An AI-based camera control method and device, and a storage medium

Abstract

This application provides an AI-based camera control method, device, and storage medium, relating to the field of intelligent monitoring. It solves the technical problems of resource competition and conflict during multi-camera collaborative tracking, and low monitoring efficiency due to a single control strategy. The method includes: acquiring the current state information of at least two cameras within a target monitoring area, and the motion trajectory of at least one target object within the field of view of the at least two cameras; constructing a conflict detection model based on historical current state information and motion trajectories; identifying camera combinations with tracking competition within the target monitoring area as game participants based on the conflict detection model; constructing a multi-objective optimization-based payoff factor for the game participants; determining the optimal control strategy combination based on the payoff factor, and sending control operation commands. This application can improve the overall efficiency and intelligence level of the monitoring system.

Description

Technical Field

[0001] This application relates to the field of intelligent monitoring, and in particular to an AI-based camera control method, device, and storage medium. Background Technology

[0002] Currently, in the field of security monitoring, multiple cameras are typically deployed within a monitored area to achieve continuous tracking of specific targets. When a target moves within the monitored area, multiple cameras often need to work together to avoid losing track of it. However, most existing camera control solutions are based on simple preset rules or the independent control logic of a single camera, lacking a global collaborative optimization mechanism. When multiple cameras simultaneously detect the same target and attempt to track it, a "tracking competition" phenomenon easily occurs, where multiple cameras focus on the same target simultaneously, leading to wasted monitoring resources and excessive overlap of the fields of view of different cameras, resulting in the omission of other critical information within the monitored area. Furthermore, existing technologies often neglect the comprehensive consideration of the matching degree between the camera's own hardware status and the tracking task when allocating camera tasks, resulting in high overall energy consumption of the monitoring system and difficulty in guaranteeing tracking image quality.

[0003] Therefore, the present invention provides an AI-based camera control method, device, and storage medium to solve the above-mentioned technical problems. Summary of the Invention

[0004] This application provides an AI-based camera control method, device, and storage medium, which solves the technical problem in the prior art of achieving globally optimal configuration of monitoring resources when multiple cameras are collaboratively tracking.

[0005] To achieve the above objectives, this application adopts the following technical solution: Firstly, an AI-based camera control method is provided, including: S1: Obtain the current status information of at least two cameras within the target monitoring area, and the motion trajectory of at least one target object within the field of view of the at least two cameras; wherein, the current status information includes the pan-tilt angle, focal length status, and field of view coverage. S2: Construct a conflict detection model based on the current state information and motion trajectory in history, and identify the combination of cameras with tracking competition in the target monitoring area as game participants based on the conflict detection model; S3: Construct a payoff factor based on multi-objective optimization for the game participants; S4: Determine the optimal combination of control strategies based on the profit factors and send control operation instructions.

[0006] In conjunction with the first aspect above, in one possible implementation, acquiring the current status information of at least two cameras within the target monitoring area, and the motion trajectory of at least one target object within the field of view of the at least two cameras, includes: The data repository obtains the pan-tilt angle, focal length, and field of view of the cameras covering the target monitoring area, and the cameras acquire the movement trajectory of the target object in real time.

[0007] In conjunction with the first aspect mentioned above, one possible implementation involves constructing a conflict detection model based on historical current state information and motion trajectories, including: Extract current state information, movement trajectories, and game participants from the historical reference database; the historical reference data includes several current state information and movement trajectories, as well as game participants set by experts based on the current state information and movement trajectories. The current state information and motion trajectory are integrated into several sets of training data and test data; the training data is used to train the artificial intelligence model, and the test data is used to test the trained artificial intelligence model. The artificial intelligence model is adjusted according to the test results; finally, the input is the current state information and motion trajectory, and the output is the conflict detection model of the game participants; among them, the artificial intelligence model is a nonlinear regression model based on neural network, which is implemented by using BP neural network structure and / or RBF neural network structure.

[0008] In conjunction with the first aspect mentioned above, one possible implementation involves identifying, based on a conflict detection model, combinations of cameras within the target monitoring area that are tracking competing cameras as game participants, including: The current status information of the camera corresponding to the target monitoring area and the movement trajectory of the target object are input into the conflict detection model to obtain the game participants that are tracking and competing within the target monitoring area.

