Automatic adjustment device for coverage range in distributed multi-camera environment

Abstract

The invention provides a distributed multi-camera environment coverage area automatic adjusting device, and relates to the technical field of video monitoring and image data processing, the distributed multi-camera environment coverage area automatic adjusting device comprises a camera main body and a control optimization system, the camera main body comprises a multi-source sensing acquisition end, the bottom of the multi-source sensing acquisition end is provided with a multi-source analysis control end, and the multi-source analysis control end is connected with the control optimization system. An intelligent base is installed at the bottom end of the multi-source analysis control end, adjusting holders are installed on the two sides and the front face of the bottom of the intelligent base, and cameras are installed at the bottoms of the adjusting holders. The problem that an existing monitoring system is insufficient in adaptive capacity in a dynamic environment is solved. Through cooperative work of the multi-source sensing acquisition end and the multi-source analysis control end, the system can acquire and fuse multi-source data such as illumination, humidity, temperature and the like in real time, environment perception is performed in combination with image data, and the people flow density and the environment change trend are dynamically analyzed.

Description

Technical Field

[0001] The present invention relates to the technical fields of video surveillance and image data processing, and particularly to an automatic coverage adjustment device in a distributed multi-camera environment. Background Art

[0002] In recent years, with the continuous improvement of the needs for smart cities and public security, multi-camera surveillance systems have been widely used in fields such as urban management, traffic monitoring, and industrial park security. Through a distributed camera network, these systems can achieve comprehensive coverage of complex scenarios, improving surveillance efficiency and incident response capabilities. However, most existing multi-camera surveillance systems still have some limitations. Traditional surveillance systems usually use cameras with fixed installations, and the viewing angles and positions cannot be dynamically adjusted, making it difficult to adapt to dynamic environments such as changes in lighting, fluctuations in the density of people flow, or extreme weather, resulting in the emergence of coverage blind spots or overlapping areas and affecting the surveillance effect. In addition, existing systems have limited capabilities in multi-source data fusion and real-time analysis, and it is difficult to quickly respond to environmental changes and target dynamics. Especially in high-density crowds or complex scenarios, the accuracy of target detection and tracking decreases, and key events are easily missed.

[0003] Although some surveillance systems have introduced intelligent algorithms, such as target detection, tracking, and simple scene analysis functions, there are still the following deficiencies: First, there is a lack of an adaptive adjustment mechanism, and the cameras cannot automatically optimize the coverage range according to real-time environmental changes; second, the collaborative analysis ability between multi-source data (such as light, humidity, and image data) is insufficient, making it difficult to comprehensively evaluate the environmental state and generate effective adjustment strategies; finally, the system lacks closed-loop feedback and self-learning capabilities and cannot continuously optimize performance through historical data and user feedback, restricting its adaptability and intelligence level in dynamic environments.

[0004] Therefore, an automatic coverage adjustment device in a distributed multi-camera environment is needed to solve the above problems. Summary of the Invention

[0005] Technical Problems to be Solved Aiming at the deficiencies of the prior art, the present invention provides an automatic coverage adjustment device in a distributed multi-camera environment, which solves the problems in the above background art.

[0006] Technical Solutions

[0007] To achieve the above objectives, the present invention is implemented through the following technical solutions: A device for automatically adjusting the coverage range in a distributed multi-camera environment, including a camera main body and a control optimization system. The camera main body includes a multi-source sensing and acquisition terminal. At the bottom of the multi-source sensing and acquisition terminal, a multi-source analysis and control terminal is installed. At the bottom end of the multi-source analysis and control terminal, an intelligent base is installed. On both sides and the front of the bottom of the intelligent base, adjustment pan-tilt heads are installed. Cameras are installed at the bottom of the adjustment pan-tilt heads. Inside the multi-source sensing and acquisition terminal, a data acquisition subsystem is provided. The data acquisition subsystem includes a data acquisition module, a data preprocessing module, and a data transmission module. Inside the camera, an auxiliary acquisition subsystem is provided. The auxiliary acquisition subsystem includes an image acquisition module and an edge processing module. Inside the multi-source analysis and control terminal, an environment perception subsystem, a coverage optimization subsystem, a feedback closed-loop subsystem, and a user interaction subsystem are implanted. Inside the environment perception subsystem, a feature extraction module, a data fusion module, and a feature analysis module are provided. Inside the coverage optimization subsystem, a task scheduling module, a coverage evaluation module, a strategy optimization module, and a verification module are provided. Inside the feedback closed-loop subsystem, a feedback analysis module and a feedback analysis module are provided. Inside the user interaction subsystem, a visualization module and an interaction control module are provided. Inside the intelligent base, a dynamic adjustment subsystem is provided. The dynamic adjustment subsystem includes an instruction parsing module, a drive control module, and a collaborative scheduling module. Inside the adjustment pan-tilt head, an auxiliary adjustment system is provided. The auxiliary adjustment system includes a microcontroller module and an attitude feedback module. The data acquisition subsystem, the auxiliary acquisition subsystem, the environment perception subsystem, the coverage optimization subsystem, the feedback closed-loop subsystem, the user interaction subsystem, the dynamic adjustment subsystem, and the auxiliary adjustment system together constitute the control optimization system.

[0008] Preferably, the data acquisition module in the data acquisition subsystem collects real-time multi-source environment data around through a light sensor, a temperature sensor, and a distance sensor. The data preprocessing module denoises and standardizes and formats the collected data to generate formatted environment data. The data transmission module transmits the formatted environment data to the environment perception subsystem inside the multi-source analysis and control terminal through a low-latency communication protocol. The image acquisition module in the auxiliary acquisition subsystem inside the camera receives real-time image data from each corresponding camera. The edge processing module performs preliminary target detection and compression on the image data through an embedded chip to generate compressed image data, and transmits it to the environment perception subsystem inside the multi-source analysis and control terminal through a low-latency communication protocol. The control flow between the data acquisition subsystem and the auxiliary acquisition subsystem and the environment perception subsystem is designed as follows: The data acquisition subsystem and the auxiliary acquisition subsystem regularly collect data and push it to the environment perception subsystem. If the environment perception subsystem detects data loss and anomalies, it feeds back to the data acquisition subsystem and the auxiliary acquisition subsystem through control instructions, requiring re - collection and adjustment of the acquisition frequency.

[0009] Preferably, the environment perception subsystem performs identification and analysis on multi - source data through a multi - source analysis control terminal, specifically including: The feature extraction module receives the formatted environment data transmitted by the data acquisition subsystem and the compressed image data transmitted by the auxiliary acquisition subsystem, extracts key environmental features such as the light change rate and the pedestrian flow density distribution using a multi - modal feature extraction algorithm, generates an environmental feature set, and transmits it to the data fusion module; The data fusion module fuses the environmental feature set and the image features through Bayesian network probability inference, generates an environmental feature vector, and transmits it to the feature analysis module; The feature analysis module analyzes the environmental feature vector through a time - series prediction algorithm, identifies the environmental change trend, uses a clustering algorithm to identify hot spots, uses an anomaly detection algorithm to detect emergencies, generates an environmental status report, stores it in the local feature library of the multi - source analysis control terminal, and transmits the environmental status report to the coverage optimization subsystem; The control flow between the environment perception subsystem and the coverage optimization subsystem is designed as follows: After the environment perception subsystem generates an environmental status report, it actively pushes it to the coverage optimization subsystem. If the coverage optimization subsystem needs specific environmental features, it requests the environment perception subsystem to re - analyze and supplement feature extraction through control instructions.

[0010] Preferably, the coverage optimization subsystem is used for target detection and tracking, specifically including: The task scheduling module receives the environmental status report of the environment perception subsystem and the adjustment feedback data of the dynamic adjustment subsystem through the feedback interface, generates the monitoring task priorities of each camera using a priority - weighted scheduling algorithm, generates a task allocation table, and transmits it to the policy optimization module; The coverage evaluation module receives the environmental status report and the compressed image data of the auxiliary acquisition subsystem, quickly identifies dynamic targets in the monitoring area in combination with the YOLO algorithm, predicts and tracks the target position through the Kalman filter algorithm, performs multi - target tracking and data association through the Deep SORT algorithm, constructs a dynamic coverage efficiency model, analyzes the current camera coverage range and compares it with the preset optimal coverage range, generates a coverage efficiency heat map, and transmits it to the policy optimization module; The control flow between the coverage optimization subsystem and the environment perception subsystem is designed as follows: The coverage optimization subsystem triggers the environment perception subsystem to adjust the feature extraction strategy through the target detection result.

