Intelligent security patrol method and system for multiple unmanned aerial vehicles at night

Abstract

This invention relates to the field of UAV collaborative technology, providing a method and system for intelligent security patrol inspection using multiple UAVs at night. The method divides the inspection area into a continuous three-dimensional virtual grid, assigns a deployment position to each UAV, and monitors the surrounding grid, sharing grid status information through inter-UAV communication. When a moving target is detected, the grid it occupies is marked as a triggered state grid, and only surrounding UAVs are dispatched to observe it closely. Multiple observation UAVs share position data, calculate the target's speed and direction in real time, predict the grid sequence it may pass through in the future, and mark these as pre-triggered state grids. Warning commands are sent to UAVs surrounding the pre-triggered grids to adjust their positions and viewing angles in advance. All detection data is fused in real time, the target motion model is corrected, and the triggered and pre-triggered state grids are dynamically updated. This invention achieves proactive prediction and relay tracking of moving targets at night, improving inspection efficiency and target acquisition capabilities.

Description

Technical Field

[0001] This invention belongs to the field of drone collaborative technology, and in particular relates to a method and system for intelligent security patrol and inspection by multiple drones at night. Background Technology

[0002] With the rapid development of drone technology and the increasing demand for security, drone inspections have been widely applied in border patrols, park security, and facility monitoring. Nighttime environments, due to poor lighting and limited visibility, have always been a challenge and a key focus of security inspections. Traditional nighttime security mainly relies on fixed monitoring equipment or single drone patrols, which are insufficient to address the inspection needs of large-scale, complex terrain. In recent years, multi-drone collaborative technology has gradually matured. Through multi-drone formation flights and mission coordination, the inspection range can be effectively expanded and coverage density increased, providing a new technological approach for nighttime security inspections.

[0003] Existing technical solutions suffer from the following problems in practical applications: In low-light conditions at night, it is difficult for a single drone to maintain stable tracking after detecting a target, and the target easily escapes the monitoring field of view; multiple drones execute tasks independently, lacking information sharing and collaborative decision-making, resulting in delayed tracking responses to moving targets and an inability to form effective relay tracking; drone swarms often enter high-power tracking mode after detecting a target, wasting energy and communication resources; they lack the ability to predict the trajectory of moving targets, only able to respond passively after the target appears, making it difficult to deploy observation resources in advance before the target enters a critical area; and the detection data acquired by multiple drones is processed in a scattered manner, unable to be fused in real time to form a unified target motion model, leading to deviations in target positioning and tracking. Summary of the Invention

[0004] The purpose of this invention is to provide a method and system for multi-drone collaborative intelligent security patrol inspection at night, which aims to solve the technical problems existing in the prior art as identified in the background art.

[0005] This invention is implemented as follows: a multi-drone collaborative intelligent security patrol method for nighttime patrols, the method comprising:

[0006] The inspection area is divided into a continuous three-dimensional virtual grid in three-dimensional space. Each drone is assigned a deployment position. Each drone monitors the three-dimensional virtual grid around its corresponding deployment position and shares the status information of the entire three-dimensional virtual grid through inter-drone communication.

[0007] When any UAV detects a moving target within the 3D virtual grid, the 3D virtual grid is marked as a trigger state grid, and the trigger state grid's number and spatial location are determined.

[0008] UAVs deployed within the three-layer 3D virtual grid centered on the trigger state grid are selected as close-range observation UAVs. The close-range observation UAVs are scheduled to observe and their flight attitude and sensor pointing are adjusted to continuously observe moving targets.

[0009] Acquire all moving target location data collected by close-range observation drones, calculate the instantaneous speed, direction vector and movement trend of the moving target in real time, and dynamically deduce the predicted path grid sequence that the moving target may pass through in the future within a preset time period in the shared three-dimensional virtual grid map, and mark it as a pre-trigger state grid.

[0010] Send early warning commands to drones around the pre-triggered state grid so that the drones around the pre-triggered state grid can adjust their sensor views in advance and align them with the predicted path grid.

[0011] The detection data from all UAVs are fused in real time to correct the position and motion model of moving targets, and the triggered state grid and pre-triggered state grid are updated in real time.

[0012] As a further aspect of the present invention, the allocation of deployment locations for each drone specifically includes:

[0013] The inspection area is divided into continuous three-dimensional virtual grids of equal size in three-dimensional space;

[0014] Each drone is assigned a deployment location, ensuring that the deployment locations of each drone are evenly distributed across the inspection area. Each drone monitors eight adjacent three-dimensional virtual grids in three-dimensional space, centered on its own deployment location. This ensures that the three-dimensional virtual grids surrounding the corresponding deployment location monitored by each drone cover the entire inspection area, and that the monitoring ranges of adjacent drones overlap.

[0015] Each drone broadcasts its detected status information to other drones via an inter-drone communication link and receives status information of the 3D virtual mesh broadcast by other drones.

[0016] As a further embodiment of the present invention, the marker is a trigger state grid, specifically including:

[0017] Based on the position coordinates of any UAV when it detects a moving target within a 3D virtual grid, generate a trigger state grid marker;

[0018] In the shared 3D virtual mesh status information, the status field of the 3D virtual mesh that detects a moving target is updated to the trigger status, and the trigger time is recorded;

[0019] Extract the unique number and spatial coordinate range of the 3D virtual mesh corresponding to the trigger state mesh, and store the number and spatial coordinate range in association with the trigger time.

[0020] As a further aspect of the present invention, the scheduling of close-range observation drones for observation specifically includes:

[0021] Based on the trigger state grid number and spatial coordinate range, calculate the spatial area covered by the three-layer three-dimensional virtual grid extending outward from the trigger state grid, and designate all UAVs deployed within this spatial area as close-range observation UAVs;

[0022] Send close-range observation instructions to the selected close-range observation drones. The close-range observation instructions include the trigger state grid number, spatial coordinate range, and the current position of the moving target.

[0023] Control each close-in observation UAV to adjust its flight course toward the trigger state grid according to the close-in observation command, and adjust the center of the sensor field of view to align with the moving target;

[0024] Control all close-range observation drones to photograph and track moving targets from different angles, and share real-time moving target location data through inter-drone communication links.