[0009] In conjunction with the first aspect above, in one possible implementation, a payoff factor based on multi-objective optimization is constructed for the game participants, including: The cameras corresponding to the game participants are extracted sequentially, and the tracking confidence factor of the camera for the target object is obtained in real time through the calculation formula (1). ; Calculate the area of ​​the difference between the union of the field of view coverage of the camera and the field of view coverage of all other game participants excluding the camera, and use the ratio of the area of ​​the difference to the total field of view area of ​​the current camera as the uniqueness score. ; Acquire the characteristic power consumption of the camera Based on the aforementioned tracking confidence factor Uniqueness score and characteristic energy consumption The return factor is determined by formula (2). ; The calculation formula (1) is: ; In equation (1), For the number The confidence score output by the camera through a preset target detection algorithm reflects the probability of successfully identifying the target object in the field of view; The distance sensitivity coefficient, obtained through the hardware optical characteristics parameters of the camera, is used to characterize the sensitivity of the camera's imaging quality to changes in target distance. It determines the rate at which the tracking confidence factor gain decreases as the target deviates from the optimal imaging distance. The distance number of the target object is The physical distance of the camera The optimal imaging reference distance is set according to the camera type; The calculation formula (2) is: ; In equation (2), The average tracking confidence factor of the game participants for the target object. The average uniqueness score of the game participants. The average characteristic energy consumption of the game participants.

[0010] In conjunction with the first aspect described above, one possible implementation involves obtaining the characteristic power consumption of the camera. ,include: The energy consumption of the camera when receiving control commands in history is obtained, and several energy consumption data are integrated into an energy consumption group. It is then determined whether the variance of the energy consumption group is less than the corresponding standard energy consumption variance. If so, the average value of the energy consumption in the energy consumption group is calculated to obtain the characteristic energy consumption. If not, remove the energy consumption group with the largest difference from the average energy consumption, recalculate the variance of the remaining energy consumption, and re-evaluate the variance until the variance of the remaining energy consumption in the energy consumption group is less than the corresponding standard energy consumption variance. Then, calculate the average of the remaining energy consumption in the energy consumption group to obtain the characteristic energy consumption. The standard energy consumption variance is obtained through empirical settings.

[0011] In conjunction with the first aspect mentioned above, one possible implementation involves determining the optimal control strategy combination based on return factors, including: Extract the payoff factors of the game participants, number the payoff factors from largest to smallest, and extract the game participants sequentially based on the numbers. When the field of vision of the extracted game participants can cover the target monitoring area, the extracted game participants are integrated into a control strategy combination.

[0012] In conjunction with the first aspect above, in one possible implementation, sending control operation instructions includes: Real-time tracking instructions are issued to the player with the highest payoff factor in the control strategy combination, while fixed-area monitoring instructions are issued to the player with a non-maximum payoff factor in the control strategy combination; the fixed area is determined by the area where the player's field of vision overlaps with the target monitoring area.

[0013] Secondly, an AI-based camera control device is provided, comprising: a communication unit and a processing unit; The communication unit is used to acquire the current status information of at least two cameras within the target monitoring area, and the motion trajectory of at least one target object within the field of view of the at least two cameras; wherein, the current status information includes the pan-tilt angle, focal length status, and field of view coverage. The processing unit is used to construct a conflict detection model based on the current state information and motion trajectory in history, identify the combination of cameras with tracking competition in the target monitoring area as game participants based on the conflict detection model, construct a payoff factor based on multi-objective optimization for the game participants, determine the optimal control strategy combination based on the payoff factor, and send control operation instructions.

[0014] Thirdly, this application provides a storage medium storing instructions that, when executed on an AI-based camera control device, cause the AI-based camera control device to perform the methods described in the first aspect and any possible implementation thereof.

[0015] This application provides an AI-based camera control method, device, and storage medium, with the following advantages: 1. This invention constructs a conflict detection model that can intelligently identify camera combinations with tracking competition based on the current state information of the cameras and the motion trajectory of the target object, avoiding resource waste caused by blind tracking. By constructing a multi-objective optimization benefit factor that includes tracking confidence factors, uniqueness scores, and feature energy consumption, it comprehensively considers imaging quality, field of view coverage contribution, and energy consumption costs, making the control strategy more scientific and reasonable. By determining the optimal control strategy combination based on the benefit factor, it realizes the priority scheduling of high-benefit cameras and the regional monitoring division of labor with other cameras, effectively solving the conflict problem in multi-camera collaborative tracking and improving the overall efficiency and intelligence level of the monitoring system.

