A continuous monitoring system and method for unmanned aerial vehicles (UAVs)
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- SHANDONG ZHIYANG ELECTRIC
- Filing Date
- 2026-01-26
- Publication Date
- 2026-06-02
Smart Images
Figure CN122135603A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of drone video surveillance technology, specifically a drone uninterrupted monitoring system and method based on deep learning and collaborative decision-making. Background Technology
[0002] With the continuous advancement of unmanned aerial vehicle (UAV) platforms, sensing, and communication technologies, UAV systems are increasingly being used in military reconnaissance, public safety monitoring, environmental protection, and disaster emergency response. Existing UAV surveillance systems typically consist of a single UAV or a small number of UAVs, relying on pre-set flight paths or manual remote control to perform patrol missions. They can transmit captured video data back to ground stations in real time for operators to analyze. In recent years, research has attempted to apply simple computer vision algorithms to UAV-transmitted video to achieve motion detection or basic target identification, thereby alleviating the burden of manual monitoring to some extent.
[0003] However, existing drone surveillance technologies still have significant limitations, making it difficult to meet the demands for continuous, intelligent, and highly reliable surveillance in complex scenarios. Specifically, the main problems include the following three aspects: First, the level of intelligence and automation in the monitoring process is low. Existing systems mostly rely on operators to monitor video footage in real time or respond based on preset fixed rules. They are unable to autonomously and accurately identify and continuously track specific targets (such as people, vehicles, and abnormal behaviors) in the monitored scene, resulting in low monitoring efficiency and easy omissions due to human fatigue. Second, the system's continuous monitoring capability is insufficient. Due to the limited endurance of a single drone, monitoring must be interrupted to return for charging or battery replacement when performing long-term missions, resulting in time gaps and spatial blind spots in monitoring coverage, making it impossible to achieve true 24 / 7 uninterrupted monitoring. Third, the system lacks multi-drone collaboration and adaptive scheduling capabilities. Existing multi-drone systems often lack a unified intelligent decision-making center, making it impossible to dynamically allocate tasks and coordinate scheduling based on real-time identification results and the status of each drone (such as battery level, location, and load). As a result, the overall system efficiency and robustness are poor.
[0004] The root causes of these problems can be summarized into the following three points: First, at the data processing level, traditional image processing methods have limited feature extraction capabilities, making it difficult to cope with complex and ever-changing real-world monitoring environments. Furthermore, the lack of efficient and accurate deep learning models prevents the system from achieving a high level of autonomous intelligent perception. Secondly, in terms of energy supply, there is a lack of automated energy supply solutions that are seamlessly integrated with monitoring tasks, and energy management has not been deeply integrated with task scheduling, resulting in the drone's operation cycle being limited by battery capacity. Finally, at the system architecture level, there is a lack of a high-level decision-making mechanism that integrates intelligent perception, collaborative decision-making, and resource management. The various drones and the drones and ground stations have failed to form an organic and collaborative whole, resulting in weak system adaptability and low resource utilization efficiency.
[0005] Therefore, there is an urgent need for a continuous drone surveillance system and method based on deep learning and collaborative decision-making to solve the above problems. Summary of the Invention
[0006] The purpose of this invention is to provide a continuous monitoring system and method for unmanned aerial vehicles (UAVs) based on deep learning and collaborative decision-making. This system improves the accuracy and efficiency of monitoring, expands the monitoring range, enhances the system's adaptability, and enables continuous monitoring.
[0007] To achieve the above objectives, the present invention employs the following technical solution: On one hand, the present invention provides a continuous monitoring system for unmanned aerial vehicles (UAVs), comprising: Several drones, each equipped with image acquisition equipment, sensors and communication equipment, are used to collect monitoring data in a designated monitoring area and transmit it in real time; The ground control station is used to receive and process the monitoring data sent by the UAV and to send control commands to the UAV. A deep learning module, deployed at the ground control station, is used to intelligently process the received monitoring data, extract target features, and perform classification and identification. A collaborative decision-making module, deployed at the ground control station, is used to make collaborative decisions and generate control commands based on the processing results of the deep learning module and the status information of each UAV. The energy replenishment module is used to replenish the drone's energy when it falls below a preset threshold, in order to maintain the system's uninterrupted monitoring capability.