[0025] As a further aspect of the present invention, the step of dynamically deducing the predicted path grid sequence that a moving target may traverse within a preset time period in a shared three-dimensional virtual grid map specifically includes:

[0026] The location coordinates of the moving target in the three-dimensional virtual grid, measured in real time by each close-range observation UAV, are collected through the inter-machine communication link to form a location data set.

[0027] The spatial coordinate system is transformed into the position data set. Using the position data set at the same moment, the instantaneous velocity vector, instantaneous velocity magnitude and direction of motion of the moving target in three-dimensional space are calculated by geometric solution method. The motion trend curve of the moving target is fitted according to the velocity vector changes at consecutive moments.

[0028] Starting from the current three-dimensional virtual grid where the moving target is located, along the direction of the motion trend curve, and combined with the instantaneous speed of the moving target, predict the three-dimensional virtual grids that the moving target will pass through in a future preset time period, and generate a predicted path grid sequence according to the predicted time sequence.

[0029] In the shared 3D virtual mesh state information, a pre-triggering state identifier is assigned to each 3D virtual mesh in the predicted path mesh sequence. The pre-triggering state identifier includes the prediction time window in which a moving target is predicted to enter the 3D virtual mesh, as well as the predicted motion direction and velocity information of the moving target.

[0030] As a further aspect of the present invention, sending early warning commands to drones surrounding the pre-triggered state grid specifically includes:

[0031] Based on the number of each pre-triggering state grid and the prediction time window in the predicted path grid sequence, drones whose deployment locations can cover the pre-triggering state grid are selected, and warning instructions are sent to the selected drones. The warning instructions include the pre-triggering state grid number, the prediction time window, the predicted direction of movement, and the predicted instantaneous speed.

[0032] The drone that receives the early warning command adjusts its flight path according to the predicted time window and its current position to approach the periphery of the pre-triggered state grid in advance, and adjusts the pointing angle of the sensor so that the sensor's field of view covers the entrance direction of the pre-triggered state grid in advance, so that the moving target can be captured by the sensor when it enters the pre-triggered state grid.

[0033] As a further aspect of the present invention, the real-time updating of the triggered state grid and the pre-triggered state grid specifically includes:

[0034] Establish a data fusion node, and aggregate the mobile target detection data acquired by all close-in observation UAVs and UAVs that have received early warning commands to the data fusion node through the inter-machine communication link. Perform spatiotemporal alignment and data association processing on the multi-source detection data to generate a fused mobile target state estimate.

[0035] The position coordinates of the moving target are updated using the fused moving target state estimate. The updated position coordinates are then input into the motion trend curve fitting process to recalculate the instantaneous velocity vector, instantaneous velocity magnitude, and motion direction of the moving target, so that the motion model reflects the latest motion state of the moving target in real time.

[0036] Based on the updated position coordinates of the moving target, determine whether the moving target has entered a new 3D virtual grid. If it has entered a new 3D virtual grid, mark the newly entered 3D virtual grid as a trigger state grid. At the same time, re-determine the predicted path grid sequence based on the updated motion trend curve.

[0037] The pre-triggering status mark is removed from the original predicted path grid sequence for grids that have become invalid because the prediction time window has expired and the moving target has not actually entered. The pre-triggering status mark is removed from the grids in the original predicted path grid sequence that are not included in the re-derived predicted path grid sequence. The newly derived predicted path grids are marked as pre-triggering state grids, thereby realizing the dynamic update of the triggering state grids and the pre-triggering state grids.

[0038] Another object of the present invention is to provide a multi-drone collaborative intelligent security patrol system for nighttime unmanned aerial vehicles (UAVs), the system comprising:

[0039] The grid division module is used to divide the inspection area into continuous three-dimensional virtual grids in three-dimensional space, assign deployment positions to each UAV, and each UAV monitors the three-dimensional virtual grids around its corresponding deployment position and shares the status information of the entire three-dimensional virtual grid through inter-UAV communication.

[0040] The target detection module is used to mark the three-dimensional virtual grid as a trigger state grid when any UAV detects a moving target within the three-dimensional virtual grid, and to locate the number and spatial position of the trigger state grid.

[0041] The UAV screening and scheduling module is used to screen UAVs whose deployment location is within the three-layer three-dimensional virtual grid centered on the trigger state grid as close-range observation UAVs, schedule close-range observation UAVs to view them, adjust the flight attitude and sensor pointing of close-range observation UAVs, and continuously observe moving targets.

[0042] The target motion data analysis module is used to acquire all the moving target position data collected by the close-range observation UAV, calculate the instantaneous motion speed, direction vector and motion trend of the moving target in real time, and dynamically deduce the predicted path grid sequence that the moving target may pass through in the future within a preset time period in the shared three-dimensional virtual grid map, and mark it as a pre-trigger state grid.

[0043] The early warning command sending module is used to send early warning commands to drones around the pre-triggered state grid, so that the drones around the pre-triggered state grid can adjust their sensor perspective in advance and align with the predicted path grid.

[0044] The data fusion and update module is used to fuse all UAV detection data in real time, correct the position and motion model of moving targets, and update the triggered state grid and pre-triggered state grid in real time.

[0045] The beneficial effects of this invention are:

[0046] This invention constructs a fully covered, real-time-aware digital inspection space by dividing the inspection area into a continuous three-dimensional virtual grid, assigning a monitoring range to each drone and sharing grid status information.

[0047] When any UAV detects a moving target, it only triggers close-range observation of UAVs within a limited range around the target's grid, preventing the entire fleet from interrupting its routine tasks due to a single event, thus achieving precise resource deployment and effective energy consumption control.

[0048] Multiple drones in the vicinity collaboratively observe from different angles, using shared position data to calculate the target's instantaneous speed and direction in real time, and fit a motion trend curve. Based on this, the system dynamically extrapolates a grid sequence of predicted paths the target might traverse in the future, enabling it to predict the trajectory of moving targets. By sending early warning commands to drones around the predicted grid, they can adjust their positions and sensor perspectives in advance, completing observation deployment before the target arrives. This achieves a shift from passive response to proactive prediction, ensuring that a drone is on standby to observe the target when it enters each predicted grid, forming a continuous and seamless relay tracking network.