[0016] It should be understood that the descriptions of technical features, technical solutions, beneficial effects, or similar language in this application do not imply that all features and advantages can be achieved in any single embodiment. Rather, it is understood that the description of a feature or beneficial effect means that a specific technical feature, technical solution, or beneficial effect is included in at least one embodiment. Therefore, the descriptions of technical features, technical solutions, or beneficial effects in this specification do not necessarily refer to the same embodiment. Furthermore, the technical features, technical solutions, and beneficial effects described in this embodiment can be combined in any suitable manner. Those skilled in the art will understand that embodiments can be implemented without one or more specific technical features, technical solutions, or beneficial effects of a particular embodiment. In other embodiments, additional technical features and beneficial effects may be identified in specific embodiments that do not embody all embodiments. Attached Figure Description

[0017] Figure 1 A schematic diagram illustrating the steps of an AI-based camera control method provided in this application embodiment; Figure 2 A schematic diagram of the steps provided for the embodiments of this application; Figure 3 This is a schematic diagram of the structure of an AI-based camera control device provided in an embodiment of this application. Detailed Implementation

[0018] In the description of this application, unless otherwise stated, " / " means "or," for example, A / B can mean A or B. The "and / or" in this document is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, and B alone. Furthermore, "at least one" means one or more, and "multiple" means two or more. The terms "first," "second," etc., do not limit the quantity or order of execution, and "first," "second," etc., do not necessarily imply differences.

[0019] It should be noted that, in this application, the terms "exemplary" or "for example" are used to indicate that something is being described as an example, illustration, or illustration. Any embodiment or design described as "exemplary" or "for example" in this application should not be construed as being more preferred or advantageous than other embodiments or design solutions. Specifically, the use of terms such as "exemplary" or "for example" is intended to present the relevant concepts in a concrete manner.

[0020] like Figure 1 As shown in the figure, an AI-based camera control method provided in this application includes: S1: Obtain the current status information of at least two cameras within the target monitoring area, and the motion trajectory of at least one target object within the field of view of the at least two cameras; wherein, the current status information includes the pan-tilt angle, focal length status, and field of view coverage. Specifically, the target monitoring area refers to the geographical area requiring security monitoring or specific target tracking, such as logistics parks, commercial plazas, or traffic intersections. The current status information of the cameras is the fundamental parameter determining their monitoring capabilities. The pan-tilt angle determines the camera's orientation, the focal length determines the distance and clarity of the image, and the field of view is the physical area that the camera can currently observe, calculated based on the pan-tilt angle and focal length. Motion trajectories reflect the movement path of target objects, such as vehicles, pedestrians, or suspicious objects, within the monitored area, typically obtained through continuous frame image analysis. By acquiring this information, the system can construct a monitoring situation map of the current moment, clearly defining the "line of sight" of each camera and the dynamic position of the target object, providing a data foundation for subsequent conflict assessment.

[0021] S2: Construct a conflict detection model based on the current state information and motion trajectory in history, and identify the combination of cameras with tracking competition in the target monitoring area as game participants based on the conflict detection model; Specifically, in multi-camera environments, it's common for multiple cameras to simultaneously capture the same target object. If all cameras track this target, it leads to wasted computational resources and may even cause other areas to be missed due to simultaneous rotation. Therefore, this step introduces the concept of "game participants." Here, game participants specifically refer to camera combinations with overlapping fields of view and competing intentions to track the same target object. The conflict detection model is an AI model trained on historical data. It can intelligently determine which cameras will "compete" for the target based on the current camera status and target trajectory. For example, if camera A and camera B have overlapping fields of view and both identify the same target, the model can identify them as a group of game participants. This process transforms the complex physical space competition relationship into a computable set of game model participants.

[0022] S3: Construct a payoff factor based on multi-objective optimization for the game participants; Specifically, after identifying the participants in the game, tracking tasks cannot be simply assigned randomly; instead, the "value" of each camera's participation in tracking needs to be evaluated. The benefit factor is precisely the indicator used to quantify this value. It is not a single-dimensional score, but a quantitative indicator based on multi-objective optimization, comprehensively considering tracking effectiveness, such as image clarity, distance, resource contribution (e.g., unique field of view, coverage of blind spots), and operating costs (e.g., energy consumption, mechanical wear). By constructing the benefit factor, the fuzzy question of "who is more suitable for tracking" can be transformed into a concrete numerical comparison problem, providing a mathematical basis for decision-making.

[0023] S4: Determine the optimal combination of control strategies based on the profit factors and send control operation instructions; Specifically, the optimal control strategy combination refers to the camera collaboration scheme selected from the game participants that achieves the best overall monitoring effect. The system determines the specific task of each camera based on the magnitude of the payoff factor. For example, the camera with the highest payoff factor may be assigned to perform high-precision real-time tracking tasks, while cameras with lower payoff factors may be assigned to perform auxiliary monitoring tasks in fixed areas to fill blind spots left by the tracking camera's rotation. Finally, the system generates corresponding control operation commands, sends them to the corresponding cameras via the network, and drives the pan-tilt unit to rotate or adjust its focus, thereby implementing the aforementioned strategy. Through the above steps, this embodiment achieves closed-loop control throughout the entire process from situational awareness, conflict identification, value assessment to collaborative decision-making, effectively solving the problem of multi-camera tracking competition and improving the overall efficiency of the monitoring system.