[0008] Preferably, the deep learning module uses a convolutional neural network to perform target detection and recognition on the surveillance video data, and its target recognition function can be expressed as: ; in, To identify probabilities, For activation function, This is the weight matrix. For the input feature vector, For bias terms; The deep learning module supports online learning and model updates.
[0009] Preferably, the collaborative decision-making module constructs a decision model based on a reinforcement learning algorithm, and its decision function is: ; in, For optimal action, For state-action value function, The current system status includes: battery level, location, task load, and target identification information for each drone; The collaborative decision-making module can dynamically schedule drones to perform tasks such as return to home, take-off, tracking, zooming, and taking pictures.
[0010] Preferably, the energy replenishment module includes a ground charging base station or an aerial charging drone, supporting automatic docking and wireless charging functions; when a drone's battery level falls below a threshold... When this happens, the module initiates the resupply process, and the collaborative decision-making module schedules a backup drone to take over the surveillance mission.
[0011] Preferably, the image acquisition device includes a high-definition camera and an infrared sensor, supporting multispectral data acquisition; the sensors include GPS, IMU, and barometer, used to provide UAV position, attitude, and environmental data.
[0012] On the other hand, the present invention also provides a method for continuous monitoring of unmanned aerial vehicles (UAVs), applied to an unmanned aerial vehicle (UAV) continuous monitoring system as described above, comprising the following steps: Step S1: Several drones conduct collaborative surveillance within a designated area and collect monitoring data; Step S2: Transmit the monitoring data to the ground control station in real time; Step S3: Process the data using a deep learning module to extract target features and perform classification and recognition; Step S4: The collaborative decision-making module generates control commands based on the recognition results and the UAV status; Step S5: Send the control command to the corresponding drone to control it to perform actions; Step S6: When the drone's energy is lower than the preset threshold, activate the energy replenishment module to replenish it; Step S7: Repeat steps S1 to S6 to achieve uninterrupted monitoring.
[0013] Preferably, the deep learning processing in step S3 includes: The video stream is analyzed in real time using a pre-trained neural network model to identify specific targets, including people, vehicles, and buildings. Once the target is detected, the gimbal attitude is automatically adjusted to keep the target centered in the field of view.
[0014] Preferably, the collaborative decision-making in step S4 includes: Monitor the battery status of each drone. If the battery level of a drone is lower than the threshold, dispatch a backup drone to the monitored area to take over. Tracking tasks are dynamically assigned based on target recognition results, enabling multi-UAV collaborative tracking and complementary perspectives.
[0015] Preferably, the energy replenishment in step S6 includes: Wireless power transfer via ground charging stations or aerial charging drones; During the resupply process, the collaborative decision-making module ensures that the monitoring task is not interrupted and achieves seamless switching.
[0016] Preferably, it also supports exception handling, including: When a suspicious target or unusual event is detected, an alarm is automatically triggered and video footage is recorded. It supports a manual intervention mode, allowing operators to override system commands and manually control the drone to perform specific tasks.
[0017] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. Significantly improves the intelligence level and monitoring efficiency of the surveillance system: By introducing a deep learning module to analyze and process the surveillance video stream in real time, the system can automatically and accurately identify and classify specific targets (such as people, vehicles, and buildings), and automatically control the pan-tilt unit to track them. This greatly reduces the reliance on manual operation and avoids negligence and omissions caused by human fatigue, thereby improving the accuracy and efficiency of monitoring.