[0049] All participating drones will fuse the detection data in real time, continuously correct the target's position and motion model, and use the deviation between the actual trigger grid and the predicted grid as feedback to optimize the subsequent prediction accuracy, forming a closed-loop feedback mechanism that continuously improves the accuracy and stability of target tracking as observation continues.

[0050] This invention addresses the unique characteristics of nighttime environments by comprehensively considering factors such as low illumination and differences in thermal radiation, adjusting sensor configuration and data fusion strategies to ensure the reliability of grid triggering and target detection under various nighttime conditions. The entire method, through spatial grid management, multi-drone collaborative sensing, motion trajectory prediction, and closed-loop feedback optimization, significantly improves the efficiency and target acquisition capabilities of nighttime multi-drone collaborative inspections, saves UAV energy and communication resources, and makes security inspections more intelligent, efficient, and adaptable to complex nighttime scenarios. Attached Figure Description

[0051] Figure 1 A flowchart of the nighttime multi-drone collaborative intelligent security patrol method provided in this embodiment of the invention;

[0052] Figure 2 A flowchart for assigning deployment locations to each drone, provided as an embodiment of the present invention;

[0053] Figure 3 A flowchart labeled as a trigger state grid is provided for an embodiment of the present invention;

[0054] Figure 4 This is a flowchart of a scheduling and close-range observation UAV viewing process provided in an embodiment of the present invention;

[0055] Figure 5 A flowchart illustrating how to dynamically deduce a predicted path grid sequence that a moving target may traverse within a preset time period in a shared three-dimensional virtual grid map, as provided in an embodiment of the present invention.

[0056] Figure 6 This is a flowchart illustrating the process of sending early warning commands to drones surrounding a pre-triggered state grid, as provided in an embodiment of the present invention.

[0057] Figure 7 This is a flowchart of real-time updates of the triggered state grid and the pre-triggered state grid provided in an embodiment of the present invention;

[0058] Figure 8 The structural block diagram of the multi-drone collaborative intelligent security patrol system for nighttime use provided in this embodiment of the invention. Detailed Implementation

[0059] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0060] Figure 1 The flowchart of the multi-drone collaborative intelligent security patrol method for nighttime use provided in this embodiment of the invention is as follows: Figure 1 As shown, the method includes:

[0061] The S100 divides the inspection area into a continuous three-dimensional virtual grid in three-dimensional space, assigns a deployment position to each drone, and each drone monitors the three-dimensional virtual grid around its corresponding deployment position and shares the status information of the entire three-dimensional virtual grid through inter-drone communication.

[0062] The entire inspection area is broken down into continuous three-dimensional virtual grids of equal size, completing the standardization and precise quantification of the inspection space. This transforms the originally vague inspection area into grid units that can be accurately identified and controlled by drones. Based on the quantified three-dimensional virtual grid layout, deployment positions are assigned to each drone. The deployment process follows the principle of uniform distribution, ensuring that the spatial distribution of drones in the inspection area matches the grid layout. Each drone monitors eight adjacent three-dimensional virtual grids in three-dimensional space, with its own deployment position as the center. In the planning of deployment and monitoring range, it is ensured that the monitoring range of all drones can completely cover the entire inspection area, while allowing the monitoring ranges of adjacent drones to form natural overlapping areas.

[0063] A stable inter-drone communication link is established, and each drone broadcasts the grid status information it detects to all other drones in the cluster in real time through this link. At the same time, each drone can also continuously receive grid status information broadcast by other drones, realizing the full-domain sharing of 3D virtual grid status information in the entire inspection area.

[0064] In practical nighttime security patrol applications, the three-dimensional virtual grid can adjust the grid size according to the actual space size and security needs of the patrol area. Drones are evenly distributed in the three-dimensional space of the patrol area according to the grid layout. The space at low and even medium altitudes can be broken down into standardized grids and included in the monitoring range. Each drone focuses on basic monitoring of the surrounding eight grids. The overlapping monitoring range of adjacent drones can form multiple monitoring of the grid boundary area. The inter-drone communication link allows all drones to synchronize the grid status of the entire patrol area in real time. The idle or abnormal status of various grids can be monitored in real time by all drones in the cluster.

[0065] S200: When any UAV detects a moving target within the three-dimensional virtual grid, it marks the three-dimensional virtual grid as a trigger state grid and locates the trigger state grid's number and spatial position.

[0066] Once any drone detects a moving target within its monitored 3D virtual grid, the process of identifying and solidifying abnormal grid information is immediately initiated. Based on the pre-set 3D virtual grid spatial coding and coordinate system of the inspection area, the real-time position coordinates of the drone when it detects the moving target are accurately matched with the corresponding 3D virtual grid, generating a trigger status grid marker that is compatible with the global shared grid system. This marker is directly associated with the spatial attributes and abnormal status of the grid.

[0067] In the shared 3D virtual mesh status information of the UAV cluster, the status field corresponding to the mesh that detects a moving target is updated in real time, completing the switch from the normal monitoring state to the triggered state. At the same time, the instant when the mesh is triggered is accurately recorded, so that the abnormal information has a unified time reference. Then, the unique number of the triggered state mesh is extracted from the dedicated coding library of the 3D virtual mesh, and the complete 3D spatial coordinate range it covers is determined. The mesh number, spatial coordinate range and trigger time are integrated and stored to form a complete abnormal mesh information package containing spatiotemporal attributes and location attributes. This information package is synchronized to the entire UAV cluster in real time through the inter-machine communication link, so that all UAVs can accurately obtain the specific mesh location where the abnormality occurred.

[0068] In practical applications of nighttime security patrols, in various areas requiring full-area 3D security, after drones detect suspicious moving targets such as people and vehicles in low-light and limited-view environments at night, they can quickly complete the standardized marking and information synchronization of abnormal grids through this process. This allows the dispersed drone clusters to quickly form a unified abnormal perception and cognition. The transmission and identification of various abnormal information are all completed based on the preset grid system, without the need for additional manual intervention and information conversion.