[0024] In one possible implementation of this application embodiment, the above-mentioned S1 can be implemented by the following S101, which will be described in detail below: S101: Obtain the current status information of at least two cameras within the target monitoring area, and the motion trajectory of at least one target object within the field of view of the at least two cameras, including: The data repository obtains the pan-tilt angle, focal length, and field of view of the cameras covering the target monitoring area, and the cameras acquire the movement trajectory of the target object in real time.

[0025] It should be noted that the data repository can be a relational database deployed on a cloud server or a local control center, pre-stored with device profile information for all cameras within the monitored area. Since the pan / tilt angle and focal length of the cameras typically remain unchanged or undergo only low-frequency automatic scanning changes when no control commands are received, these parameters are considered static or quasi-static data. By directly reading this information from the data repository, the need to send query commands to the cameras each time is avoided, significantly reducing communication latency and improving system response speed. Simultaneously, the field of view coverage, a geometric area calculated based on the pan / tilt angle and focal length, can also be pre-calculated and stored in the database, or calculated in real-time based on the read angle and focal length.

[0026] It should be noted that, after obtaining the above data, this embodiment further describes in detail the application process of the conflict detection model. Specifically, identifying camera combinations with tracking competition within the target monitoring area as game participants based on the conflict detection model involves inputting the current state information of the corresponding cameras in the target monitoring area and the movement trajectory of the target object into the conflict detection model to obtain the game participants with tracking competition within the target monitoring area. This process demonstrates the advantages of AI models in real-time reasoning. The conflict detection model has been trained offline using a large amount of historical data and has the ability to judge potential conflicts based on the input state. In practical applications, the system only needs to use the real-time collected camera state vector and target trajectory vector as input, and the model can quickly output the camera combinations with competing relationships through forward computation. For example, when the fields of view of both camera A and camera B cover the current position of the target object, and the trajectory predicts that the target is about to enter the overlapping field of view area of ​​both, the model can quickly identify A and B as a group of game participants. This model-based identification method, compared with traditional judgment methods based on fixed rules, can more accurately capture complex spatiotemporal competitive relationships and has stronger generalization ability, adapting to diverse tracking needs in different monitoring scenarios.

[0027] In one possible implementation of this application embodiment, the above-mentioned S2 can be implemented by the following S201 and S202, which are described in detail below: S201: Construct a collision detection model based on historical current state information and motion trajectory, including: Extract current state information, movement trajectories, and game participants from the historical reference database; the historical reference data includes several current state information and movement trajectories, as well as game participants set by experts based on the current state information and movement trajectories. The current state information and motion trajectory are integrated into several sets of training data and test data; the training data is used to train the artificial intelligence model, and the test data is used to test the trained artificial intelligence model. The artificial intelligence model is adjusted according to the test results; finally, the input is the current state information and motion trajectory, and the output is the conflict detection model of the game participants; among them, the artificial intelligence model is a nonlinear regression model based on neural network, which is implemented by using BP neural network structure and / or RBF neural network structure.

[0028] It's important to note that the historical reference database forms the foundation for model training, storing a vast amount of historical monitoring data. This data includes not only past camera status information such as gimbal angles and focal lengths, as well as the corresponding target object's trajectory, but more importantly, it also contains game participant labels set by experts based on the current status information and trajectory. In other words, the historical data has been pre-labeled by humans or expert systems to indicate which camera combinations are competing for tracking under specific conditions. This high-quality, labeled data provides a reliable benchmark for supervised learning.

[0029] It should be noted that the data integration process typically follows the principles of random partitioning or stratified sampling. For example, 70% of the total dataset is used as training data for model learning, while the remaining 30% is used as test data to evaluate the model's generalization ability. This partitioning ensures that the model performs well not only on known data but also maintains accuracy in unknown new scenarios. In this embodiment, the artificial intelligence model is a nonlinear regression model based on a neural network. As a specific implementation, a BP (Back Propagation) neural network structure can be used. This network continuously adjusts the network weights through the error backpropagation algorithm, possessing good nonlinear mapping capabilities and able to capture the complex logical relationships between state information and competitive relationships. As another parallel implementation, an RBF (Radial Basis Function) neural network structure can also be used. RBF networks have the characteristic of local approximation and have unique advantages in processing small sample data or scenarios requiring fast convergence. Those skilled in the art can flexibly choose any of the above network structures according to the actual amount of data and computing resources.