[0011] Preferably, the coverage optimization subsystem generates an adjustment strategy through a multi-source analysis control terminal, specifically including: The policy optimization module receives a task assignment table and a coverage efficiency heat map, dynamically adjusts the coverage range through a fuzzy logic control algorithm, optimizes the position and angle of the camera using a genetic algorithm, learns the best adjustment strategy using a reinforcement learning algorithm, generates an adjustment strategy, and transmits it to the dynamic adjustment subsystem in the intelligent base; The verification module receives the adjustment strategy, verifies it using historical scenario data and simulation environment data, generates a verification report, feeds it back to the environment perception subsystem for optimizing feature extraction, and simultaneously transmits the verification report to the feedback closed-loop subsystem; The control flow between the coverage optimization subsystem and the dynamic adjustment subsystem is designed as follows: After the coverage optimization subsystem generates an adjustment strategy, it is sent to the dynamic adjustment subsystem through a control interface. If the dynamic adjustment subsystem fails to execute or the deviation is too large, the coverage optimization subsystem is notified through a feedback interface to regenerate the strategy.

[0012] Preferably, the dynamic adjustment subsystem and the auxiliary adjustment system respectively realize the dynamic adjustment of the camera through the intelligent base and the adjustment pan-tilt, specifically including: The instruction parsing module in the intelligent base receives the adjustment strategy of the coverage optimization subsystem, parses it into specific control instructions, where the control instructions mainly include viewing angle adjustment and position offset parameters, and then transmits them to the drive control module; The drive control module drives a single adjustment pan-tilt to perform viewing angle adjustment through a stepper motor, generates an adjustment execution result, and transmits it to the collaborative scheduling module; The collaborative scheduling module coordinates the actions of multiple groups of adjustment pan-tilts through a time synchronization algorithm, generates a final adjustment execution result, and feeds it back to the coverage optimization subsystem in the multi-source analysis control terminal; The microcontroller module in the auxiliary adjustment system in the adjustment pan-tilt executes the instructions of the drive control module to perform fine viewing angle adjustment. The attitude feedback module real-time collects the adjusted angle data, generates attitude feedback data, and feeds it back to the coverage optimization subsystem; The control flow between the dynamic adjustment subsystem and the coverage optimization subsystem is designed as follows: After the dynamic adjustment subsystem executes the adjustment strategy, it pushes the adjustment execution result and the attitude feedback data to the coverage optimization subsystem. If the execution deviation exceeds the threshold, the coverage optimization subsystem requires the dynamic adjustment subsystem to readjust through a control instruction.

[0013] Preferably, the coverage optimization subsystem realizes background modeling and scene change perception, specifically including: The coverage evaluation module dynamically analyzes the scene background through a background modeling algorithm, eliminates interference factors, generates background model data, and transmits it to the policy optimization module to improve the accuracy of target detection; The coverage evaluation module analyzes the environmental status report through a time series prediction algorithm, identifies the trends of the flow of people and environmental changes, identifies hot spots using a clustering algorithm, detects emergencies using an anomaly detection algorithm, generates a scene change report, and transmits it to the policy optimization module to optimize the adjustment policy.

[0014] Preferably, the feedback closed-loop subsystem realizes dynamic optimization, specifically including: The feedback analysis module receives the user feedback from the user interaction subsystem, the verification report from the coverage optimization subsystem, and the adjustment execution result from the dynamic adjustment subsystem. After comprehensive analysis, it generates optimization instructions including data acquisition optimization instructions and policy adjustment instructions, and then issues the data acquisition optimization instructions to the data acquisition subsystem in the multi-source sensing acquisition terminal through the control interface, and issues the policy adjustment instructions to the coverage optimization subsystem; The feedback closed-loop subsystem has the ability of self-learning and continuously optimizes the algorithms and policies of each module according to historical data and user feedback; The control flow between the feedback closed-loop subsystem, the data acquisition subsystem, and the coverage optimization subsystem is designed as follows: after the feedback closed-loop subsystem generates optimization instructions, it actively issues them to the data acquisition subsystem and the coverage optimization subsystem. If the optimization effect does not meet the expectation, the data acquisition subsystem or the coverage optimization subsystem requests the feedback closed-loop subsystem to re-analyze through the feedback interface.

[0015] Preferably, the user interaction subsystem supports multi-dimensional interaction, specifically including: The visualization module receives the coverage efficiency heat map from the coverage optimization subsystem, the environmental status report from the environmental perception subsystem, and the real-time image data from the auxiliary acquisition subsystem, generates a visualized monitoring area map, and transmits it to the interaction control module; The interaction control module supports the user to adjust the task scheduling priority and coverage optimization parameters through the interaction interface, generates user instructions, and feeds them back to the feedback closed-loop subsystem.

[0016] Beneficial effects

[0017] The present invention provides a device for automatically adjusting the coverage range in a distributed multi-camera environment. It has the following beneficial effects: 1. The present invention proposes a device for automatically adjusting the coverage range in a distributed multi-camera environment, which solves the problem of insufficient adaptability of existing monitoring systems in dynamic environments. Through the collaborative work of multi-source sensing acquisition terminals and multi-source analysis and control terminals, the system can collect and fuse multi-source data such as light, humidity, and temperature in real time, perform environmental perception by combining image data, and dynamically analyze the density of people flow and the trend of environmental changes. The coverage optimization subsystem uses YOLO, Kalman filtering, and DeepSORT algorithms to achieve target detection and tracking, generates a heat map of coverage efficiency, and the intelligent pedestal and adjustment pan-tilt head dynamically adjust the viewing angle and position of the camera according to the adjustment strategy, thereby effectively eliminating coverage blind spots and overlapping areas, significantly improving the comprehensiveness and accuracy of monitoring, especially the target recognition ability in complex scenarios such as low light and heavy fog.

[0018] 2. Through the design of the environmental perception subsystem, coverage optimization subsystem, and feedback closed-loop subsystem, the present invention endows the system with powerful self-adaptive and self-learning capabilities. The feedback closed-loop subsystem can comprehensively consider user feedback, historical data, and adjustment execution results, generate optimization instructions, dynamically adjust the data acquisition frequency and task priority, and continuously optimize algorithm parameters (such as the YOLO detection threshold). This closed-loop mechanism enables the system to always maintain the best coverage effect in different scenarios (such as high-density crowds, extreme weather), improving the intelligent level of monitoring. At the same time, through the multi-source data fusion and strategy optimization module, the system comprehensively analyzes the environmental state and target dynamics, provides accurate decision-making support for monitoring personnel, and enhances the flexibility and response speed in dealing with emergencies.

[0019] 3. The present invention introduces the collaborative design of the dynamic adjustment subsystem and the auxiliary adjustment system to ensure the real-time and precise adjustment of the camera. The dynamic adjustment subsystem coordinates the actions of multiple groups of adjustment pan-tilt heads through a time synchronization algorithm, and the auxiliary adjustment system uses an attitude feedback module to correct adjustment deviations in real time. The adjustment accuracy is controlled within 0.5°, and the overall response time is controlled within 200 ms. This efficient adjustment mechanism enables the system to quickly adapt to environmental changes, such as quickly optimizing the coverage range during peak hours of people flow or in extreme weather, significantly improving the stability and real-time performance of monitoring, and providing a reliable solution for scenarios such as public safety, traffic management, and industrial park security in smart cities. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 is the overall framework diagram of the control optimization system of the present invention; Figure 2 is the operation flow chart of the control optimization system of the present invention; Figure 3 is the main structure diagram of the camera of the present invention; Figure 4 is the experimental data table of the present invention; Figure 5 This is the system simulation diagram of the present invention.

[0021] Legend Explanation: 1. Camera body; 2. Multi-source sensing and acquisition end; 3. Multi-source analysis and control end; 4. Intelligent base; 5. Adjusting pan-tilt head; 6. Camera. Specific Embodiment

[0022] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention. Specific Embodiment 1:

[0024] As Figures 1-5 shown, a device for automatically adjusting the coverage range in a distributed multi-camera environment embeds multiple adapted subsystems and modules in each component of the entire camera body 1 to form a control and optimization system, including a data acquisition subsystem, an auxiliary acquisition subsystem, an environmental perception subsystem, a coverage optimization subsystem, a dynamic adjustment subsystem, a feedback closed-loop subsystem, a user interaction subsystem, and an auxiliary adjustment system. The purpose of this device is to achieve automatic optimization of the coverage range of the distributed multi-camera system in a dynamic environment through multi-source data fusion, dynamic adjustment, and a closed-loop feedback mechanism. The following is a description of the working principle and operation details.

