Real-time monitoring system of intelligent unmanned aerial vehicle
By utilizing the environmental perception, positioning control, and linkage decision-making modules of the intelligent drone real-time monitoring system, accurate identification of bird targets and multi-level bird deterrence decisions are achieved, solving the problems of low efficiency and poor accuracy of existing drone bird deterrence methods and improving the bird deterrence effect.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- NANJING NEW YUEYANG TECH CO LTD
- Filing Date
- 2025-06-03
- Publication Date
- 2026-06-09
AI Technical Summary
Existing methods of using drones to scare away birds suffer from low efficiency, poor accuracy, and a lack of prediction and coordination of bird targets, resulting in poor bird-scaring effects.
An intelligent drone real-time monitoring system is adopted, which combines an environmental perception module, an intelligent positioning and control module, and a linkage decision-making module. It uses a visual recognition model to identify and locate targets, record and predict target movement trajectories, and builds a linkage model to make multi-level bird deterrence decisions.
It enables accurate identification of bird targets in the operation scene and multi-level bird deterrence decision-making, improves bird deterrence efficiency and drone linkage, and improves the low efficiency, low precision and single bird deterrence mode of traditional bird deterrence methods.
Smart Images
Figure CN120635752B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of unmanned aerial vehicle (UAV) monitoring, specifically a real-time monitoring system for intelligent UAVs. Background Technology
[0002] With the rapid development of drone technology, its application in agriculture, airports, power line inspection and other fields is becoming increasingly widespread. However, in these scenarios, bird activity poses a threat to the safe operation of the operation, and traditional bird control methods suffer from low efficiency and poor accuracy. Using bird control drones for scene operations has become a new trend. However, current drone bird control still has shortcomings. It is mostly a single scene inspection and target tracking to drive away birds. The bird drive-away route cannot be predicted, and there is a lack of coordination in the deployment of drones, resulting in poor bird control effect and low accuracy of current drones. Summary of the Invention
[0003] The purpose of this invention is to provide a real-time monitoring system for intelligent unmanned aerial vehicles (UAVs) to solve the problems raised in the prior art.
[0004] To achieve the above objectives, the present invention provides the following technical solution:
[0005] A real-time monitoring system for intelligent unmanned aerial vehicles (UAVs), the system comprising an environmental perception module, an intelligent positioning and control module, a linkage decision-making module, and a data feedback module:
[0006] The environmental perception module acquires real-time environmental data of the operational scenario through the environmental acquisition drone, and formulates the drone's operational flight trajectory based on the operational scenario environmental data; the intelligent positioning and control module acquires real-time spatial image data corresponding to the drone's operational flight trajectory according to the drone's operational flight, identifies and retrieves targets in the spatial image data by constructing a visual recognition model, and performs real-time target positioning analysis based on the recognition results; the linkage decision module identifies target positioning data based on the visual recognition model, records the real-time target motion trajectory, and performs real-time target motion trajectory prediction simulation; it performs anomaly probability distribution analysis based on the real-time target motion trajectory prediction simulation, determines the drone linkage model based on the analysis results, and plans drone decisions based on the drone linkage model.
[0007] Furthermore, the environmental perception module includes a work scene environmental data acquisition unit and a flight trajectory planning unit;
[0008] The work scenario environmental data acquisition unit acquires the safe working area in the work scenario and collects real-time environmental data based on the safe working area; it achieves environmental data collection through an environmental acquisition drone combined with sensing devices; the real-time environmental data includes temperature data, wind speed data, humidity data, pressure data, etc.
[0009] The flight trajectory planning unit determines the drone's operational parameter settings based on real-time operational scenario environmental data and plans the drone's operational inspection route. The drone's operational parameter settings refer to determining the drone's flight parameters based on the real-time environment, including speed parameters, battery parameters, and turning angle.
[0010] Furthermore, the safe operating range is the safe hemispherical space in the operating scenario where equipment or personnel are performing operations; the operating inspection route refers to the inspection flight route of the drone within the safe space of the operating scenario; interference detection is performed by the drone on the inspection flight route.
[0011] Furthermore, the intelligent positioning control module includes a visual retrieval unit and a positioning analysis unit;
[0012] The visual retrieval unit collects real-time spatial image data along the drone's operation and inspection route and transmits it remotely to the monitoring terminal; at the monitoring terminal, a visual recognition model is constructed to perform continuous time target recognition on the spatial image data.