[0030] It should be noted that the specific steps for using test data to test the trained artificial intelligence model and adjusting the model based on the test results are as follows: The current state information and movement trajectory from the test data are input into the trained artificial intelligence model to obtain the corresponding game participants. The corresponding game participants are compared with the corresponding game participants in the test data. If the overlap is greater than the overlap threshold (obtained empirically), no parameter adjustment is needed, and the next set of test data is performed. If the overlap is not greater than the overlap threshold, the corresponding parameters are adjusted until the overlap is greater than the overlap threshold, and then the next set of test data is performed. When the number of test data points where the overlap of game participants is greater than the overlap threshold accounts for 90% or more of the total test data, a conflict detection model is obtained, with the input being the current state information and movement trajectory, and the output being the game participants. The overlap threshold is obtained empirically, for example, it can be set to 85% or 95%, depending on the required model accuracy.

[0031] It should be noted that the participants in the game can also be selected manually.

[0032] S202: Based on the conflict detection model, combinations of cameras within the target monitoring area that are tracking competitors are identified as game participants, including: The current status information of the camera corresponding to the target monitoring area and the movement trajectory of the target object are input into the conflict detection model to obtain the game participants that are tracking and competing within the target monitoring area.

[0033] In one possible implementation of this application embodiment, the above-mentioned S3 can be implemented by the following S301 and S302, which are described in detail below: S301: Construct payoff factors based on multi-objective optimization for the aforementioned game participants, including: The cameras corresponding to the game participants are extracted sequentially, and the tracking confidence factor of the camera for the target object is obtained in real time through the calculation formula (1). Tracking confidence factors It is an indicator for measuring the image quality of a target object by a single camera; Calculate the area of ​​the difference between the union of the field of view coverage of the camera and the field of view coverage of all other game participants excluding the camera, and use the ratio of the area of ​​the difference to the total field of view area of ​​the current camera as the uniqueness score. Among them, uniqueness score This reflects the irreplaceability of the camera's field of view. If the field of view of a certain camera highly overlaps with that of other cameras, its difference area is small and its uniqueness score is low. Conversely, if the camera covers blind spots that other cameras cannot cover, its uniqueness score is high. The significance of introducing this indicator is to avoid all cameras focusing on tracking the same target and ignoring the monitoring of the surrounding area, and to encourage the system to retain those cameras that have unique field of view contributions to perform auxiliary monitoring tasks. Acquire the characteristic power consumption of the camera Based on the aforementioned tracking confidence factor Uniqueness score and characteristic energy consumption The return factor is determined by formula (2). ; The calculation formula (1) is: ; In equation (1), For the number The confidence score output by the camera through the preset target detection model reflects the probability of successfully identifying the target object in the field of view. This confidence score is usually directly output by target detection models such as YOLO or SSD, and the value ranges from 0 to 1. The higher the value, the clearer the target features and the easier it is to capture. The distance sensitivity coefficient, obtained through the hardware optical characteristics parameters of the camera, is used to characterize the sensitivity of the camera's imaging quality to changes in the target distance. It determines the rate at which the tracking confidence factor gains decrease as the target deviates from the optimal imaging distance. Different camera models have different focal lengths and aperture characteristics, so their distance sensitivity coefficients k are also different. By introducing this parameter, the calculation of the gain factor can be adapted to monitoring equipment with different hardware characteristics. The distance number of the target object is The physical distance of the camera The optimal imaging reference distance is set according to the camera type; The calculation formula (2) is: ; In equation (2), The average tracking confidence factor of the game participants for the target object. The average uniqueness score of the game participants. The average characteristic energy consumption of the game participants.

[0034] S302: Obtain the characteristic power consumption of the camera. ,include: For characteristic energy consumption To obtain the data, this embodiment proposes a calculation method with data cleaning capabilities to ensure the robustness of the parameters. This iterative elimination method can gradually remove interfering data and retain valid data that reflects the true energy consumption characteristics of the camera. Specifically, the energy consumption of the camera when receiving control commands in history is obtained, several energy consumption values ​​are integrated into an energy consumption group, and it is determined whether the variance of the energy consumption group is less than the corresponding standard energy consumption variance; if so, the characteristic energy consumption is obtained by averaging the energy consumption in the energy consumption group. If not, remove the energy consumption group with the largest difference from the average energy consumption, recalculate the variance of the remaining energy consumption, and re-evaluate the variance until the variance of the remaining energy consumption in the energy consumption group is less than the corresponding standard energy consumption variance. Then, calculate the average of the remaining energy consumption in the energy consumption group to obtain the characteristic energy consumption. The standard energy consumption variance is obtained through empirical settings.