[0018] 2. Achieves truly uninterrupted, blind-spot-free continuous monitoring: By integrating an energy replenishment module and linking it with the collaborative decision-making module, the system can automatically schedule the drone to replenish its energy before it runs out, while simultaneously instructing a backup drone to seamlessly take over the monitoring task. This "relay-style" monitoring mode breaks the limitation of single-drone endurance, fundamentally solving the monitoring interruption problem caused by returning to base for charging, and ensuring the continuity of monitoring tasks and the integrity of coverage.
[0019] 3. Enhanced system collaboration and adaptability, improving overall robustness: The collaborative decision-making module, based on algorithms such as reinforcement learning, acts as the "system brain," intelligently making task allocation and scheduling decisions based on real-time monitoring results (such as the detection of suspicious targets) and the dynamic status of each drone (such as battery level and location). This enables multiple drones to work collaboratively as an organic whole, achieving complex behaviors such as automatic task handover, complementary perspectives, and collaborative tracking. The system can adaptively respond to unexpected situations and task changes, resulting in more efficient resource utilization and a significant improvement in overall reliability and mission success rate. Attached Figure Description
[0020] Figure 1This is a schematic diagram of the system structure of the present invention; Figure 2 This is a flowchart of the method of the present invention. Detailed Implementation
[0021] The present invention will be further illustrated below with reference to specific embodiments. It should be understood that these embodiments are for illustrative purposes only and are not intended to limit the scope of the invention. Furthermore, it should be understood that after reading the teachings of this invention, those skilled in the art can make various alterations or modifications to the invention, and these equivalent forms also fall within the scope defined in this application.
[0022] In this invention, terms such as "upper," "lower," "left," "right," "front," "back," "vertical," "horizontal," "side," and "bottom" indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. These terms are used only to facilitate the description of the structural relationships of the various components or elements of this invention and do not specifically refer to any component or element in this invention. They should not be construed as limiting the invention.
[0023] Example: like Figure 1 As shown, this embodiment provides a continuous monitoring system for unmanned aerial vehicles (UAVs), including: Several drones, each equipped with image acquisition equipment, sensors and communication equipment, are used to collect monitoring data in a designated monitoring area and transmit it in real time; The ground control station is used to receive and process the monitoring data sent by the UAV and to send control commands to the UAV. A deep learning module, deployed at the ground control station, is used to intelligently process the received monitoring data, extract target features, and perform classification and identification. A collaborative decision-making module, deployed at the ground control station, is used to make collaborative decisions and generate control commands based on the processing results of the deep learning module and the status information of each UAV. The energy replenishment module is used to replenish the drone's energy when it falls below a preset threshold, in order to maintain the system's uninterrupted monitoring capability.
[0024] like Figure 2 As shown, this embodiment also provides a method for uninterrupted monitoring by a drone, including the following steps: Step S1: Several drones conduct collaborative surveillance within a designated area and collect monitoring data; Step S2: Transmit the monitoring data to the ground control station in real time; Step S3: Process the data using a deep learning module to extract target features and perform classification and recognition; Step S4: The collaborative decision-making module generates control commands based on the recognition results and the UAV status; Step S5: Send the control command to the corresponding drone to control it to perform actions; Step S6: When the drone's energy is lower than the preset threshold, activate the energy replenishment module to replenish it; Step S7: Repeat steps S1 to S6 to achieve uninterrupted monitoring.
[0025] The following embodiment will explain the system and method in conjunction with a specific application scenario: I. Application Scenarios Overview: This embodiment relates to the practical application of a drone-based uninterrupted surveillance system and method based on deep learning and collaborative decision-making in the field of border security monitoring. Border areas are typically characterized by vast territory, complex terrain, variable climate, and difficulties in manual patrols. Traditional methods of manual patrols and fixed camera monitoring suffer from limited coverage, slow response times, and susceptibility to environmental conditions. Furthermore, illegal border crossings, smuggling, and sabotage frequently occur in border areas, necessitating 24 / 7 uninterrupted monitoring of specific areas to promptly detect and address anomalies.