[0013] The positioning analysis unit performs positioning analysis on the target recognition results of real-time spatial image data on the UAV's operation and inspection route based on the visual recognition model. If the recognition result is no target, the UAV's flight positioning data is continuously recorded and transmitted to the monitoring terminal. If the recognition result is that a target exists, the positioning of both the UAV and the target is recorded simultaneously, and the spatial relative position between the UAV and the target is determined. The spatial relative position is determined by performing spatial point processing on the UAV and the target respectively, taking the spatial outer spheres of the UAV and the target respectively, using the center of the sphere as the spatial positioning node of the UAV and the target respectively, and connecting the two nodes to obtain the spatial relative position data between the UAV and the target. The positioning record of the UAV and the target and the spatial relative position data are transmitted to the monitoring terminal.
[0014] Furthermore, the specific steps for performing continuous-time target recognition on spatial image data by constructing a visual recognition model are as follows:
[0015] By filtering the scene background pixels of the real-time spatial image data collected by the UAV, dynamic flying targets in the spatial image data can be extracted.
[0016] By decomposing the dynamic flight target image data into pixels, the contour pixels of the dynamic flight target image are obtained. The contour pixels are then fused, and adjacent contour pixels are connected to determine the slope of each line segment. The difference between the slopes of adjacent line segments is calculated, and a slope judgment threshold is set. If the absolute value of the slope difference between adjacent line segments is less than or equal to the slope judgment threshold, the pixels on the adjacent line segments are filtered out; otherwise, they are retained. This process yields the contour attitude image data of the dynamic flight target.
[0017] Based on the fusion processing results of the contour pixels of the dynamic flight target image, the remaining contour pixels of the dynamic flight target image data are labeled. By mapping the attitude image data of the dynamic flight target in the continuous temporal and spatial image data to the same coordinate system, the spatial position changes of the same-labeled contour pixels in the continuous temporal and spatial image data are recorded, and the motion trajectories of the same-labeled pixels are obtained through connection processing. By comparing the curvature values of the continuous-time motion trajectories of adjacent labeled contour pixels, combined with the contour attitude image data of the dynamic flight target in the continuous time, a matching search is performed in the database to determine the target. During the matching search, an initial search is performed first using the contour attitude image data of the dynamic flight target in the continuous time; secondly, a further search is performed based on the curvature values of the continuous-time motion trajectories of adjacent labeled contour pixels, by setting a curvature threshold. The system filters out database search targets with curvature values below a certain threshold. It then calculates the type distribution of the remaining targets and determines their weight. If a target's weight exceeds a set threshold, it is considered a valid target; otherwise, the data is fed back to the monitoring station for manual judgment. Specifically, recording continuous temporal and spatial image data shows at least one complete trajectory movement for pixels with the same outline. This means recording stops when the spatial position changes of pixels with the same outline begin to repeat. Conversely, if pixels with the same outline are all at the same point within a continuous time frame, a response time is set, and recording continues until a complete trajectory movement is observed. If no correct data is recorded within the response time, recording stops, and the data is fed back to the monitoring station for manual judgment.
[0018] Furthermore, the linkage decision-making module includes a target trajectory tracking and analysis unit and a linkage decision-making unit;
[0019] The target trajectory tracking and analysis unit determines the target's real-time moving trajectory based on the target's real-time positioning data and the recorded target positioning data; and predicts and simulates the target's movement trajectory based on the target's real-time moving trajectory.
[0020] The linkage decision-making unit performs anomaly analysis of the operation scenario based on the real-time motion trajectory prediction simulation, divides the safe operation range of the operation scenario into levels, determines the probability distribution analysis of the impact of the real-time motion trajectory prediction simulation generated route on the anomalies of different levels of space in the safe operation range of the operation scenario, and constructs a UAV linkage model based on the analysis results to determine the decision.
[0021] Furthermore, the specific steps for predicting and simulating the target's trajectory based on the target's real-time and actual movement trajectory are as follows:
[0022] By collecting positioning data of the target and recording its movement trajectory, and using the target spatial positioning point recorded at the current real-time point as the starting node of the predicted route, the movement trajectory of the target in the subsequent prediction period is predicted; the prediction period is the time required for the UAV with the shortest spatial relative position to the target to reach the target spatial position node at the current time.