[0069] S300 selects drones deployed within the three-layer 3D virtual grid centered on the trigger state grid as close-range observation drones, schedules close-range observation drones to observe them, adjusts the flight attitude and sensor pointing of the close-range observation drones, and continuously observes moving targets.

[0070] Based on the three-dimensional virtual grid space system preset in the inspection area, the drone screening and collaborative scheduling operation is carried out. First, the marked trigger state grid is used as the core of the three-dimensional space. The complete spatial area covered by the three-layer three-dimensional virtual grid is accurately calculated and used as the spatial boundary for drone screening. All drones deployed in this spatial area are accurately identified from the entire drone inspection cluster and designated as close-range observation drones, thus clarifying the main body of nighttime collaborative observation.

[0071] A unified close-range observation command is issued to all selected close-range observation drones. The command integrates core information such as the trigger state grid number, spatial coordinate range, and the current real-time position of the moving target, providing a unified and accurate basis for the close-range operation of each drone. Then, the flight attitude and sensors of each close-range observation drone are precisely controlled to guide the drone to adjust its flight course and fly smoothly in the spatial direction of the trigger state grid. At the same time, the center of the sensor's field of view is adjusted to accurately align it with the real-time position of the moving target.

[0072] After completing the observation preparation, all close-range observation drones are coordinated to form a three-dimensional multi-view collaborative observation system, allowing each drone to conduct synchronous shooting and continuous tracking of the moving target from different spatial angles. Furthermore, all close-range observation drones will share the real-time moving target position data with the entire drone cluster through a stable inter-drone communication link, enabling all drones participating in the collaborative inspection within the cluster to obtain complete target observation data in real time.

[0073] In various areas requiring 3D full-domain security patrols, once the triggered state grid is marked, this process can quickly select suitable drones from the surrounding area to form a close-range observation cluster. Different drones will observe moving targets from different directions at high and low altitudes. Even at night, when there are factors affecting detection such as insufficient light, building obstructions, and terrain occlusion, the multi-directional observation angles can still achieve all-round capture of the target. Furthermore, the real-time synchronization of all observation data allows the cluster to form a unified target perception system.

[0074] S400 acquires all the moving target position data collected by the close-in observation UAV, calculates the instantaneous speed, direction vector and movement trend of the moving target in real time, and dynamically infers the predicted path grid sequence that the moving target may pass through in the future within a preset time period in the shared three-dimensional virtual grid map, and marks it as a pre-trigger state grid.

[0075] Relying on the inter-machine communication link, the system completes the full-domain collection of mobile target position data collected by all close-range observation UAVs, integrates the discrete position data obtained by UAVs from different spatial angles and detection positions into a complete position data group, and allows the scattered single-node detection data to form a complementary organic data whole.

[0076] Subsequently, a spatial coordinate system transformation operation was performed on the location data group to eliminate spatial data deviations caused by the local coordinate systems of different UAVs, so that all location data could be uniformly incorporated into a standardized geodetic coordinate system, forming effective analytical data with a unified spatial reference standard.

[0077] Based on this standardized location data set, geometric calculation methods are used to process multi-source data at the same time to accurately obtain the instantaneous velocity and direction vector of the moving target in three-dimensional space. Simultaneously, by combining the velocity vector changes over consecutive time intervals, a motion trend curve that accurately reflects the motion pattern of the moving target is generated through curve fitting, transforming discrete location data into a motion state model with continuity and regularity. Then, using the current three-dimensional virtual grid of the moving target as the spatial starting point, along the direction of the fitted motion trend curve and combining the calculated instantaneous velocity magnitude, the spatial position changes of the moving target within a preset time period are accurately predicted. This clearly identifies all the three-dimensional virtual grids the target will sequentially pass through during this time period, and these grids are arranged in an orderly manner according to the expected chronological order of the target's passage, generating a standardized predicted path grid sequence.

[0078] In the shared 3D virtual grid status information of the UAV cluster, each 3D virtual grid in the sequence is assigned a unique pre-trigger status identifier. The identifier integrates the predicted time window when a moving target is predicted to enter the grid, as well as the predicted motion direction and velocity information when the target enters the grid. This prediction information is deeply bound to the 3D virtual grid system, and the pre-trigger status information is synchronized to the entire UAV inspection cluster through the inter-UAV communication link, so that all UAVs participating in the collaborative inspection can accurately obtain the future motion prediction information of the moving target.

[0079] In various areas requiring 3D full-domain security patrols, once the drone swarm completes multi-angle close-up observations of moving targets, this process can accurately calculate the target's motion status and predict its path. Even at night, with environmental factors affecting target tracking such as insufficient light, complex terrain, and building obstructions, the integration of multi-source data and accurate calculation fitting can maintain high accuracy in judging the target's motion trend. Path extrapolation based on the grid system allows the prediction results to be directly connected to the subsequent drone early warning and scheduling process, transforming nighttime patrols from passive anomaly detection and tracking into proactive motion prediction and advance deployment.

[0080] The S500 sends warning commands to drones around the pre-triggered state grid, causing the drones around the pre-triggered state grid to adjust their sensor perspectives in advance and align them with the predicted path grid.

[0081] Based on the standardized attributes of the 3D virtual mesh and the deployment information of the UAV swarm, precise UAV screening and early warning command issuance are carried out. First, based on the unique number of each pre-triggered state grid in the predicted path grid sequence and the corresponding prediction time window, combined with the deployment location of the UAV and the effective detection range of the sensor, UAVs whose deployment location can form effective detection coverage for the pre-triggered state grid are accurately screened from the entire inspection swarm. The UAVs that are responsible for issuing early warnings for each pre-triggered grid are determined. Then, standardized early warning commands are sent to these selected UAVs. The commands integrate core information such as the pre-triggered state grid number, prediction time window, predicted target movement direction, and predicted instantaneous movement speed, so that each UAV that issues the early warning can accurately grasp the predicted movement parameters of the target.