[0025] This device is deployed in public safety scenarios of smart cities such as city squares. The operation process starts from the initialization stage. The data acquisition subsystem in the multi-source sensing and acquisition end 2 collects environmental data such as light intensity 10 lux, temperature 5°C, and distance 3 meters in real time through data acquisition modules including a light sensor, a temperature sensor, and a distance sensor. The data preprocessing module denoises and standardizes the collected data to generate formatted environmental data. The auxiliary acquisition subsystem in the camera 6 collects real-time image data through an image acquisition module, and the edge processing module uses an embedded chip for preliminary target detection and compression to generate compressed image data. In low-light environments where the light intensity is lower than 20 lux, the data acquisition module detects insufficient light, and the edge processing module automatically adjusts the image gain and enables the infrared mode to ensure the quality of the image data. The formatted environmental data and the compressed image data are transmitted to the environmental perception subsystem in the multi-source analysis and control end 3 through a low-latency communication protocol.

[0026] The environmental perception subsystem receives formatted environmental data and compressed image data. The feature extraction module uses a multi-modal feature extraction algorithm to extract key features such as the light change rate and the pedestrian flow density distribution, and generates an environmental feature set. The data fusion module fuses the environmental feature set and the image features through Bayesian network probabilistic inference to generate an environmental feature vector. In extreme weather conditions such as foggy scenes, the light, temperature, and humidity data are fused to generate an environmental feature vector reflecting visibility, solving the problems of heterogeneity and noise interference in multi-source data fusion. The feature analysis module analyzes the environmental feature vector through a time series prediction algorithm to identify the environmental change trend (the pedestrian flow density increases by 10% per minute), uses a clustering algorithm to identify hot spots such as the center of the square, and uses an anomaly detection algorithm to detect emergencies such as crowd gatherings, generates an environmental status report, stores it in the local feature library, and pushes it to the coverage optimization subsystem. This solution ensures the robustness of the fusion result by dynamically adjusting the weights of each data source, such as increasing the weight of infrared images in low-light environments and smoothing noise through time series prediction.

[0027] The coverage optimization subsystem receives the environmental status report. The task scheduling module uses a priority weighted scheduling algorithm to generate the monitoring task priorities of each camera 6. In low-light environments, the infrared mode camera 6 is preferentially scheduled to generate a task allocation table, which is transmitted to the policy optimization module. The coverage evaluation module combines the YOLO algorithm to quickly identify dynamic targets (pedestrians and vehicles), predicts the target positions through Kalman filtering, and performs multi-target tracking and data association through Deep SORT. In heavy snow scenes, YOLO combines infrared images to optimize detection, Kalman filtering smooths the prediction through historical trajectories, and Deep SORT maintains tracking continuity through feature matching, dynamically adjusting the detection threshold to improve the low-light recognition rate. The coverage evaluation module constructs a dynamic model of coverage efficiency, analyzes the current coverage rate (80%) compared with the preset optimal coverage rate (95%), generates a coverage efficiency heat map, and transmits it to the policy optimization module. The policy optimization module dynamically adjusts the coverage range through a fuzzy logic control algorithm (if the coverage rate is low and the pedestrian flow density is high, increase the viewing angle), uses a genetic algorithm to optimize the position and angle of camera 6, and uses reinforcement learning to learn the best adjustment strategy to optimize the coverage rate through a reward function, generates an adjustment strategy, and transmits it to the dynamic adjustment subsystem. The algorithm is dynamically optimized in combination with the environmental feature vector. Kalman filtering corrects the prediction model through humidity and temperature, and reinforcement learning iteratively updates the strategy through the environmental status to adapt to dynamic requirements.

[0028] The dynamic adjustment subsystem receives the adjustment strategy. The instruction parsing module in the intelligent base 4 parses it into specific control instructions: adjust the viewing angle by 30° and offset the position by 5 cm, and then transmits them to the drive control module. The drive control module drives a single adjustment gimbal 5 to perform the viewing angle adjustment through a stepper motor. The cooperative scheduling module coordinates the actions of multiple groups of adjustment gimbals 5 through a time synchronization algorithm to avoid overlapping viewing angles and generates an adjustment execution result. The auxiliary adjustment system in the adjustment gimbal 5 executes the instructions through the microcontroller module. The attitude feedback module real-time collects the adjusted angle data such as the pitch angle of 15° and generates attitude feedback data, which is fed back to the coverage optimization subsystem. Through the time synchronization algorithm and the attitude feedback mechanism, this solution ensures the action consistency of multiple groups of adjustment gimbals 5 and controls the adjustment delay within 50 ms, solving the technical difficulties of multi-camera cooperation and high real-time requirements.

[0029] The feedback closed-loop subsystem receives the user feedback from the user interaction subsystem (coverage of hot spots), the verification report of the coverage optimization subsystem (coverage rate increased to 90%), and the adjustment execution result deviation of the dynamic adjustment subsystem (less than 2°). After comprehensive analysis, it generates an optimization instruction to adjust the acquisition frequency to 5 times per second and update the task priority, and issues it to the data acquisition subsystem and the coverage optimization subsystem through the control interface. In low-light environments, the optimization instruction increases the infrared data acquisition frequency, and the coverage optimization subsystem re-adjusts the priority of the camera 6. The feedback closed-loop subsystem has self-learning ability. It iteratively optimizes the algorithm parameters through historical data and user feedback (the YOLO detection threshold is adjusted from 0.5 to 0.6) to improve adaptability. Through low-latency communication and priority-weighted scheduling, it optimizes the feedback path and controls the delay within 100 ms, solving the technical difficulties of feedback delay and optimization conflicts.

[0030] The user interaction subsystem displays the coverage efficiency heat map, environmental status report, and real-time image data through the visualization module. The user adjusts the task scheduling priority through the interaction control module (raising the priority of the hot spot area to the highest and setting the coverage rate target of the coverage optimization parameter to 98%), generates a user instruction, and feeds it back to the feedback closed-loop subsystem. In a rainstorm scenario, the user views the coverage efficiency heat map, discovers the blind area, and adjusts the strategy, and the system automatically optimizes the adjustment.

[0031] The operation process includes initializing to collect initial data to generate an environmental status report, real-time monitoring, dynamically scheduling tasks to adjust the perspective, dynamically responding to switch to the infrared mode in low-light or foggy environments, optimizing object detection, user intervention for users to adjust parameters, system response, and continuous optimization feedback loop to iteratively optimize the strategy. The specific scenario optimization mechanism includes: optimizing the YOLO detection threshold through the infrared mode and light data in low-light environments, and the edge processing module enhances the image contrast; in foggy scenarios, adjusting the Kalman filter prediction model in combination with humidity and temperature, increasing the multi-source data fusion weight, and preferentially adjusting the camera with the best visibility 6. In practical applications, the coverage rate is increased from 80% to 95%, the target tracking accuracy reaches 98%, and the response time is controlled within 200 ms, showing strong adaptability and stability. Specific Embodiment Two:

[0033] Such as Figures 1-5As shown, this device realizes the control of the camera through the organic cooperation of various components. It can not only perform single adjustment to meet local requirements, but also optimize the overall coverage effect through coordinated operations to ensure efficient operation in complex environments. The following details the control mechanism, the specific implementation of single and coordinated control, and how to ensure better effects. Overview of the control mechanism: Each component completes the camera control task through the division of labor and cooperation of subsystems and modules in the control optimization system. The multi-source sensing acquisition end is responsible for environmental data acquisition, the multi-source analysis and control end conducts data analysis and strategy generation, the intelligent base and the adjustment pan-tilt execute adjustment instructions, the camera acquires images and feedbacks the status, the camera body serves as the overall bearing structure, and the multi-source analysis and control end serves as the core decision-making unit, coordinating the functions of each component to ensure the accuracy of single control and the efficiency of coordinated control. Implementation of single control: Single control focuses on the independent adjustment of a single camera to cope with local environmental changes or specific task requirements. The implementation process is that the data acquisition subsystem in the multi-source sensing acquisition end collects local environmental data (such as the light intensity of a certain area is 15 lux) through light, temperature, and distance sensors. The auxiliary acquisition subsystem in the camera acquires images of this area through the image acquisition module. The edge processing module conducts preliminary target detection and generates compressed image data. These data are transmitted to the multi-source analysis and control end through a low-latency communication protocol. The environmental perception subsystem in the multi-source analysis and control end analyzes the data, extracts key features (such as insufficient light), and generates an environmental status report. The task scheduling module of the coverage optimization subsystem identifies the camera that needs to be independently adjusted according to the report, generates a specific task priority, and the coverage evaluation module combines the YOLO algorithm to identify dynamic targets (such as pedestrians) in the area, predicts the target position through Kalman filtering, and generates a local coverage efficiency heat map. The strategy optimization module generates a single adjustment strategy (such as adjusting the viewing angle of this camera to 30° to cover the target area) through the fuzzy logic control algorithm according to the heat map. The adjustment strategy is transmitted to the dynamic adjustment subsystem in the intelligent base, and the instruction parsing module parses it into a single control instruction (viewing angle adjustment of 30°). The drive control module drives the corresponding adjustment pan-tilt through a stepper motor. The auxiliary adjustment system in the adjustment pan-tilt executes the instruction through the microcontroller module. The attitude feedback module real-time collects the adjusted angle (actually adjusted to 29.8°) and generates attitude feedback data, which is fed back to the coverage optimization subsystem to ensure adjustment accuracy. Single control ensures that the camera can independently cope with local requirements through precise environmental perception and strategy optimization. For example, in a low-light environment, the infrared mode can be enabled separately to avoid affecting the working status of other cameras. The attitude feedback mechanism ensures that the adjustment deviation is controlled within 0.5°, improving the accuracy of single control.Implementation of Coordinated Control: Coordinated control aims to enable multiple cameras to work together to optimize the overall coverage, avoid overlaps or blind spots, and is particularly suitable for scenarios with large-scale dynamic environmental changes. The implementation process is as follows: The multi-source sensing acquisition end collects global environmental data (such as the overall light intensity of 20 lux and humidity of 80%), all cameras collect real-time images, generate compressed image data, and uniformly transmit it to the multi-source analysis and control end. The environmental perception subsystem fuses multi-source data to generate a global environmental feature vector. The feature analysis module uses a time series prediction algorithm to identify the overall environmental change trend (such as the flow of people moving eastward), generates a global environmental status report. The task scheduling module of the coverage optimization subsystem assigns task priorities to all cameras according to the global environmental status report using a priority weighted scheduling algorithm (for example, the priority of the east-side camera is increased), generates a global task allocation table. The coverage evaluation module combines the Deep SORT algorithm to track multi-target dynamics (such as the movement trajectory of the crowd), constructs a global coverage efficiency dynamic model, generates a coverage efficiency heat map, and identifies coverage blind spots (such as the coverage rate in the southeast corner is only 60%). The strategy optimization module optimizes the angles and positions of all cameras through a genetic algorithm, learns the global optimal adjustment strategy through a reinforcement learning algorithm, and generates a coordinated adjustment strategy (for example, the viewing angle of the east-side camera is expanded to 45°, and the south-side camera is offset by 5 cm). The adjustment strategy is transmitted to the dynamic adjustment subsystem in the intelligent base, and the instruction parsing module parses it into multiple groups of control instructions. The coordinated scheduling module coordinates the actions of multiple groups of adjustment pan-tilts through a time synchronization algorithm (for example, the east-side adjustment pan-tilt starts to adjust at t = 0 ms, and the south-side adjustment pan-tilt starts at t = 10 ms) to avoid adjustment conflicts. The auxiliary adjustment system in the adjustment pan-tilt executes the instructions, and the attitude feedback module real-time collects the adjusted angles of each camera and feeds them back to the coverage optimization subsystem to ensure the consistency of coordinated actions. Coordinated control is optimized through a time synchronization algorithm and a global coverage efficiency heat map to ensure that the viewing angles of multiple cameras do not overlap and the coverage rate is maximized. The coordinated scheduling module monitors the status of each adjustment pan-tilt in real time. If the adjustment delay of a certain camera exceeds 20 ms, other cameras are preferentially adjusted to make up for the coverage, and the overall adjustment delay is controlled within 100 ms. Combination Mechanism of Single and Coordinated Control: The combination of single control and coordinated control is achieved through the dynamic switching of the coverage optimization subsystem of the multi-source analysis and control end. The task scheduling module judges the current demand according to the environmental status report. If a local area (such as a certain hotspot) needs to be independently optimized, single control is triggered. If the global coverage rate is insufficient (such as the overall coverage rate is lower than 90%), it switches to coordinated control. The two modes are seamlessly connected through the coordinated scheduling module of the dynamic adjustment subsystem. For example, during the coordinated control process, if a certain camera detects an unexpected event (such as a crowd gathering), the task scheduling module temporarily switches to single control, preferentially adjusts the viewing angle of this camera, and then re-integrates into the coordinated control to ensure the overall coverage effect.Control Optimization in Specific Scenarios: In low-light environments (light intensity below 20 lux), single control preferentially enables cameras in infrared mode, and coordinated control adjusts the working modes of all cameras through global lighting data to avoid resource waste. In extreme weather scenarios (such as heavy fog with visibility below 5 meters), coordinated control preferentially schedules the camera with the best visibility and optimizes the overall view through a global coverage efficiency heat map. Single control focuses on hotspots (such as crowded areas), and precisely adjusts by combining infrared data and humidity data to ensure that the target is not lost. Technical Effects and Advantages: Through the combination of single control and coordinated control, the accuracy of this device within single adjustment is within 0.5°, the coverage rate during coordinated adjustment is increased to over 95%, and the overall response time is controlled within 200 ms. Compared with the prior art, single control avoids adjustment deviation through attitude feedback and precise strategy optimization, and coordinated control eliminates coverage blind spots and overlapping areas through time synchronization and global heat map optimization, ensuring an efficient coverage effect in complex dynamic environments. The feedback closed-loop subsystem further optimizes the control strategy through historical data iteration, improving the stability and adaptability during long-term operation. Specific Embodiment Three: As Figures 1-5 shown, the key algorithms mentioned in Embodiment One are analyzed in detail below: In the entire system, the YOLO algorithm is used to achieve real-time object detection and multi-object tracking. Its speed and efficiency are suitable for the dynamic requirements of the monitoring system. The YOLO algorithm can directly output the category and location of objects (such as pedestrians, vehicles, etc.), and adjust the detection accuracy and speed of each target according to the system's needs. The intersection over union (IOU) is used as an index to measure the overlap degree between the detection box and the ground truth box, enabling this module, that is, the coverage evaluation module, to maintain accurate detection in high-density scenarios.

[0035] The mathematical formula of the YOLO algorithm is as follows:

[0036] Where is the grid size (for example, a grid of size S is used to divide the image); is the number of bounding boxes predicted for each grid, is the intersection over union, used to measure the overlap degree between the predicted box and the ground truth box; and are weight coefficients, used to balance the coordinate loss and classification loss; is the ground truth value, used to indicate whether the target exists within the grid; is the predicted probability, representing the possibility of the target category.

[0037] In the coverage evaluation module, Kalman filtering is used to predict and track the target position.

[0038] The Kalman filtering algorithm formula is as follows: State transition equation:

[0039] Observation equation:

[0040] Where is the state vector at the current time (such as the position and velocity of the target); State transition matrix, describing how the system state transfers from time to time ; is the control input matrix, describing how the control input affects the state; is the control input (such as acceleration and external force); is the process noise, is the current observation value, is the observation matrix that maps the state space to the observation space, is the observation noise.

[0041] For the multi-target tracking requirements that may exist in the system, the Deep SORT algorithm adds the Hungarian algorithm on the basis of Kalman filtering to achieve target association through distance measurement. This algorithm effectively solves problems such as multiple target overlaps and occlusions in complex environments.

[0042]

[0043] Deep SORT combines the detection and feature extraction modules to perform Euclidean distance measurement on the features of the target, so as to match and associate multiple targets. The camera inputs the features of multiple targets into this module, tracks the movement paths of different targets and ensures that there are no mis-tracking or missed-tracking situations.

[0044] Where is the distance measurement between target and target ; is the center coordinate of target ; is the height and width of target ; ; is the target The corresponding parameters.

[0045] The time series prediction algorithm analyzes the environmental status report to identify the trends of the flow of people and environmental changes. The time series prediction algorithm can also detect the trends of changes.

[0046] The specific formula is as follows:

[0047] Where The observed value at the current moment; ; The constant term, representing the basic level; The autoregressive coefficient, representing the influence of the previous moment's value on the current value; The white noise (error term) at the current moment; The moving average coefficient, representing the influence of the previous moment's error on the current value.

[0048] The K-means clustering algorithm is used to identify the hot spots in the monitored area. For example, in the crowded area during the peak period of the flow of people, the system can analyze the specific patterns in the area based on background modeling, automatically identify the areas with high frequency, and cluster the location points in the monitoring data through the K-means algorithm to automatically identify the hot spots.

[0049] The system divides the data into classes, making the samples in the same class closer and the distance between classes farther.

[0050]

[0051] Where : The sample set of the th class; : The center of the th class; : Represents the square distance from the sample to the class center.

[0052] The anomaly detection algorithm is used in the environmental change analysis module to detect sudden events or abnormal activities in the area, such as sudden congestion of people or aggregation of vehicles. Through methods such as Z-score, the system can detect the phenomenon that the behavior in the monitored area deviates from the normal range. The anomaly detection algorithm can set reasonable thresholds, such as the density or the flow of people in the area monitored by the camera. When the Z-score exceeds the preset value, the system will trigger an alarm signal and adjust the camera angle to observe the specific situation of the sudden event more clearly.

