight is assigned according to the security sensitivity of the target type, and a priority weight is calculated in combination with the real-time motion speed and position sensitivity of the dynamic target. The calculation formula of the priority weight is: P i =α·W+β·v+γ·S Among them, P i is the priority weight of the i-th dynamic target, α, β, and γ are the preset normalization coefficients of type weight W, motion speed v, and position sensitivity S, respectively.

5. The method for coordinated viewing angle scheduling of a multi-pan-tilt camera according to claim 4, characterized in that: Each PTZ camera calculates a bid value for each dynamic target according to its own dynamic state parameter, and allocates the dynamic target to the PTZ camera with the highest bid value through a bidding allocation mechanism, thereby generating an allocation mapping table, including: Each PTZ camera calculates a bid value for each dynamic target based on its current position, viewing angle range, and current load state in its dynamic state parameters, wherein the calculation formula for the bid value is: Among them, T k,i represents the estimated time for the PTZ camera k to rotate to the dynamic target i, L k represents the current load state of camera k, λ represents the initial load balancing coefficient, B k,i represents the bid value of the dynamic target; The highest bid value submitted by each camera is screened through central arbitration to generate an allocation mapping table containing target-camera binding relationships.

6. The method for coordinated viewing angle scheduling of a multi-pan-tilt camera according to claim 5, characterized in that: The generation of the initial rotation path specifically includes: Parsing the real-time position of the dynamic target corresponding to each pan-tilt camera and its associated priority weight from the allocation mapping table, wherein the priority weight is used to represent the scheduling priority of the dynamic target; The real-time position of the dynamic target is used as the destination point of the path planning, and the current three-dimensional position of the pan-tilt camera is used as the starting point of the path planning, forming a start-end point constraint pair for the path planning; Under the constraint of obstacle distribution in a known environment map, an improved RRT algorithm is used for path search, wherein the improved RRT algorithm uses the priority weight value of the dynamic target as a path optimization factor during the path generation process, and the priority weight is proportional to the planning priority; Based on obstacle constraints and path optimization factors, an initial rotation path is generated starting from the starting position of the pan-tilt camera and passing through the target points of each dynamic target in sequence, wherein the initial rotation path includes a node sequence and a rotation time sequence.

7. The method for coordinated viewing angle scheduling of a multi-pan-tilt camera according to claim 6, characterized in that: The rolling optimization process of the model predictive control method includes: Constructing a reference trajectory within a rolling time domain window based on the node sequence of the initial rotation path and the corresponding rotation time sequence; Taking target tracking error and gimbal energy consumption as optimization objectives, a multi-objective cost function is constructed; Using the dynamic state parameters as constraints, physical constraints of the kinematic model of the pan-tilt camera are constructed; In each control cycle, the optimal solution of the multi-objective cost function under the constraint conditions is solved to generate real-time rotation angle and zoom parameters.

8. The method for coordinated viewing angle scheduling of multi-pan-tilt cameras according to claim 7, characterized in that: The triggering and execution process of the dynamic optimization strategy includes: Based on the generated real-time rotation angle and zoom parameters, the following monitoring operations are performed: Calculate the real-time tracking error and zoom error of the dynamic target. If any error exceeds the error threshold defined in the multi-objective cost function, call the model predictive control method to perform rolling horizon optimization adjustment. When a dynamic target is detected outside the defined viewing angle range and / or the current load status exceeds a preset threshold, the bidding allocation mechanism is triggered to recalculate the bidding value and update the allocation mapping table, and the load balancing coefficient is dynamically adjusted based on the current load status.

9. A multi-pan-tilt camera view angle collaborative scheduling system, used to execute a multi-pan-tilt camera view angle collaborative scheduling method according to any one of claims 1 to 8, characterized in that: The system comprises: A priority weight generation module is used to generate respective priority weights according to the dynamic attributes of each dynamic target based on a preset dynamic target list; The bidding allocation mechanism module is used for each PTZ camera to calculate the bidding value for each dynamic target according to its own dynamic state parameters, and allocate the dynamic target to the PTZ camera with the highest bidding value through the bidding allocation mechanism to generate an allocation mapping table; An initial path planning module is used to extract the real-time position and priority weight of the dynamic target assigned to each PTZ camera according to the allocation mapping table, and generate an initial rotation path based on the improved RRT algorithm under obstacle constraints; A rolling optimization control module is used to perform rolling optimization on the kinematic model of the corresponding PTZ camera based on the initial rotation path using a model predictive control method to generate real-time rotation angles and zoom parameters; The dynamic optimization trigger module is used to perform status monitoring of dynamic targets based on real-time rotation angle and zoom parameters, and trigger the dynamic optimization strategy when the status does not meet the preset conditions.

Images