[0026] This embodiment selects a land border line approximately 50 kilometers long for implementation. This area includes diverse terrain such as plains, hills, and rivers, with varying vegetation cover, large diurnal temperature variations, and occasional sandstorms. The main monitoring objectives include: identifying illegal border crossers and vehicles; monitoring the placement of suspicious items; detecting anomalies in border facilities (such as fence damage); and providing security for regular patrol personnel.
[0027] II. System Composition and Configuration: The uninterrupted drone surveillance system implemented in this embodiment consists of the following parts: 1. Unmanned Aerial Vehicle (UAV) Platform: Equipped with 6 medium-sized rotary-wing drones (model: XM-8Pro), each drone carries: High-definition visible light camera (4K resolution, 30x optical zoom); Infrared thermal imaging camera (640×512 resolution, temperature measurement range -20°C to 150°C); Laser rangefinder (maximum distance 1500m, accuracy ±0.5m); Multispectral sensors (for vegetation analysis and camouflage identification); GPS / BeiDou dual-mode positioning system (positioning accuracy ±0.5m); IMU (Inertial Measurement Unit) (for attitude stabilization); Weather sensors (temperature, humidity, air pressure, wind speed); Data link system (maximum transmission distance 50km, encrypted transmission); The single drone has a flight time of 45 minutes, a maximum flight speed of 60 km / h, and an operating temperature range of -10°C to 50°C.
[0028] 2. Ground control station: Deployed within border outposts, including: Main computer (dual Xeon processors, 512GB RAM, 4×RTX 4090 GPUs). Data storage system (capacity 1PB, read / write speed 2GB / s); Communication base station (maximum coverage radius 60km, multi-band adaptive); Large monitoring screen (4×4 55-inch LCD splicing screen); Backup power system (diesel generator + UPS, providing continuous power for 72 hours).
[0029] 3. Deep Learning Module: An improved YOLOv7 object detection algorithm is used, and the training dataset includes: 20,000 images of people in border areas (different clothing, postures, and lighting conditions); 15,000 vehicle images (various models, colors, angles); 10,000 animal images (common wild animals, reducing false alarms); 8,000 images of unusual situations (fence breaches, object drops, fires, etc.); Model accuracy: Personnel detection mAP@0.5 = 95.2%, Vehicle detection mAP@0.5 = 96.8%; Inference speed: Real-time processing of 1080p video stream (30fps).
[0030] 4. Collaborative Decision-Making Module: Developed based on the Multi-Agent Reinforcement Learning (MADDPG) algorithm; The state space includes: the location of each drone, battery level, sensor status, and target identification information; The action space includes: flight path adjustment, gimbal control, and mission handover decisions; The reward function design takes into account: target tracking quality, coverage, and energy efficiency.
[0031] 5. Energy supply system: Three ground-based automated charging stations (deployed in key locations); Two dedicated charging drones (with large-capacity batteries, capable of aerial docking and charging). Wireless charging system (85% efficiency, charging time to 80% in 15 minutes).
[0032] III. System Workflow 1. Initial Deployment and Task Assignment: After the system starts up, the initial deployment will begin: Three drones are deployed on the south side of the border, two on the north side, and one is kept as a backup. Each drone is responsible for monitoring a section of approximately 8-10 kilometers. The deep learning module loads the latest trained model (updated weekly); The collaborative decision-making module initializes the status and task weights of each UAV; The monitored area is divided into multiple sub-areas, and different monitoring weights are assigned to different areas based on historical data and analysis results. For example: Known areas with frequent illegal border crossings: weight 0.9; Normal passage area: weight 0.6; Areas with difficult-to-pass natural barriers: weight 0.3.