[0035] (It should be noted that when removing the energy consumption group with the largest difference from the average energy consumption, the average energy consumption is the average energy consumption retained in the current variance judgment step.)

[0036] It should be noted that, considering extreme cases where the camera may be in an unstable state for an extended period, leading to a vicious cycle in data cleaning, this embodiment includes a fallback strategy: if, after removing 90% of the energy consumption data, the variance of the remaining energy consumption is still not less than the standard energy consumption variance, then the average of the last 20% of unremoved energy consumption is used as the feature energy consumption. This strategy ensures that even with extremely poor data quality, the system can still output a usable energy consumption reference value, avoiding control process interruptions due to missing parameters. Through the rigorous calculation and cleaning logic described above, the benefit factor constructed in this embodiment not only comprehensively considers imaging quality, field of view contribution, and energy consumption cost, but also ensures the reliability of input parameters through a data cleaning mechanism, providing a solid quantitative basis for subsequently formulating the optimal control strategy. In one possible implementation of this application embodiment, the above-mentioned S4 can be implemented by the following S401 and S402, which are described in detail below: S401: Determining the optimal control strategy combination based on return factors, including: The payoff factors of the game participants are extracted and numbered from largest to smallest. This numbering process prioritizes the game participants. For example, if there are three game participants, cameras A, B, and C, with calculated payoff factors of 0.9, 0.7, and 0.5 respectively, then camera A is numbered 1, camera B is numbered 2, and camera C is numbered 3. A higher payoff factor indicates that the camera is optimal in terms of current tracking confidence, unique field of view, and energy consumption, and should therefore receive higher scheduling priority.

[0037] Based on their assigned numbers, game participants are extracted sequentially. When the field of view of each extracted participant covers the target monitoring area, the participants are integrated into a control strategy combination. This step employs a cyclical filtering logic: the system first extracts camera number 1 (the one with the highest payoff factor) and determines if its field of view completely covers the target monitoring area. If it does, the monitoring requirement is met by this camera alone, and no other cameras are needed; the control strategy combination then includes only this camera. However, in practical applications, the field of view of a single camera is often limited and cannot completely cover the entire target monitoring area. Therefore, if camera number 1 cannot completely cover the area, the system continues to extract camera number 2 and takes the union of the fields of view of cameras numbered 1 and 2, again determining if the union covers the target monitoring area. If it still does not cover the area, the system continues to extract camera number 3, and so on, until the union of the fields of view of the extracted camera combinations completely covers the target monitoring area. This "sorting-extraction-coverage judgment" loop logic ensures that the final generated control strategy combination not only includes high-priority preferred cameras, but also eliminates monitoring blind spots by supplementing with secondary-priority cameras, achieving comprehensive coverage of the target monitoring area.

[0038] S402: Send control operation commands, including: The game player with the highest payoff factor in the control strategy combination is given a real-time tracking operation command. The player with the highest payoff factor usually has the best imaging quality and tracking conditions, so it is designated as the "main tracker" and is responsible for continuous and dynamic real-time tracking of the target object to ensure that the target is always in the center of the field of view and the image is clear. For players whose payoff factor is not the highest in the control strategy combination, fixed-area monitoring commands are issued. Although these cameras are slightly inferior to the main tracker in overall score, their inclusion in the strategy combination means that their field of view makes an indispensable contribution to covering the target monitoring area. Therefore, they are designated as "auxiliary monitors." Unlike the dynamic tracking of the main tracker, the auxiliary monitors perform fixed-area monitoring tasks. The fixed area is determined by the overlap between the field of view of the game participants and the target monitoring area. Specifically, the system calculates the geometric intersection of the camera's field of view and the target monitoring area; this intersection area is the fixed area that the camera needs to focus on monitoring. The system controls the camera's pan-tilt-zoom (PTZ) angle to align its field of view center with this fixed area, or adjusts the focus to clearly cover the area. In this way, when the primary tracker moves its field of view to follow the target, the auxiliary monitor can fill in the blind spots left by the primary tracker's field of view movement, or conduct predictive monitoring of the surrounding areas where the target might escape. This differentiated command generation mechanism fully embodies the idea of ​​multi-camera collaborative control, leveraging the tracking advantages of high-yield cameras while utilizing the complementary field of view characteristics of other cameras, achieving optimized allocation and efficient collaboration of monitoring resources.