[0053] The specific formula is as follows:

[0054] Where Current observed value Mean value of the observed values Standard deviation Z-score, if is usually regarded as an anomaly

[0055] The Bayesian network plays a role in probabilistic reasoning in the data fusion module. By combining multiple sensor data, when the system detects a possible abnormal situation, it can calculate the occurrence probability of an event in the Bayesian inference way to help the system determine the countermeasures to be taken. After receiving the multi-sensor data of an emergency, the system calculates the probability of the best response plan through Bayesian inference. For example, when there is an abnormal crowd gathering in a certain area, the Bayesian network analyzes the previous data (such as time, historical flow, etc.) to infer whether the nature of the event belongs to a dangerous event or a temporary event, and then provides the reliability to support decision-making

[0056] The specific formula is as follows

[0057] where : Posterior probability of given when ; : Likelihood probability of given when ; Prior probability of Marginal probability of Specific embodiment four As Figures 1-5 shown, the following is the detailed hardware composition and hardware description in each module of an automatic coverage adjustment device in a distributed multi-camera environment Data Acquisition Subsystem (located at the multi-source sensing acquisition end): Responsible for collecting multi-source environmental data and providing basic data support. The hardware components include a data acquisition module, a data preprocessing module, and a data transmission module. The data acquisition module includes a light sensor (model: TSL2561, accuracy ±0.1 lux, working range 0 - 60,000 lux), a temperature sensor (model: DHT22, accuracy ±0.5°C, working range -40°C to 80°C), a distance sensor (model: VL53L0X, accuracy ±3mm, working range 0 - 2m), a microcontroller (model: STM32F103, 32-bit ARM Cortex-M3, operating frequency 72MHz), and an ADC converter (model: ADS1115, 16-bit accuracy, sampling rate 860 SPS). The light, temperature, and distance sensors communicate with the microcontroller through the I2C interface. The microcontroller triggers the sensors to collect data at regular intervals. The ADC converter converts the analog signal into a digital signal, generates the raw data and stores it in the internal buffer (256KB SRAM), and transmits it to the data preprocessing module through the SPI interface. The data preprocessing module includes a digital signal processor (DSP, model: TMS320C6748, operating frequency 456MHz) and a memory (model: AT45DB161D, 16Mb Flash). The DSP obtains the raw data from the microcontroller, performs denoising (median filtering) and normalization (normalized to the 0 - 1 range), generates the formatted environmental data and stores it in the Flash, and transmits it to the data transmission module through the high-speed SPI interface. The data transmission module includes a wireless communication module (model: nRF24L01, 2.4GHz, rate 2Mbps), a microcontroller (model: ESP32, built-in Wi-Fi), and a cache memory (model: IS62WV51216, 8Mb SRAM). The ESP32 obtains the formatted environmental data from the DSP, encapsulates it into data packets, and transmits it to the multi-source analysis control end through the nRF24L01. The cache memory ensures that data is not lost during transmission interruptions and is connected to the environmental perception subsystem through wireless communication.

[0059] Auxiliary Acquisition Subsystem (located in the camera): Responsible for collecting image data and performing preliminary processing to provide basic data for target detection. The hardware components include an image acquisition module and an edge processing module. The image acquisition module includes a high-definition camera (model: OV5640, 5 million pixels, supporting 1080p, frame rate 30fps, with infrared mode), a lens (focal length: 4mm, field of view angle 90°), and a microcontroller (model: ESP32-CAM, built-in Wi-Fi). OV5640 is connected to ESP32-CAM through the MIPI interface, supporting the switching between normal light and infrared modes (automatically switched in low-light environments). The collected images are stored in the internal buffer (4MB PSRAM) and transmitted to the edge processing module through the I2C interface. The edge processing module includes an embedded GPU (model: NVIDIA Jetson Nano, 128-core Maxwell GPU, inference speed 10fps), a memory (model: SD card, 32GB), and a microcontroller (model: Raspberry Pi Pico). Jetson Nano obtains image data from ESP32-CAM, runs a lightweight YOLO model for preliminary target detection (identifying pedestrians and vehicles), generates target position data, and Raspberry Pi Pico performs JPEG compression on the images (compression ratio 80%) to generate compressed image data, which is transmitted to the environmental perception subsystem of the multi-source analysis control terminal through the Wi-Fi module (built-in in ESP32).

[0060] Environmental Perception Subsystem (Located at the Multi-source Analysis Control Terminal): Analyze multi-source data and generate an environmental status report. The hardware components include a feature extraction module, a data fusion module, and a feature analysis module. The feature extraction module includes an FPGA (model: Xilinx Zynq-7000, dual-core ARM Cortex-A9, operating frequency 800 MHz) and a memory (model: DDR3, 2 GB). The FPGA receives formatted environmental data and compressed image data from the data acquisition subsystem and the auxiliary acquisition subsystem, runs a multi-modal feature extraction algorithm (convolutional neural network), extracts features such as the light change rate and the pedestrian flow density distribution, generates an environmental feature set and stores it in the DDR3, and transmits it to the data fusion module through the AXI bus. The data fusion module includes a server processor (model: Intel Xeon E5-2620, 6 cores and 12 threads, operating frequency 2.1 GHz) and a memory (model: NVMe SSD, 512 GB). The Xeon processor receives the environmental feature set from the FPGA, runs a Bayesian network probability inference algorithm, dynamically adjusts the weights of each data source (the weight of the infrared in low-light environments is increased to 0.7), fuses and generates an environmental feature vector and stores it in the NVMe SSD, and transmits it to the feature analysis module through the PCIe interface. The feature analysis module includes an embedded AI chip (model: NVIDIA Jetson TX2, 256-core Pascal GPU) and a memory (model: eMMC, 32 GB). The Jetson TX2 receives the environmental feature vector from the Xeon processor, runs an LSTM time series prediction algorithm to analyze trends (the pedestrian flow density increases by 5%), a K-means clustering algorithm to identify hot spots (K = 3), and a Z-score anomaly detection algorithm to mark emergencies (Z > 3), generates an environmental status report and stores it in the eMMC, and transmits it to the coverage optimization subsystem through Ethernet.

[0061] Coverage Optimization Subsystem (Located at the Multi-source Analysis Control Terminal): Optimize the camera coverage range and generate adjustment strategies. The hardware components include a task scheduling module, a coverage evaluation module, a policy optimization module, and a verification module. The task scheduling module includes a microprocessor (model: ARM Cortex-A72, 4 cores, operating frequency 1.5 GHz) and a memory (model: DDR4, 4GB). The ARM Cortex-A72 receives the environmental status report from the environmental perception subsystem, runs the priority weighted scheduling algorithm, generates a task allocation table (priority of the east-side camera is 0.9), stores it in the DDR4, and transmits it to the policy optimization module through a high-speed bus. The coverage evaluation module includes a GPU server (model: NVIDIA RTX 2080 Ti, 4352 CUDA cores) and a memory (model: GDDR6, 11GB). The RTX 2080 Ti receives data from the environmental perception subsystem and the auxiliary acquisition subsystem, runs YOLO to identify targets (pedestrians, vehicles), uses Kalman filtering to predict positions, and Deep SORT to associate multiple targets, generates a coverage efficiency heat map (coverage rate 80%), stores it in the GDDR6, and transmits it to the policy optimization module through PCIe. The policy optimization module includes a high-performance computing cluster (model: Intel Core i9-9900K, 8 cores and 16 threads) and a memory (model: NVMe SSD, 1TB). The i9-9900K receives the task allocation table and the coverage efficiency heat map, runs fuzzy logic control to adjust the coverage range, genetic algorithm to optimize the camera position, reinforcement learning to learn the best strategy, Bayesian network to evaluate the success probability, and decision tree to generate the final adjustment strategy, stores it in the NVMe SSD, and transmits it to the dynamic adjustment subsystem through Ethernet. The verification module includes a microprocessor (model: ARM Cortex-A53, 4 cores, operating frequency 1.2 GHz) and a memory (model: DDR3, 2GB). The ARM Cortex-A53 receives the adjustment strategy, runs historical scenario and simulation environment verification (coverage rate increased to 95%), generates a verification report, stores it in the DDR3, and transmits it to the feedback closed-loop subsystem through Ethernet.