[0033] 2. Data Acquisition and Transmission: The drone flies along a planned path to collect data. Visible light cameras are used for daytime monitoring, while infrared thermal imaging is used for nighttime and inclement weather. Video data is transmitted to the ground control station in real time via data link; Critical data (when a target is detected) is transmitted first, while non-critical data can be cached locally on the drone. The communication protocol uses adaptive coding and modulation to adjust the transmission rate according to channel conditions; Data transmission latency test results: Under normal circumstances: end-to-end latency <200ms; In severe weather: latency <500ms (automatic degradation of transmission quality); When communication is interrupted: Data is stored locally for a maximum of 30 minutes.
[0034] 3. Deep learning processing and target recognition: After the ground control station receives the video data, the deep learning module processes it: Target recognition process: def target_detection(video_stream): Video frame extraction and preprocessing frames = extract_frames(video_stream, fps=30) preprocessed_frames = preprocess(frames) Object detection and classification detections = yolo_model.predict(preprocessed_frames) Target tracking and trajectory analysis For detection in detections: if detection.confidence > 0.7: Confidence threshold target_id = assign_target_id(detection) trajectory = update_trajectory(target_id, detection.position) Behavioral Analysis behavior=analyze_behavior(trajectory, detection.target_type) Threat Assessment threat_level=evaluate_threat(behavior,detection.target_type) if threat_level>THREAT_THRESHOLD: alert_control_station(target_id, threat_level, position) return detections, trajectories, alerts ``` The types of targets that the system can identify include: People (walking, running, crawling, carrying items); Vehicles (cars, motorcycles, all-terrain vehicles); Animals (large wild animals, to avoid false alarms); Abnormal items (packages, weapons, tools); Abnormal environment (smoke, fire, fence damage).
[0035] 4. Collaborative decision-making and task scheduling: The collaborative decision-making module makes decisions based on the identification results and the drone's status: Collaborative decision-making process def collaborative_decision_making(detections, drone_statuses): Task priority calculation tasks = [] For detection in detections: priority = calculate_priority(detection) tasks.append({ 'target_id': detection.id, 'position': detection.position, 'priority': priority, 'required_sensors': get_required_sensors(detection.type) }) Unmanned Aerial Vehicle Capability Assessment drone_capabilities = [] For drone in drone_statuses: capability = evaluate_capability(drone) drone_capabilities.append(capability) Task allocation optimization (Hungarian algorithm) assignment = hungarian_algorithm(tasks, drone_capabilities) Special Circumstances Handling For drone in drone_statuses: if drone.battery <LOW_BATTERY_THRESHOLD: handle_low_battery(drone) if drone.communication_down: handle_communication_loss(drone) return assignment, control_commands; Typical decision-making scenarios include: When a high-priority target is detected, the nearest drone is scheduled to track it; When the drone's battery level drops below 30%, a backup drone is dispatched to take over and instructed to go to the charging station. When multiple targets appear, coordinate multiple drones to form a monitoring network; Adjust the drone's altitude and sensor modes when the weather changes abruptly.
[0036] 5. Energy Management and Replenishment: The power replenishment module monitors the battery level of each drone in real time. Battery level below 35%: Warning, prepare for dispatch and takeover; Battery level below 25%: Initiate return-to-home charging procedure; Battery level below 15%: Emergency return to base (maintaining a safety margin); The charging process is fully automated: The drone autonomously flew to the nearest charging station; Visual guidance enables precise positioning (error <2cm); Automatic wireless charging docking (99.8% success rate); During the charging process, the system automatically dispatches a backup drone to take over the mission.
[0037] IV. Practical Application Examples: Scenario: Nighttime illegal border crossing monitoring: At 02:30 on October 15, 2023, the system detected an anomaly in the S07 section (hilly area) of the border line: 1. Initial detection: The infrared sensor of drone D03 (on patrol) detected three heat sources; Initial classification: People (confidence level 0.82); Location: 42.135°N, 123.768°E, 120 meters from the boundary line.