[0082] The system performs precise dual control over the flight path and sensor attitude of the drone that receives the early warning command. The drone autonomously plans and adjusts its flight path based on its current spatial position and the prediction time window of the pre-triggered grid, so as to smoothly approach the periphery of the pre-triggered state grid in advance. At the same time, it adjusts the pointing angle of the sensor so that the effective field of view of the sensor can accurately and in advance cover the entrance direction of the pre-triggered state grid. This ensures that the detection range of the sensor is precisely matched with the path of the target expected to enter the grid, so that the moving target can be accurately captured by the sensor the moment it enters the pre-triggered state grid. The whole process relies on the inter-drone communication link to realize the real-time issuance of commands and the real-time feedback of the drone status, so that the early warning deployment and the predicted movement rhythm of the target are highly consistent.

[0083] In practical applications of nighttime security patrols, once the path prediction of moving targets is completed and the pre-trigger grid is marked, the screening and deployment of early warning drones can be completed quickly through this process. Whether it is a low-altitude ground moving target or a low-altitude flying target in the middle and low altitudes, the drones around the corresponding pre-trigger grid can complete the position adjustment and sensor calibration in advance according to the early warning instructions. Even in environments with insufficient light at night, complex terrain, or building obstructions, the sensors can always be aligned with the direction the target is expected to enter through advance deployment, achieving full-process predictive monitoring of the target's movement path.

[0084] The S600 integrates all UAV detection data in real time, corrects the position and motion model of moving targets, and updates the triggered state grid and pre-triggered state grid in real time.

[0085] Within the collaborative system of the UAV swarm, a dedicated data fusion node is established. This node serves as the core processing hub for multi-source detection data. Through a stable inter-machine communication link, it comprehensively aggregates the mobile target detection data collected by all UAVs participating in close-range observation and UAVs that receive early warning commands for pre-deployment. This data includes information such as target position and motion status from different perspectives, locations, and times. The data fusion node performs professional spatiotemporal alignment and data correlation processing on the aggregated multi-source detection data to eliminate data biases caused by different UAVs in terms of detection time, spatial location, and sensor characteristics, generating more accurate fused mobile target state estimates.

[0086] Based on this state estimate, the real-time position coordinates of the moving target are updated. At the same time, the updated position coordinates are re-inputted into the fitting process of the moving target's motion trend curve. The instantaneous velocity vector, instantaneous velocity magnitude, and motion direction of the target are recalculated accurately, so that the target's motion model can fit its actual motion state in real time and eliminate the deviation of the previous motion model caused by changes in the target's motion state.

[0087] Subsequently, based on the updated precise location coordinates of the moving target and the spatial division rules of the three-dimensional virtual grid in the inspection area, it is determined whether the moving target has entered a new three-dimensional virtual grid. If it is determined that it has entered a new grid, the new grid is immediately marked as a trigger state grid. At the same time, based on the refitted motion trend curve, the predicted path grid sequence that the moving target may pass through in the future within a preset time period is dynamically deduced again, so that the path prediction is always based on the latest motion state of the target.

[0088] The pre-triggered state grids in the shared 3D virtual mesh system are dynamically cleaned and updated. The pre-triggered state markers of grids in the original predicted path grid sequence whose predicted time windows have expired and whose moving targets have not actually entered are removed. At the same time, the pre-triggered state markers of grids in the original sequence that were not included in the re-deduced predicted path grid sequence are also removed. All grids in the re-deduced predicted path grid sequence are uniformly marked as pre-triggered state grids. This completes the full-domain dynamic update of the triggered state grids and pre-triggered state grids. Moreover, the update information of all grid states is synchronized to the entire UAV inspection cluster in real time through the inter-machine communication link, allowing each UAV to carry out subsequent collaborative inspection operations based on the latest grid state.

[0089] When a moving target undergoes changes in motion such as changing direction, speed, or even brief pauses during nighttime patrols, this process can quickly complete data fusion correction and dynamic updates of the grid status. Even if there are problems such as sensor detection noise or local data loss due to terrain obstruction at night, the fusion processing of multi-source data can make up for the deficiencies of single data, ensuring that the grid status updates always match the actual movement trajectory of the target, and enabling the collaborative deployment of the drone swarm to adjust in real time to follow the changes in the target's movement.

[0090] like Figure 2 As shown, assigning deployment locations to each drone specifically includes:

[0091] S110 divides the inspection area into a continuous three-dimensional virtual grid of equal size in three-dimensional space;

[0092] S120 assigns a deployment position to each drone, ensuring that the deployment positions of each drone are evenly distributed in the inspection area. Each drone monitors eight adjacent three-dimensional virtual grids in three-dimensional space with its own deployment position as the center, ensuring that the three-dimensional virtual grids around the corresponding deployment position monitored by each drone cover the entire inspection area and that the monitoring ranges of adjacent drones overlap.

[0093] Evenly distributed deployment can avoid monitoring blind spots, balance the load of each drone, ensure full coverage of the inspection area, and prevent the decrease in accuracy caused by overloading of a single drone.

[0094] Monitoring eight adjacent grids can match the effective detection radius of the drone's sensors. Using eight grids as the basic unit simplifies management, maximizes the use of detection capabilities, and covers the core area around the deployment location.

[0095] The S130 broadcasts the status information detected by each UAV to other UAVs via the inter-UAV communication link, and receives the status information of the three-dimensional virtual mesh broadcast by other UAVs.

[0096] like Figure 3 As shown, the marker is a trigger state grid, specifically including:

[0097] S210: Generate a trigger state grid marker based on the position coordinates of any UAV when it detects a moving target in the three-dimensional virtual grid;

[0098] S220 updates the state field of the 3D virtual mesh that detects a moving target to the trigger state in the shared 3D virtual mesh state information and records the trigger time; recording the trigger time can provide a unified time starting point for multi-machine collaboration and avoid time asynchrony between scheduling / prediction.

[0099] S230, extract the unique number and spatial coordinate range of the three-dimensional virtual mesh corresponding to the trigger state mesh, and store the number and spatial coordinate range in association with the trigger time.