[0062] Dynamic adjustment subsystem (located in the intelligent base): Executes adjustment strategies to adjust the camera's perspective and position. The hardware components include an instruction parsing module, a drive control module, and a cooperative scheduling module. The instruction parsing module contains a microcontroller (model: STM32F407, 32-bit ARM Cortex-M4, operating frequency 168 MHz) and a memory (model: EEPROM, 512 KB). STM32F407 receives the adjustment strategy from the coverage optimization subsystem, parses it into control instructions (adjust the perspective by 45°), stores them in the EEPROM, and transmits them to the drive control module through the SPI interface. The drive control module contains a stepper motor (model: NEMA 17, step angle 1.8°), a motor driver (model: DRV8825, current 2.5 A), and a microcontroller (model: Arduino Nano). Arduino Nano receives the control instructions, drives the NEMA 17 stepper motor through DRV8825 to execute the adjustment (accuracy 0.1°), records the adjustment time (delay <50 ms), and transmits the adjustment execution result to the cooperative scheduling module through the I2C interface. The cooperative scheduling module contains a microcontroller (model: ESP32, built-in Wi-Fi) and a memory (model: Flash, 4 MB). ESP32 receives multiple groups of adjustment execution results, runs a time synchronization algorithm (NTP protocol), coordinates the actions of multiple groups of adjustment gimbals (starts at t = 0 ms on the east side and t = 10 ms on the south side), stores the results in the Flash, and transmits the final adjustment execution result to the coverage optimization subsystem through Wi-Fi.

[0063] Auxiliary adjustment system (located on the adjustment gimbal): Assists in executing adjustment instructions and provides attitude feedback. The hardware components include a microcontroller module and an attitude feedback module. The microcontroller module contains a microcontroller (model: ATmega328P, 8-bit AVR, operating frequency 16 MHz) and a memory (model: EEPROM, 1 KB). ATmega328P receives instructions from the drive control module, performs fine-tuning of the perspective (accuracy 0.1°), records the energy consumption data and stores it in the EEPROM, and communicates with the attitude feedback module through the I2C interface. The attitude feedback module contains a gyroscope sensor (model: MPU-6050, 6-axis, accuracy ±0.05°) and a microcontroller (model: Raspberry Pi Pico). MPU-6050 collects the adjusted angle data (pitch angle 45.2°) through the I2C interface, Raspberry Pi Pico generates attitude feedback data, checks the deviation (<0.5°), and transmits it to the coverage optimization subsystem through the SPI interface.

[0064] Feedback closed-loop subsystem (located at the multi-source analysis control end): Optimize the operation of the system through comprehensive feedback. The hardware components include a feedback analysis module, which consists of a server processor (model: Intel Core i7-9700, 8 cores, operating frequency 3.0 GHz) and a memory (model: NVMe SSD, 512 GB). The i7-9700 receives user feedback, verification reports, and adjustment execution results, runs analysis algorithms to generate optimization instructions (acquisition frequency up to 10 times per second), stores them in the NVMe SSD, and transmits the optimization instructions to the data acquisition and coverage optimization subsystem via Ethernet.

[0065] User interaction subsystem (located at the multi-source analysis control end): Provide user interaction and visualization support. The hardware components include a visualization module and an interaction control module. The visualization module consists of a graphics processing unit (model: NVIDIA GTX 1660, 1408 CUDA cores) and a monitor (model: Dell UltraSharp, 4K resolution). The GTX 1660 receives the coverage efficiency heat map, environmental status report, and real-time image data, generates a visual mapping and displays it on the Dell UltraSharp monitor, and outputs it to the interaction control module via the HDMI interface. The interaction control module consists of a touch screen (model: Elo Touch, 15 inches) and a microprocessor (model: Raspberry Pi 4, 4 cores). The Raspberry Pi 4 receives user input (task priority 0.9) through the Elo Touch touch screen, generates user instructions, and transmits them to the feedback closed-loop subsystem via Ethernet. Specific embodiment five: As Figures 1-5 shown, the following provides specific use cases: Case 1: Monitoring of the crowd at a smart city public square at night Scenario description: A festival is being held at a city center square at night. The crowd is dense and the light intensity is low (10 lux). It is necessary to monitor the crowd distribution and ensure safety to avoid overcrowding or emergencies.

[0067] Device deployment: Deploy a distributed multi-camera device at the center of the square, which includes a multi-source sensing and acquisition end, a multi-source analysis control end, a smart pedestal, and three cameras (one each on the east, south, and west sides). The viewing angle can be adjusted by adjusting the pan-tilt head.

[0068] Usage process: Data Acquisition and Environment Perception: The data acquisition subsystem within the multi-source sensing acquisition terminal detects low-light environments (10 lux) through a light sensor, records temperature (5°C) with a temperature sensor, and measures the distance of the crowd (average 3 meters) with a distance sensor. After denoising and normalizing the data, the data preprocessing module transmits it to the multi-source analysis control terminal. The auxiliary acquisition subsystem within the camera switches to the infrared mode to acquire real-time images. The edge processing module uses an embedded chip (NVIDIA Jetson Nano) for preliminary target detection, generates compressed image data, and transmits it. The feature extraction module of the environment perception subsystem extracts the light change rate and the distribution of pedestrian flow density. The data fusion module generates an environment feature vector through Bayesian network probability inference. The feature analysis module predicts the pedestrian flow trend (increasing by 15% per minute) through a time series prediction algorithm (LSTM), identifies hotspots (the center of the square) using the K-means clustering algorithm, and no abnormalities are detected by the Z-score anomaly detection. An environment status report is generated and pushed to the coverage optimization subsystem.

[0069] Target Detection and Coverage Optimization: The task scheduling module of the coverage optimization subsystem, based on the environment status report, preferentially schedules the infrared mode camera on the east side (priority 0.9) and generates a task allocation table. The coverage evaluation module uses the YOLO algorithm combined with infrared data to quickly identify dynamic targets (pedestrians), adjusts the detection threshold (from 0.5 to 0.6), predicts the target position using Kalman filtering, associates multiple targets with Deep SORT, and generates a coverage efficiency heat map (coverage rate 75%). The strategy optimization module, through a fuzzy logic control algorithm (if the coverage rate is low and the pedestrian flow density is high, increase the viewing angle), optimizes the camera position using a genetic algorithm (expand the viewing angle on the east side to 50°), generates an adjustment strategy through reinforcement learning, and transmits it to the dynamic adjustment subsystem.

[0070] Dynamic Adjustment: The instruction parsing module of the dynamic adjustment subsystem parses the adjustment strategy into control instructions (adjust the viewing angle on the east side by 50°). The drive control module drives the adjustment pan-tilt through a stepper motor (NEMA 17). The cooperative scheduling module coordinates the actions of the three groups of adjustment pan-tilts through a time synchronization algorithm (start on the east side at t = 0ms and start on the south side at t = 10ms). The auxiliary adjustment system within the adjustment pan-tilt performs fine-tuning. The attitude feedback module collects the adjusted angle (49.8°) and the feedback deviation (0.2°) and sends them back to the coverage optimization subsystem.

[0071] Feedback and Optimization: The feedback closed-loop subsystem receives the adjustment execution results (coverage rate increased to 90%) and user feedback (focus on hotspots), generates optimization instructions (increase the infrared data acquisition frequency to 10 times per second), and issues them to the data acquisition subsystem and the coverage optimization subsystem. The user interaction subsystem displays the coverage efficiency heat map through a visualization module, and the user adjusts the task priority (increase the priority of the hotspot area to 0.95) through the interaction control module.

[0072] Usage effect: The night coverage rate is increased from 75% to 90%, the target tracking accuracy reaches 98%, the response time is 200 ms, it can successfully cope with low-light environments, and ensure no blind spots in crowd safety monitoring.

[0073] Case 2: Vehicle monitoring in the parking lot of a commercial complex (foggy weather) Scenario description: In the winter foggy weather (visibility 5 meters, humidity 90%), the parking lot of a commercial complex needs to monitor vehicle entry and exit to prevent congestion and accidents.

[0074] Device deployment: One device is deployed at each of the parking lot entrance and exit. The multi-source sensing acquisition terminal collects environmental data, the multi-source analysis and control terminal analyzes the data, and the intelligent base coordinates two cameras (entrance and exit).

[0075] Usage process: Data collection and environmental perception: The multi-source sensing acquisition terminal collects data through a humidity sensor (humidity 90%), a temperature sensor (0°C), and a light sensor (20 lux). After being standardized by the data preprocessing module, the data is transmitted. The camera switches to the infrared mode, the image acquisition module acquires images, and after being compressed by the edge processing module, the images are transmitted. The environmental perception subsystem extracts visibility features through the feature extraction module. The data fusion module uses the Bayesian network to adjust the weight of infrared data (0.8). The feature analysis module predicts the vehicle flow trend (increasing by 20 vehicles per hour) through the time series prediction algorithm, identifies the hot spot area (at the entrance) by K-means, detects anomalies (entrance congestion, Z>3) by Z-score, and generates an environmental status report.

[0076] Target detection and coverage optimization: The task scheduling module of the coverage optimization subsystem preferentially schedules the entrance camera (priority 0.9). The coverage evaluation module uses YOLO combined with infrared data to identify vehicles, combines Kalman filtering with humidity to correct the prediction model, and Deep SORT associates vehicle trajectories to generate a coverage efficiency heat map (coverage rate 70%). The strategy optimization module uses fuzzy logic control (low coverage rate and high traffic, increase the viewing angle), and the genetic algorithm optimizes the viewing angle (adjust the viewing angle of the entrance camera to 60°), and transmits the adjustment strategy to the dynamic adjustment subsystem.