[0039] The foregoing primarily describes the solutions of the embodiments of this application from the perspective of device implementation. It is understood that each device, such as an AI-based camera control device, includes at least one of the hardware structures and software modules corresponding to the execution of each function in order to achieve the aforementioned functions. Those skilled in the art should readily recognize that, based on the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein, this application can be implemented in hardware or a combination of hardware and computer software. Whether a function is executed in a hardware-driven or software-driven manner depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0040] This application embodiment can divide an AI-based camera control device into functional units based on the above method example. For example, each function can be divided into separate functional units, or two or more functions can be integrated into one processing unit. The integrated unit can be implemented in hardware or as a software functional unit. It should be noted that the unit division in this application embodiment is illustrative and only represents one logical functional division; other division methods may be used in actual implementation.

[0041] When using integrated units, Figure 3 The diagram shows a possible structure of an AI-based camera control device (referred to as communication device 30) involved in the above embodiments. The communication device 30 includes a processing unit 301 and a communication unit 302, and may also include a storage medium (referred to as storage unit 303). Figure 3 The structural diagram shown can be used to illustrate the structure of an AI-based camera control device involved in the above embodiments.

[0042] when Figure 3 The schematic diagram shown illustrates the structure of an AI-based camera control device involved in the above embodiments. The processing unit 301 is used to control and manage the actions of the AI-based camera control device, the communication unit 302 is used for the AI-based camera control device to communicate with other devices, and the storage unit 303 is used to store the program code and data of the AI-based camera control device.

[0043] For example, communication unit 302 is used to acquire the current status information of at least two cameras within the target monitoring area, and the motion trajectory of at least one target object within the field of view of the at least two cameras; wherein, the current status information includes the pan-tilt angle, focal length status, and field of view coverage. Processing unit 301: Constructs a conflict detection model based on the current state information and motion trajectory in history; identifies a combination of cameras in the target monitoring area that are tracking competition as game participants based on the conflict detection model; constructs a payoff factor based on multi-objective optimization for the game participants; determines the optimal control strategy combination based on the payoff factor; and sends control operation instructions.

[0044] Although this application has been described herein in conjunction with various embodiments, those skilled in the art, by reviewing the accompanying drawings, disclosure, and appended claims, will understand and implement other variations of the disclosed embodiments in carrying out the claimed application. In the claims, the word "comprising" does not exclude other components or steps, and "a" or "an" does not exclude multiple instances. A single processor or other unit can implement several functions listed in the claims. While different dependent claims may recite certain measures, this does not mean that these measures cannot be combined to produce good results.

[0045] Although this application has been described in conjunction with specific features and embodiments, it is obvious that various modifications and combinations can be made thereto without departing from the spirit and scope of this application. Accordingly, this specification and drawings are merely illustrative descriptions of the application as defined by the appended claims, and are considered to cover any and all modifications, variations, combinations, or equivalents within the scope of this application. Clearly, those skilled in the art can make various alterations and modifications to this application without departing from the spirit and scope of this application. Thus, if such modifications and modifications fall within the scope of the claims of this application and their equivalents, this application is also intended to include such modifications and modifications.

[0046] Some of the data in the above calculation formula are obtained by removing dimensions and taking their numerical values. The calculation formula is a calculation formula that is closest to the real situation, obtained by software simulation of a large amount of collected data. The preset parameters and preset thresholds in the calculation formula are set by those skilled in the art according to the actual situation or obtained through simulation of a large amount of data.

Claims

1. An AI-based camera control method, characterized in that, include: S1: Obtain the current status information of at least two cameras within the target monitoring area, and the motion trajectory of at least one target object within the field of view of the at least two cameras; wherein, the current status information includes the pan-tilt angle, focal length status, and field of view coverage. S2: Construct a conflict detection model based on the current state information and motion trajectory in history, and identify the combination of cameras with tracking competition in the target monitoring area as game participants based on the conflict detection model; S3: Construct a payoff factor based on multi-objective optimization for the game participants; S4: Determine the optimal combination of control strategies based on the profit factors and send control operation instructions.

2. The AI-based camera control method according to claim 1, characterized in that, The acquisition of the current status information of at least two cameras within the target monitoring area, and the motion trajectory of at least one target object within the field of view of the at least two cameras, includes: The data repository obtains the pan-tilt angle, focal length, and field of view of the cameras covering the target monitoring area, and the cameras acquire the movement trajectory of the target object in real time.