[0100] like Figure 4 As shown, the dispatching of close-range observation drones for observation specifically includes:

[0101] S310, based on the trigger state grid number and spatial coordinate range, calculate the spatial area covered by the three-layer three-dimensional virtual grid extending outward from the trigger state grid, and designate all UAVs whose deployment locations are within this spatial area as close-range observation UAVs;

[0102] The three-layer design aims to balance observation accuracy with system overhead (too few layers are insufficient for drones, too many layers result in excessive load), and the three-layer range matches the short-range flight capability of drones, resulting in short response times.

[0103] S320, send a close-up observation command to the selected close-up observation UAV, the close-up observation command including the trigger state grid number, spatial coordinate range and the current position of the moving target;

[0104] The S330 controls each close-in observation UAV to adjust its flight course toward the triggered state grid according to the close-in observation command, and adjusts the center of the sensor's field of view to align with the moving target;

[0105] The S340 controls all close-in observation UAVs to photograph and track moving targets from different angles, and shares the real-time location data of the moving targets through inter-UAV communication links.

[0106] like Figure 5 As shown, the step of dynamically deducing the predicted path grid sequence that a moving target may traverse within a preset time period in a shared 3D virtual grid map specifically includes:

[0107] S410 aggregates the position coordinates of moving targets in a three-dimensional virtual grid measured in real time by each close-in observation UAV through an inter-machine communication link, forming a position data set.

[0108] S420 performs a spatial coordinate system transformation on the position data set. Using the position data set at the same moment, the instantaneous velocity vector, instantaneous velocity magnitude, and direction of motion of the moving target in three-dimensional space are calculated by geometric solution method. The motion trend curve of the moving target is fitted based on the velocity vector changes at consecutive moments.

[0109] (1) Spatial coordinate system transformation (ENU geodetic coordinate system)

[0110] ;

[0111] : Attitude rotation matrix of UAV i (α=roll angle, β=pitch angle, γ=yaw angle);

[0112] : The relative position of the target in the local coordinate system of UAV i;

[0113] : The absolute position of UAV i in the ENU coordinate system;

[0114] The target's unified position in the ENU coordinate system.

[0115] (2) Target fusion position at time t

[0116] ;

[0117] Number of drones used for close-range observation; These are the weighting coefficients. ( (where i is the distance from drone i to the target; the closer the distance, the higher the weight).

[0118] : The precise fusion position of the target at time t.

[0119] (3) Instantaneous velocity vector, instantaneous velocity magnitude and direction of motion

[0120] ;

[0121] ;

[0122] ;

[0123] : Sampling time interval;

[0124] Instantaneous velocity vector;

[0125] : Instantaneous velocity magnitude;

[0126] : Unit vector in the direction of velocity.

[0127] (4) Motion trend curve (3rd order polynomial fitting)

[0128] ;

[0129] (Time difference relative to the start time);

[0130] : Fitting coefficients obtained by the least squares method.

[0131] S430: Starting from the current three-dimensional virtual grid where the moving target is located, along the direction of the motion trend curve, and combined with the instantaneous speed of the moving target, predict the three-dimensional virtual grids that the moving target will pass through in a future preset time period, and generate a predicted path grid sequence according to the predicted time sequence.

[0132] (1) Definition of grid coordinates (grid side length) )

[0133] ;

[0134] (2) Predicting the time step position

[0135] ;

[0136] For the goal The three-dimensional position at any given moment; Preset prediction duration; (Time step); .

[0137] (3) Predicted grid number

[0138] ;

[0139] The target is in After a certain time, the x, y, and z direction indices of the grid in which it is located;

[0140] Round down;

[0141] After deduplication, the predicted grid sequence is obtained by sorting by time. .

[0142] S440, in the shared three-dimensional virtual mesh state information, a pre-triggering state identifier is assigned to each three-dimensional virtual mesh in the predicted path mesh sequence. The pre-triggering state identifier includes the prediction time window in which a moving target is predicted to enter the three-dimensional virtual mesh, as well as the predicted motion direction and velocity information of the moving target.

[0143] like Figure 6 As shown, sending early warning commands to drones around the pre-triggered state grid specifically includes:

[0144] S510, based on the number of each pre-triggering state grid in the predicted path grid sequence and the predicted time window, select the UAVs whose deployment locations can cover the pre-triggering state grid, and send a warning instruction to the selected UAVs. The warning instruction includes the pre-triggering state grid number, the predicted time window, the predicted direction of movement, and the predicted instantaneous speed.

[0145] (1) Pre-triggered grid center coordinates

[0146] ;

[0147] The three-dimensional center coordinates of the pre-triggered mesh; For the x, y, z direction indices of the pre-triggered mesh,

[0148] (2) Coverage judgment criteria

[0149] ;

[0150] : Deployment location of drone m; Let m be the straight-line distance from the m-th drone to the center of the pre-triggered grid. To enable drone sensors to effectively detect targets at the maximum distance, This means it is determined to be covered.

[0151] The S520 controls the UAV that receives the early warning command to adjust its flight path according to the predicted time window and its current position in order to approach the periphery of the pre-triggered state grid in advance, and adjusts the pointing angle of the sensor so that the sensor's field of view covers the entrance direction of the pre-triggered state grid in advance, so as to ensure that the moving target can be captured by the sensor when it enters the pre-triggered state grid.

[0152] Pre-adjusting the position / viewpoint can avoid monitoring interruptions, ensuring that the target is captured as soon as it enters the frame, with no response delay; improve observation stability: complete attitude stabilization in advance to reduce data noise; optimize energy consumption: avoid sudden acceleration / turning to improve battery life.

[0153] (1) Target location of the UAV

[0154] ;

[0155] : The unit vector pointing the drone towards the center of the grid; 0.8 is the offset coefficient (to avoid being too close).

[0156] (2) Flight speed / heading angle / pitch angle

[0157] ;

[0158] ;

[0159]

[0160] For the drone's flight speed, For the drone's heading angle, For the drone's pitch angle, To predict the start time, The current time;

[0161] , , For pre-triggered mesh The center coordinates;

[0162] (3) Sensor pointing angle (aligned with the grid entrance)

[0163] ;

[0164] ;

[0165] For the sensor's heading angle, The sensor's pitch angle;

[0166] , , These are the coordinates of the grid entry point, i.e., the position where the target is expected to enter the pre-triggered grid.