[0077] Dynamic adjustment: The instruction parsing module parses the instruction (entrance viewing angle 60°), the drive control module drives and adjusts the pan-tilt, the cooperative scheduling module coordinates the actions of the entrance and exit cameras, the attitude feedback module collects the adjusted angle (59.7°), and feeds back the deviation (0.3°).

[0078] Feedback and Optimization: The feedback closed-loop subsystem receives the adjustment result (coverage rate increased to 92%), generates an optimization instruction (increase the humidity data collection frequency), the user interaction subsystem displays the coverage efficiency heat map, and the user adjusts the entrance priority (0.95).

[0079] Usage Effect: In a foggy environment, the coverage rate is increased from 70% to 92%, the vehicle tracking accuracy is 96%, the response time is 180 ms, successfully coping with low visibility scenarios and preventing entrance congestion.

[0080] Case 3: Security Monitoring at the Entrance and Exit of an Industrial Park (During Peak Pedestrian Flow) Scenario Description: An industrial park needs to monitor the entrance and exit during peak commuting hours (pedestrian flow density increases by 30% per minute) and identify abnormal behaviors (such as running fast).

[0081] Device Deployment: One device is deployed at the entrance and exit. The intelligent base controls two cameras (main entrance and side door), and the multi-source analysis control terminal analyzes the pedestrian flow data.

[0082] Usage Process: Data Collection and Environmental Sensing: The multi-source sensing collection terminal collects light (50 lux), temperature (20°C), distance (2 meters), the camera collects images, and the edge processing module compresses and transmits them. The environmental sensing subsystem extracts the pedestrian flow density features, the data fusion module generates the environmental feature vector, the feature analysis module predicts the pedestrian flow trend (increase by 30%) through the time series prediction algorithm, K-means identifies the hot spot area (main entrance), Z-score detects abnormalities (running fast, Z>3), and generates an environmental status report.

[0083] Object Detection and Coverage Optimization: The task scheduling module preferentially schedules the main entrance camera (priority 0.9). The coverage evaluation module uses YOLO to identify pedestrians, Kalman filter to predict positions, Deep SORT to associate trajectories, and generates a coverage efficiency heat map (coverage rate 85%). The strategy optimization module adjusts the perspective through fuzzy logic control (low coverage rate and high density, increase the viewing angle), and genetic algorithm adjusts the viewing angle (main entrance viewing angle 50°), and transmits the adjustment strategy.

[0084] Dynamic Adjustment: The instruction parsing module parses the instruction (viewing angle 50°), the drive control module drives the adjustment pan-tilt, the collaborative scheduling module coordinates the actions of the two cameras, the attitude feedback module collects the angle (49.9°), and the feedback deviation is (0.1°).

[0085] Feedback and Optimization: The feedback closed-loop subsystem receives the result (coverage rate increased to 94%), generates an optimization instruction (increase the collection frequency), the user interaction subsystem displays the heat map, and the user adjusts the priority (0.9).

[0086] Usage effect: The peak - period coverage rate is increased from 85% to 94%, the accuracy rate of abnormal behavior detection is 95%, the response time is 150 ms, ensuring efficient security monitoring.

[0087] Case 4: Monitoring of school playground activities (rainy weather) Scenario description: A school sports meeting is held on the school playground in rainy weather (humidity 95%, visibility 10 m). It is necessary to monitor the distribution of students to prevent accidents.

[0088] Device deployment: Two devices are deployed around the playground. The intelligent base controls four cameras (one each in the east, south, west, and north), and the multi - source analysis control terminal analyzes the data.

[0089] Usage process: Data collection and environmental perception: The multi - source sensing collection terminal collects humidity (95%), temperature (15 °C), and light (30 lux). The camera switches to the infrared mode, and the edge processing module compresses the image and then transmits it. The environmental perception subsystem extracts visibility features, the data fusion module adjusts the infrared weight (0.85), the feature analysis module predicts the trend of the flow of people (a 10% decrease), K - means identifies the hot spot (the podium), Z - score does not detect abnormalities, and an environmental status report is generated.

[0090] Object detection and coverage optimization: The task scheduling module preferentially schedules the podium camera (priority 0.9). The coverage evaluation module uses YOLO to identify students, the Kalman filter corrects the prediction model, Deep SORT associates the trajectories, and a coverage efficiency heat map is generated (coverage rate 80%). The strategy optimization module adjusts the viewing angle through fuzzy logic control (the podium viewing angle is 55°) and transmits the adjustment strategy.

[0091] Dynamic adjustment: The instruction parsing module parses the instruction (viewing angle 55°), the drive control module drives and adjusts the pan - tilt head, the cooperative scheduling module coordinates the actions of the four cameras, the attitude feedback module collects the angle (54.8°), and the feedback deviation is 0.2°.

[0092] Feedback and optimization: The feedback closed - loop subsystem receives the result (the coverage rate is increased to 93%), generates an optimization instruction (increase the humidity collection frequency), the user interaction subsystem displays the heat map, and the user adjusts the priority (0.9).

[0093] Usage effect: In the rainy environment, the coverage rate is increased from 80% to 93%, the target tracking accuracy rate is 97%, the response time is 190 ms, ensuring the safety of the activity. Specific embodiment six: As Figures 1-5 shown, Figure 4 Provide complete experimental data, showing the performance comparison before and after optimization of each scenario. Figure 5It is the simulation diagram of the system. The horizontal axis (time step): It represents 5 time steps (from 1 to 5) of the simulation run, and each time step represents a system adjustment cycle. The vertical axis (coverage rate %): It represents the coverage rate of the camera, ranging from 70% to 95%.

[0095] Data line: Blue: Coverage rate of the east camera; Red: Coverage rate of the south camera; Yellow: Coverage rate of the west camera.

[0096] It can be seen from this simulation diagram that: Coverage rate improvement: The coverage rates of the three cameras gradually increase from the initial 80%, 75%, and 70% to 95%, reaching the target.

[0097] Adjustment efficiency: The east camera takes 3 time steps, the south takes 4 time steps, and the west takes 5 time steps to complete the adjustment. The camera with higher priority (east) adjusts faster.

[0098] System performance: The system can optimize the coverage rates of all cameras to the target within 5 time steps, showing good dynamic adjustment ability and adapting to low-light environments.

[0099] It should be noted that in this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variant thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements not only includes those elements, but also includes other elements not expressly listed, or also includes elements inherent to such process, method, article or device. Without further limitation, an element defined by the statement "including a reference structure" does not exclude the existence of additional identical elements in the process, method, article or device including the element.

[0100] Although the embodiments of the present invention have been shown and described, it will be understood by those of ordinary skill in the art that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the present invention, and the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A device for automatically adjusting coverage in a distributed multi-camera environment, comprising a camera body (1) and a control optimization system, characterized in that: The camera body (1) comprises a multi-source sensor acquisition terminal (2), a multi-source analysis control terminal (3) is installed at the bottom of the multi-source sensor acquisition terminal (2), an intelligent base (4) is installed at the bottom of the multi-source analysis control terminal (3), and adjustment platforms (5) are installed on both sides and the front of the bottom of the intelligent base (4), and a camera (6) is installed at the bottom of the adjustment platform (5); The multi-source sensing acquisition terminal (2) is provided with a data acquisition subsystem, wherein the data acquisition subsystem includes a data acquisition module, a data preprocessing module and a data transmission module; the camera (6) is provided with an auxiliary acquisition subsystem, wherein the auxiliary acquisition subsystem includes an image acquisition module and an edge processing module; the multi-source analysis control terminal (3) is implanted with an environment perception subsystem, a coverage optimization subsystem, a feedback closed-loop subsystem and a user interaction subsystem; the environment perception subsystem includes a feature extraction module, a data fusion module and a feature analysis module; the coverage optimization subsystem includes a task scheduling module, a coverage evaluation module, a strategy optimization module and a verification module; the feedback closed-loop subsystem includes a feedback analysis module and a feedback analysis module; the user interaction subsystem includes a visualization module and an interaction control module; the intelligent base (4) is provided with a dynamic adjustment subsystem, wherein the dynamic adjustment subsystem includes an instruction parsing module, a drive control module and a coordinated scheduling module; the adjustment gimbal (5) is provided with an auxiliary adjustment system, wherein the auxiliary adjustment system includes a microcontroller module and a posture feedback module; The data acquisition subsystem, auxiliary acquisition subsystem, environment perception subsystem, coverage optimization subsystem, feedback closed-loop subsystem, user interaction subsystem, dynamic adjustment subsystem and auxiliary adjustment system together constitute a control optimization system.