[0038] 2. Target identification and tracking: The collaborative decision-making module instructs D03 to reduce the altitude to 50 meters and initiate zoom tracking; Although the visible light camera was limited, the infrared image clearly showed the three people crawling forward; Behavioral analysis: Moving covertly, heading towards the boundary line, high threat level.
[0039] 3. Multi-machine collaboration: Dispatch nearby drone D05 to provide support, creating a cross-perspective; D03 continued tracking, while D05 climbed to 200 meters for a wider field of view; The fourth person was identified (previously obscured by the terrain).
[0040] 4. Alarms and Response: The system automatically generates a Level 3 alarm and sends it to the border control command center; At the same time, the nearest patrol team (3.2 kilometers from the scene of the incident) was notified. Record the target's trajectory, speed, and direction, and predict the border crossing point.
[0041] 5. Energy Management: During the process, the battery level of D03 dropped to 28%, and the system automatically dispatched the backup drone D06 to take over. D03 returned to base for charging, with uninterrupted monitoring throughout the entire process.
[0042] 6. Outcome of the action: The patrol team arrived at the predicted border crossing point within 10 minutes. Four people who illegally crossed the border were successfully intercepted; From discovery to completion of the response, the entire process took 23 minutes. The system automatically generates event reports (including timelines, images, and video evidence).
[0043] V. System Performance Evaluation After three months of continuous operation, the system performance indicators are as follows: 1. Monitoring coverage: Time coverage: 99.7% (interrupted only twice throughout the year due to extreme thunderstorms); Spatial coverage: 100% in key areas, 92% along the entire border; 2. Detection performance: Personnel detection accuracy: 96.3% during the day and 91.8% at night; Vehicle detection accuracy: 98.1% during the day, 95.2% at night; False alarm rate: 2.1 times / day on average (mainly caused by wild animals); 3. Response efficiency: From detection to alarm: an average of 8.5 seconds; Target tracking stability: The longest continuous tracking time for a single target is 4 hours and 17 minutes; Task handover time: The average time for drone switching is 22 seconds; 4. Battery life performance: On average, each drone performs 5.2 missions per day. Charging station usage frequency: an average of 18 times per day; Energy replenishment success rate: 99.3%; 5. System reliability: Mean time between failures: 312 hours; Communication interruption recovery time: average 35 seconds; Data completeness rate: 99.96%.
[0044] VI. Advantages: This embodiment fully demonstrates the advantages of the uninterrupted drone surveillance system: 1. Intelligent sensing capability: Deep learning algorithms can adapt to complex border environments, accurately distinguish between people, vehicles, and animals, and multi-sensor fusion can improve all-weather monitoring capabilities and reduce environmental impact factors.
[0045] 2. Advantages of collaborative decision-making: Intelligent scheduling of multiple drones maximizes the monitoring range. Automatic task allocation and handover ensure that critical objectives are not lost; Dynamically adjust monitoring strategies to adapt to different situations and priorities.
[0046] 3. Continuous operation capability: The automated energy replenishment system enables truly uninterrupted monitoring; With good redundancy design, a single point of failure will not affect the overall operation of the system.
[0047] 4. Significant practical results: Compared to traditional monitoring methods, the anomaly detection rate is increased by 3.2 times; Response time reduced by more than 60%; This significantly reduces manpower requirements and lowers the risks associated with duty.
[0048] The above is a detailed description of the preferred embodiments of the present invention, but the present invention is not limited to the embodiments described. Those skilled in the art can make various equivalent modifications or substitutions without departing from the spirit of the present invention, and these equivalent modifications or substitutions are all included within the scope defined by the claims of this application.
Claims
1. A continuous monitoring system for unmanned aerial vehicles (UAVs), characterized in that, include: Several drones, each equipped with image acquisition equipment, sensors and communication equipment, are used to collect monitoring data in a designated monitoring area and transmit it in real time; The ground control station is used to receive and process the monitoring data sent by the UAV and to send control commands to the UAV. A deep learning module, deployed at the ground control station, is used to intelligently process the received monitoring data, extract target features, and perform classification and identification. A collaborative decision-making module, deployed at the ground control station, is used to make collaborative decisions and generate control commands based on the processing results of the deep learning module and the status information of each UAV. The energy replenishment module is used to replenish the drone's energy when it falls below a preset threshold, in order to maintain the system's uninterrupted monitoring capability.