[0167] like Figure 7 As shown, the real-time updating of the triggered state grid and the pre-triggered state grid specifically includes:

[0168] S610: Establish a data fusion node. Through the inter-machine communication link, the mobile target detection data acquired by all close-in observation UAVs and UAVs that have received early warning commands are aggregated to the data fusion node. The multi-source detection data is spatiotemporally aligned and data correlation processed to generate a fused mobile target state estimate.

[0169] S620 updates the position coordinates of the moving target using the fused moving target state estimate, and inputs the updated position coordinates into the motion trend curve fitting process to recalculate the instantaneous motion velocity vector, instantaneous motion velocity magnitude and motion direction of the moving target, so that the motion model reflects the latest motion state of the moving target in real time;

[0170] S630: Based on the updated position coordinates of the moving target, determine whether the moving target has entered a new three-dimensional virtual grid. If it has entered a new three-dimensional virtual grid, mark the newly entered three-dimensional virtual grid as a trigger state grid. At the same time, re-determine the predicted path grid sequence based on the updated motion trend curve.

[0171] Identifying new grids and re-engineering paths can focus resources on the current target grid, avoiding waste, and can also update paths based on the latest location, reducing prediction error rates, canceling invalid warnings, and deploying drones to new path grids to avoid missed / false detections.

[0172] Determine whether to enter a new grid:

[0173] ;

[0174] The new grid number is the x, y, z index of the grid corresponding to the target update position; This is the old grid number, i.e., the x, y, z index of the grid where the target was located at the previous moment.

[0175] S640: Remove the pre-triggering status mark from the grids in the original predicted path grid sequence that have become invalid because the prediction time window has expired and the moving target has not actually entered. Remove the pre-triggering status mark from the grids in the original predicted path grid sequence that are not included in the re-derived predicted path grid sequence. Mark the newly derived predicted path grids as pre-triggering state grids to achieve dynamic updating of the triggering state grids and pre-triggering state grids.

[0176] Figure 8 The structural block diagram of the multi-drone collaborative intelligent security patrol system for nighttime use provided in this embodiment of the invention is as follows: Figure 8 As shown, the system includes:

[0177] The grid division module 100 is used to divide the inspection area into a continuous three-dimensional virtual grid in three-dimensional space, assign deployment positions to each UAV, monitor the three-dimensional virtual grid around the corresponding deployment position, and share the status information of the entire three-dimensional virtual grid through inter-machine communication.

[0178] The target detection module 200 is used to mark the three-dimensional virtual grid as a trigger state grid when any UAV detects a moving target in the three-dimensional virtual grid, and to locate the number and spatial position of the trigger state grid.

[0179] The UAV screening and scheduling module 300 is used to screen UAVs whose deployment location is within the three-layer three-dimensional virtual grid centered on the trigger state grid as close-range observation UAVs, schedule close-range observation UAVs to view them, adjust the flight attitude and sensor pointing of close-range observation UAVs, and continuously observe moving targets.

[0180] The target motion data analysis module 400 is used to acquire all the moving target position data collected by the close-range observation UAV, calculate the instantaneous motion speed, direction vector and motion trend of the moving target in real time, and dynamically deduce the predicted path grid sequence that the moving target may pass through in the future within a preset time period in the shared three-dimensional virtual grid map, and mark it as a pre-trigger state grid.

[0181] The warning command sending module 500 is used to send warning commands to drones around the pre-triggered state grid, so that the drones around the pre-triggered state grid can adjust their sensor perspective in advance and align with the predicted path grid.

[0182] The data fusion and update module 600 is used to fuse all UAV detection data in real time, correct the position and motion model of moving targets, and update the triggered state grid and pre-triggered state grid in real time.

[0183] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0184] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention. Therefore, the scope of protection of this patent should be determined by the appended claims.

[0185] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for multi-drone collaborative intelligent security patrol inspection at night, characterized in that: The method includes: The inspection area is divided into a continuous three-dimensional virtual grid in three-dimensional space. Each drone is assigned a deployment position. Each drone monitors the three-dimensional virtual grid around its corresponding deployment position and shares the status information of the entire three-dimensional virtual grid through inter-drone communication. When any UAV detects a moving target within the 3D virtual grid, the 3D virtual grid is marked as a trigger state grid, and the trigger state grid's number and spatial location are determined. UAVs deployed within the three-layer 3D virtual grid centered on the trigger state grid are selected as close-range observation UAVs. The close-range observation UAVs are scheduled to observe and their flight attitude and sensor pointing are adjusted to continuously observe moving targets. Acquire all moving target location data collected by close-range observation drones, calculate the instantaneous speed, direction vector and movement trend of the moving target in real time, and dynamically deduce the predicted path grid sequence that the moving target may pass through in the future within a preset time period in the shared three-dimensional virtual grid map, and mark it as a pre-trigger state grid. Send early warning commands to drones around the pre-triggered state grid so that the drones around the pre-triggered state grid can adjust their sensor views in advance and align them with the predicted path grid. The detection data from all UAVs are fused in real time to correct the position and motion model of moving targets, and the triggered state grid and pre-triggered state grid are updated in real time.

2. The method according to claim 1, characterized in that, The allocation of deployment locations for each drone specifically includes: The inspection area is divided into continuous three-dimensional virtual grids of equal size in three-dimensional space; Each drone is assigned a deployment location, ensuring that the deployment locations of each drone are evenly distributed across the inspection area. Each drone monitors eight adjacent three-dimensional virtual grids in three-dimensional space, centered on its own deployment location. This ensures that the three-dimensional virtual grids surrounding the corresponding deployment location monitored by each drone cover the entire inspection area, and that the monitoring ranges of adjacent drones overlap. Each drone broadcasts its detected status information to other drones via an inter-drone communication link and receives status information of the 3D virtual mesh broadcast by other drones.