2. According to the device for automatically adjusting coverage in a distributed multi-camera environment of claim 1, it is characterized in that: The data acquisition module in the data acquisition subsystem collects the surrounding multi-source environmental data in real time through the light sensor, temperature sensor and distance sensor, the data preprocessing module performs denoising and standardization on the collected data to generate formatted environmental data, and the data transmission module transmits the formatted environmental data to the environmental perception subsystem in the multi-source analysis control terminal (3) through a low-latency communication protocol; The image acquisition module in the auxiliary acquisition subsystem in the camera (6) receives real-time image data from each corresponding camera (6), and the edge processing module performs preliminary target detection and compression on the image data through an embedded chip, generates compressed image data, and transmits it to the environment perception subsystem in the multi-source analysis control terminal (3) through a low-latency communication protocol; The control flow between the data acquisition subsystem, the auxiliary acquisition subsystem and the environmental perception subsystem is designed as follows: the data acquisition subsystem and the auxiliary acquisition subsystem regularly collect data and push it to the environmental perception subsystem. If the environmental perception subsystem detects data missing and abnormalities, it will feedback to the data acquisition subsystem and the auxiliary acquisition subsystem through control instructions, requiring re-collection and adjustment of the collection frequency.

3. According to the device for automatically adjusting coverage in a distributed multi-camera environment of claim 1, it is characterized by: The environment perception subsystem identifies and analyzes the multi-source data through the multi-source analysis control terminal (3), specifically including: The feature extraction module receives the formatted environmental data transmitted by the data acquisition subsystem and the compressed image data transmitted by the auxiliary acquisition subsystem, uses a multimodal feature extraction algorithm to extract key environmental features such as the illumination change rate and the crowd density distribution, generates an environmental feature set, and transmits it to the data fusion module; The data fusion module fuses the environmental feature set and the image features through Bayesian network probabilistic reasoning to generate an environmental feature vector, which is transmitted to the feature analysis module; The feature analysis module analyzes the environmental feature vectors through a time series prediction algorithm, identifies environmental change trends, identifies hotspots using a clustering algorithm, detects emergencies using an anomaly detection algorithm, generates an environmental status report, stores it in a local feature library of the multi-source analysis control terminal (3), and transmits the environmental status report to the coverage optimization subsystem; The control flow between the environment perception subsystem and the coverage optimization subsystem is designed as follows: after the environment perception subsystem generates an environmental status report, it actively pushes it to the coverage optimization subsystem. If the coverage optimization subsystem requires specific environmental features, it requests the environment perception subsystem to re-analyze and supplement feature extraction through control instructions.

4. The device for automatically adjusting coverage in a distributed multi-camera environment according to claim 1, characterized in that: The coverage optimization subsystem is used for target detection and tracking, and specifically includes: The task scheduling module receives the environmental status report of the environmental perception subsystem and the adjustment feedback data of the dynamic adjustment subsystem through the feedback interface, uses the priority weighted scheduling algorithm to generate the monitoring task priority of each camera (6), generates a task allocation table, and transmits it to the strategy optimization module; The coverage assessment module receives the environmental status report and the compressed image data of the auxiliary acquisition subsystem, combines the YOLO algorithm to quickly identify dynamic targets in the monitoring area, predicts and tracks the target position through the Kalman filter algorithm, performs multi-target tracking and data association through the Deep SORT algorithm, builds a dynamic model of coverage efficiency, analyzes the coverage range of the current camera (6) and compares it with the preset optimal coverage range, generates a coverage efficiency heat map, and transmits it to the strategy optimization module; The control flow between the coverage optimization subsystem and the environment perception subsystem is designed as follows: the coverage optimization subsystem triggers the environment perception subsystem to adjust the feature extraction strategy through the target detection result.

5. The device for automatically adjusting coverage in a distributed multi-camera environment according to claim 1, characterized in that: The coverage optimization subsystem generates a regulation strategy through a multi-source analysis control terminal (3), which specifically includes: The strategy optimization module receives the task allocation table and the coverage efficiency heat map, dynamically adjusts the coverage range through a fuzzy logic control algorithm, optimizes the position and angle of the camera (6) through a genetic algorithm, learns the best adjustment strategy through a reinforcement learning algorithm, generates an adjustment strategy, and transmits it to the dynamic adjustment subsystem in the intelligent base (4); The verification module receives the adjustment strategy, verifies it using historical scene data and simulated environment data, generates a verification report, feeds it back to the environment perception subsystem for optimizing feature extraction, and transmits the verification report to the feedback closed-loop subsystem; The control flow between the coverage optimization subsystem and the dynamic adjustment subsystem is designed as follows: after the coverage optimization subsystem generates an adjustment strategy, it is sent to the dynamic adjustment subsystem through the control interface. If the dynamic adjustment subsystem fails to execute and the deviation is too large, the coverage optimization subsystem is notified through the feedback interface to regenerate the strategy.

6. The device for automatically adjusting coverage in a distributed multi-camera environment according to claim 1, characterized in that: The dynamic adjustment subsystem and the auxiliary adjustment system respectively realize dynamic adjustment of the camera (6) through the intelligent base (4) and the adjustment pan / tilt (5), and specifically include: The instruction parsing module in the intelligent base (4) receives the adjustment strategy of the coverage optimization subsystem, parses it into specific control instructions, wherein the control instructions mainly include viewing angle adjustment and position offset parameters, and then transmits it to the drive control module; The driving control module drives the single adjustment pan / tilt (5) to perform viewing angle adjustment through a stepping motor, generates an adjustment execution result, and transmits it to the coordinated scheduling module; The collaborative scheduling module coordinates the actions of multiple groups of adjustment pan-tilt platforms (5) through a time synchronization algorithm, generates a final adjustment execution result, and feeds it back to the coverage optimization subsystem in the multi-source analysis control terminal (3); The microcontroller module in the auxiliary adjustment system in the adjustment pan / tilt (5) executes the instruction of the drive control module to fine-tune the viewing angle, and the attitude feedback module collects the adjusted angle data in real time, generates attitude feedback data, and feeds it back to the coverage optimization subsystem; The control flow between the dynamic adjustment subsystem and the coverage optimization subsystem is designed as follows: after the dynamic adjustment subsystem executes the adjustment strategy, the adjustment execution result and posture feedback data are pushed to the coverage optimization subsystem. If the execution deviation exceeds the threshold, the coverage optimization subsystem requires the dynamic adjustment subsystem to readjust through control instructions.

7. The device for automatically adjusting coverage in a distributed multi-camera environment according to claim 1, characterized in that: The coverage optimization subsystem implements background modeling and scene change perception, specifically including: The coverage assessment module dynamically analyzes the scene background through a background modeling algorithm, eliminates interference factors, generates background model data, and transmits it to the strategy optimization module for improving target detection accuracy; The coverage assessment module analyzes the environmental status report through a time series prediction algorithm, identifies trends in human flow and environmental changes, uses a clustering algorithm to identify hot spots, uses an anomaly detection algorithm to detect emergencies, generates a scene change report, and transmits it to the strategy optimization module for optimizing the adjustment strategy.

8. The device for automatically adjusting coverage in a distributed multi-camera environment according to claim 1, characterized in that: The feedback closed-loop subsystem realizes dynamic optimization, specifically including: The feedback analysis module receives user feedback from the user interaction subsystem, the verification report of the coverage optimization subsystem, and the adjustment execution result of the dynamic adjustment subsystem, and generates optimization instructions including data collection optimization instructions and strategy adjustment instructions after comprehensive analysis, and then sends the data collection optimization instructions to the data collection subsystem in the multi-source sensor collection terminal (2) and sends the strategy adjustment instructions to the coverage optimization subsystem through the control interface; The feedback closed-loop subsystem has self-learning capabilities and continuously optimizes the algorithms and strategies of each module based on historical data and user feedback; The control flow between the feedback closed-loop subsystem and the data acquisition subsystem and coverage optimization subsystem is designed as follows: after the feedback closed-loop subsystem generates the optimization instruction, it actively sends it to the data acquisition subsystem and the coverage optimization subsystem. If the optimization effect does not meet expectations, the data acquisition subsystem or the coverage optimization subsystem requests the feedback closed-loop subsystem to re-analyze through the feedback interface.

9. The device for automatically adjusting coverage in a distributed multi-camera environment according to claim 1, characterized in that: The user interaction subsystem supports multi-dimensional interactions, specifically including: The visualization module receives the coverage efficiency heat map of the coverage optimization subsystem, the environmental status report of the environmental perception subsystem, and the real-time image data of the auxiliary acquisition subsystem, generates a visualization monitoring area map, and transmits it to the interactive control module; The interactive control module supports users to adjust task scheduling priorities and override optimization parameters through an interactive interface, generate user instructions, and feed back to the feedback closed-loop subsystem.

Images