3. The AI-based camera control method according to claim 1, characterized in that, The construction of the conflict detection model based on historical current state information and motion trajectory includes: Extract current state information, movement trajectories, and game participants from the historical reference database; the historical reference data includes several current state information and movement trajectories, as well as game participants set by experts based on the current state information and movement trajectories. The current state information and motion trajectory are integrated into several sets of training data and test data; the training data is used to train the artificial intelligence model, and the test data is used to test the trained artificial intelligence model. The artificial intelligence model is adjusted according to the test results; finally, the input is the current state information and motion trajectory, and the output is the conflict detection model of the game participants; among them, the artificial intelligence model is a nonlinear regression model based on neural network, which is implemented by using BP neural network structure and / or RBF neural network structure.

4. The AI-based camera control method according to claim 1, characterized in that, The combination of cameras with tracking competition within the target monitoring area, identified by the conflict detection model, as game participants includes: The current status information of the camera corresponding to the target monitoring area and the movement trajectory of the target object are input into the conflict detection model to obtain the game participants that are tracking and competing within the target monitoring area.

5. The AI-based camera control method according to claim 1, characterized in that, The construction of multi-objective optimization-based payoff factors for the game participants includes: The cameras corresponding to the game participants are extracted sequentially, and the tracking confidence factor of the camera for the target object is obtained in real time through the calculation formula (1). ; Calculate the area of ​​the difference between the union of the field of view coverage of the camera and the field of view coverage of all other game participants excluding the camera, and use the ratio of the area of ​​the difference to the total field of view area of ​​the current camera as the uniqueness score. ; Acquire the characteristic power consumption of the camera Based on the aforementioned tracking confidence factor Uniqueness score and characteristic energy consumption The return factor is determined by formula (2). ; The calculation formula (1) is: ； In equation (1), For the number The confidence score output by the camera through a preset target detection algorithm reflects the probability of successfully identifying the target object in the field of view; The distance sensitivity coefficient is obtained through the hardware optical characteristics parameters of the camera. The distance number of the target object is The physical distance of the camera The optimal imaging reference distance is set according to the camera type; The calculation formula (2) is: ； In equation (2), The average tracking confidence factor of the game participants for the target object. The average uniqueness score of the game participants. The average characteristic energy consumption of the game participants.

6. The AI-based camera control method according to claim 5, characterized in that, The characteristic power consumption of the camera is obtained. ,include: The energy consumption of the camera when receiving control commands in history is obtained, and several energy consumption data are integrated into an energy consumption group. It is then determined whether the variance of the energy consumption group is less than the corresponding standard energy consumption variance. If so, the average value of the energy consumption in the energy consumption group is calculated to obtain the characteristic energy consumption. If not, remove the energy consumption group with the largest difference from the average energy consumption, recalculate the variance of the remaining energy consumption, and re-evaluate the variance until the variance of the remaining energy consumption in the energy consumption group is less than the corresponding standard energy consumption variance. Then, calculate the average of the remaining energy consumption in the energy consumption group to obtain the characteristic energy consumption. The standard energy consumption variance is obtained through empirical settings.

7. The AI-based camera control method according to claim 1, characterized in that, The method of determining the optimal control strategy combination based on return factors includes: Extract the payoff factors of the game participants, number the payoff factors from largest to smallest, and extract the game participants sequentially based on the numbers. When the field of vision of the extracted game participants can cover the target monitoring area, the extracted game participants are integrated into a control strategy combination.

8. The AI-based camera control method according to claim 1, characterized in that, The sending of control operation instructions includes: Real-time tracking instructions are issued to the player with the highest payoff factor in the control strategy combination, while fixed-area monitoring instructions are issued to the player with a non-maximum payoff factor in the control strategy combination; the fixed area is determined by the area where the player's field of vision overlaps with the target monitoring area.

9. An AI-based camera control device, characterized in that, The device includes: a communication unit and a processing unit; The communication unit is used to acquire the current status information of at least two cameras within the target monitoring area, and the motion trajectory of at least one target object within the field of view of the at least two cameras; wherein, the current status information includes the pan-tilt angle, focal length status, and field of view coverage. The processing unit is used to construct a conflict detection model based on the current state information and motion trajectory in history, identify the combination of cameras with tracking competition in the target monitoring area as game participants based on the conflict detection model, construct a payoff factor based on multi-objective optimization for the game participants, determine the optimal control strategy combination based on the payoff factor, and send control operation instructions.

10. This application provides a storage medium storing instructions that, when executed on an AI-based camera control device, cause the AI-based camera control device to execute an AI-based camera control method.

Images