3. The method according to claim 2, characterized in that, The marker is a trigger state grid, specifically including: Based on the position coordinates of any UAV when it detects a moving target within a 3D virtual grid, generate a trigger state grid marker; In the shared 3D virtual mesh status information, the status field of the 3D virtual mesh that detects a moving target is updated to the trigger status, and the trigger time is recorded; Extract the unique number and spatial coordinate range of the 3D virtual mesh corresponding to the trigger state mesh, and store the number and spatial coordinate range in association with the trigger time.

4. The method according to claim 3, characterized in that, The aforementioned dispatching of close-range observation drones for inspection specifically includes: Based on the trigger state grid number and spatial coordinate range, calculate the spatial area covered by the three-layer three-dimensional virtual grid extending outward from the trigger state grid, and designate all UAVs deployed within this spatial area as close-range observation UAVs; Send close-range observation instructions to the selected close-range observation drones. The close-range observation instructions include the trigger state grid number, spatial coordinate range, and the current position of the moving target. Control each close-in observation UAV to adjust its flight course toward the trigger state grid according to the close-in observation command, and adjust the center of the sensor field of view to align with the moving target; Control all close-range observation drones to photograph and track moving targets from different angles, and share real-time moving target location data through inter-drone communication links.

5. The method according to claim 4, characterized in that, The process of dynamically deducing the predicted path grid sequence that a moving target may traverse within a preset time period in a shared 3D virtual grid map specifically includes: The location coordinates of the moving target in the three-dimensional virtual grid, measured in real time by each close-range observation UAV, are collected through the inter-machine communication link to form a location data set. The spatial coordinate system is transformed into the position data set. Using the position data set at the same moment, the instantaneous velocity vector, instantaneous velocity magnitude and direction of motion of the moving target in three-dimensional space are calculated by geometric solution method. The motion trend curve of the moving target is fitted according to the velocity vector changes at consecutive moments. Starting from the current three-dimensional virtual grid where the moving target is located, along the direction of the motion trend curve, and combined with the instantaneous speed of the moving target, predict the three-dimensional virtual grids that the moving target will pass through in a future preset time period, and generate a predicted path grid sequence according to the predicted time sequence. In the shared 3D virtual mesh state information, a pre-triggering state identifier is assigned to each 3D virtual mesh in the predicted path mesh sequence. The pre-triggering state identifier includes the prediction time window in which a moving target is predicted to enter the 3D virtual mesh, as well as the predicted motion direction and velocity information of the moving target.

6. The method according to claim 5, characterized in that, Sending early warning commands to drones surrounding the pre-triggered state grid specifically includes: Based on the number of each pre-triggering state grid and the prediction time window in the predicted path grid sequence, drones whose deployment locations can cover the pre-triggering state grid are selected, and warning instructions are sent to the selected drones. The warning instructions include the pre-triggering state grid number, the prediction time window, the predicted direction of movement, and the predicted instantaneous speed. The drone that receives the early warning command adjusts its flight path according to the predicted time window and its current position to approach the periphery of the pre-triggered state grid in advance, and adjusts the pointing angle of the sensor so that the sensor's field of view covers the entrance direction of the pre-triggered state grid in advance, so that the moving target can be captured by the sensor when it enters the pre-triggered state grid.

7. The method according to claim 6, characterized in that, The real-time updating of the triggered state grid and the pre-triggered state grid specifically includes: Establish a data fusion node, and aggregate the mobile target detection data acquired by all close-in observation UAVs and UAVs that have received early warning commands to the data fusion node through the inter-machine communication link. Perform spatiotemporal alignment and data association processing on the multi-source detection data to generate a fused mobile target state estimate. The position coordinates of the moving target are updated using the fused moving target state estimate. The updated position coordinates are then input into the motion trend curve fitting process to recalculate the instantaneous velocity vector, instantaneous velocity magnitude, and motion direction of the moving target, so that the motion model reflects the latest motion state of the moving target in real time. Based on the updated position coordinates of the moving target, determine whether the moving target has entered a new 3D virtual grid. If it has entered a new 3D virtual grid, mark the newly entered 3D virtual grid as a trigger state grid. At the same time, re-determine the predicted path grid sequence based on the updated motion trend curve. The pre-triggering status mark is removed from the original predicted path grid sequence for grids that have become invalid because the prediction time window has expired and the moving target has not actually entered. The pre-triggering status mark is removed from the grids in the original predicted path grid sequence that are not included in the re-derived predicted path grid sequence. The newly derived predicted path grids are marked as pre-triggering state grids, thereby realizing the dynamic update of the triggering state grids and the pre-triggering state grids.

8. A multi-drone collaborative intelligent security patrol system for nighttime operations, characterized in that: The system includes: The grid division module is used to divide the inspection area into continuous three-dimensional virtual grids in three-dimensional space, assign deployment positions to each UAV, and each UAV monitors the three-dimensional virtual grids around its corresponding deployment position and shares the status information of the entire three-dimensional virtual grid through inter-UAV communication. The target detection module is used to mark the three-dimensional virtual grid as a trigger state grid when any UAV detects a moving target within the three-dimensional virtual grid, and to locate the number and spatial position of the trigger state grid. The UAV screening and scheduling module is used to screen UAVs whose deployment location is within the three-layer three-dimensional virtual grid centered on the trigger state grid as close-range observation UAVs, schedule close-range observation UAVs to view them, adjust the flight attitude and sensor pointing of close-range observation UAVs, and continuously observe moving targets. The target motion data analysis module is used to acquire all the moving target position data collected by the close-range observation UAV, calculate the instantaneous motion speed, direction vector and motion trend of the moving target in real time, and dynamically deduce the predicted path grid sequence that the moving target may pass through in the future within a preset time period in the shared three-dimensional virtual grid map, and mark it as a pre-trigger state grid. The early warning command sending module is used to send early warning commands to drones around the pre-triggered state grid, so that the drones around the pre-triggered state grid can adjust their sensor perspective in advance and align with the predicted path grid. The data fusion and update module is used to fuse all UAV detection data in real time, correct the position and motion model of moving targets, and update the triggered state grid and pre-triggered state grid in real time.

Images