2. The uninterrupted monitoring system for unmanned aerial vehicles according to claim 1, characterized in that, The deep learning module uses a convolutional neural network to perform target detection and recognition on the surveillance video data. Its target recognition function can be expressed as: ； in, To identify probabilities, For activation function, This is the weight matrix. For the input feature vector, For bias terms; The deep learning module supports online learning and model updates.
3. The uninterrupted monitoring system for unmanned aerial vehicles according to claim 1, characterized in that, The collaborative decision-making module constructs a decision model based on a reinforcement learning algorithm, and its decision function is: ； in, For optimal action, For state-action value function, The current system status includes: battery level, location, task load, and target identification information for each drone; The collaborative decision-making module can dynamically schedule drones to perform tasks such as return to home, take-off, tracking, zooming, and taking pictures.
4. The uninterrupted monitoring system for unmanned aerial vehicles according to claim 1, characterized in that, The energy replenishment module includes a ground charging base station or an aerial charging drone, supporting automatic docking and wireless charging functions; when a drone's battery level falls below a threshold... When this happens, the module initiates the resupply process, and the collaborative decision-making module schedules a backup drone to take over the surveillance mission.
5. The uninterrupted monitoring system for unmanned aerial vehicles according to claim 1, characterized in that, The image acquisition device includes a high-definition camera and an infrared sensor, supporting multispectral data acquisition; the sensors include GPS, IMU, and barometer, used to provide UAV position, attitude, and environmental data.
6. A method for continuous monitoring of unmanned aerial vehicles (UAVs), applied to a continuous UAV monitoring system as described in any one of claims 1-5, characterized in that, Includes the following steps: Step S1: Several drones conduct collaborative surveillance within a designated area and collect monitoring data; Step S2: Transmit the monitoring data to the ground control station in real time; Step S3: Process the data using a deep learning module to extract target features and perform classification and recognition; Step S4: The collaborative decision-making module generates control commands based on the recognition results and the UAV status; Step S5: Send the control command to the corresponding drone to control it to perform actions; Step S6: When the drone's energy is lower than the preset threshold, activate the energy replenishment module to replenish it; Step S7: Repeat steps S1 to S6 to achieve uninterrupted monitoring.
7. The uninterrupted monitoring method for unmanned aerial vehicles according to claim 6, characterized in that, The deep learning processing described in step S3 includes: The video stream is analyzed in real time using a pre-trained neural network model to identify specific targets, including people, vehicles, and buildings. Once the target is detected, the gimbal attitude is automatically adjusted to keep the target centered in the field of view.
8. The uninterrupted monitoring method for unmanned aerial vehicles according to claim 6, characterized in that, The collaborative decision-making process described in step S4 includes: Monitor the battery status of each drone. If the battery level of a drone is lower than the threshold, dispatch a backup drone to the monitored area to take over. Tracking tasks are dynamically assigned based on target recognition results, enabling multi-UAV collaborative tracking and complementary perspectives.
9. A method for continuous monitoring of unmanned aerial vehicles according to claim 6, characterized in that, The energy replenishment mentioned in step S6 includes: Wireless power transfer via ground charging stations or aerial charging drones; During the resupply process, the collaborative decision-making module ensures that the monitoring task is not interrupted and achieves seamless switching.
10. A method for continuous monitoring of unmanned aerial vehicles according to claim 6, characterized in that, It also supports exception handling, including: When a suspicious target or unusual event is detected, an alarm is automatically triggered and video footage is recorded. It supports a manual intervention mode, allowing operators to override system commands and manually control the drone to perform specific tasks.