[0023] By obtaining the tangents of corresponding spatial nodes on the target's already moving trajectory, the tangent angles of the corresponding spatial nodes are recorded. The maximum value A of the tangent angles corresponding to each spatial node on the target's already moving trajectory is determined. The tangent angle is the angle between the tangent of the corresponding spatial node and the horizontal line. The extreme points of the target's already moving trajectory are determined, and the vertical distance D between the maximum and minimum points is calculated, with D as the fluctuation distance of the already moving trajectory. At the starting node of the predicted route, with [0, A] as the tangent angle variation range and [0, D] as the fluctuation distance range of the predicted trajectory, a random motion trajectory is generated through a full traversal. The full traversal generation of the random route refers to the full traversal of the tangent angle variation range [0, A] from the starting node of the predicted route towards the work scene, generating a route for each tangent angle in the range. Based on this, the fluctuation distance of the generated route is limited to the full traversal between [0, D] during the route generation process for each tangent angle.
[0024] Furthermore, the safe working range of the work scenario is hierarchically divided by determining the center of the safe working range within the work scenario, and then dividing the hemispherical space of the safe working range into a first safety layer, a second safety layer, and a third safety layer by planning a first safety distance, a second safety distance, and a third safety distance with the center of the range as the radii of the spheres; wherein the first safety distance is greater than the second safety distance, and the second safety distance is greater than the third safety distance;
[0025] Based on the hierarchical division of the safe working area of the work scenario, and using the random motion trajectory generated from the starting node of the predicted route of the real-time target, a probability distribution analysis of the abnormal impact of different spatial levels within the safe working area of the work scenario is conducted. This analysis is performed by comprehensively considering the total number of random motion trajectory routes generated from the starting node of the predicted route of the real-time target, and then comprehensively considering the distribution of each generated random motion trajectory route within different spatial levels of the safe working area of the work scenario. The percentages of random motion trajectory routes involving the third safety layer, the second safety layer, and the first safety layer at the current time point are calculated respectively. The results are then compared between each level. Based on the corresponding percentage thresholds, the drone linkage model is determined, and the decision-making levels are defined as: regular level decision, first-level decision, second-level decision, and third-level decision. According to the priority of the level comparison, the third level is greater than the second level, and the second level is greater than the first level. Therefore, if the percentage of random motion trajectories involving the third safety layer is greater than the corresponding level's percentage threshold, the third-level decision is executed directly; otherwise, the percentage of random motion trajectories involving the second safety layer is checked. If it is greater than the threshold, the second-level decision is executed directly; otherwise, the regular level decision is executed.
[0026] The conventional level decision-making involves determining the drone with the shortest relative distance to the target spatial positioning point at the current time and then tracking and driving the target away from the safe operating range of the work scene in real time.
[0027] The first-level decision-making process involves constructing a drone linkage model by linking two drones. The drone with the shortest relative distance to the target's spatial location at the current time point tracks and drives away the target in real time. The first-level spatial point with the shortest vertical distance to the target's spatial location at the current time point is determined. The drone with the shortest distance to the first-level spatial point at the current time is then directed to fly parallel to the target on the first level and to drive away the target when it enters its driving range. The first level is the outer boundary curved surface on the first safety layer.
[0028] The second level of decision-making involves constructing a drone linkage model by linking three drones. Based on the first level of decision-making, the second level spatial point with the shortest vertical distance to the target spatial positioning point at the current time is determined. The drone with the shortest distance to the second level spatial point at the current time is then mobilized to fly parallel to the target on the second level and to drive away the target when it enters its expulsion range. The second level is the outer boundary curved surface on the second safety layer.
[0029] The third-level decision-making involves constructing a drone linkage model by linking four drones. Based on the second-level decision-making, the third-level spatial point with the shortest vertical distance to the target spatial positioning point at the current time is determined. The drone with the shortest distance to the third-level spatial point at the current time is then mobilized to fly parallel to the target on the third level and to drive away the target when it enters its expulsion range. The third level is the outer boundary curved surface on the third safety layer.
[0030] Furthermore, the data feedback module includes an alert unit and an event recording unit;
[0031] When the warning unit makes a decision regarding the drone in the operational scenario, it sends a warning notification to the monitoring terminal.
[0032] The event recording unit logs the drone decision-making behavior events that occur in the operation scenario.
[0033] Compared with the prior art, the beneficial effects of the present invention are:
[0034] This invention combines a visual recognition model and a linkage model to achieve accurate target identification and multi-level bird control decision generation for bird-repelling drones in operational scenarios. It identifies targets and records their continuous motion trajectories to predict subsequent routes. Based on this, it analyzes the abnormal distribution of the predicted routes within the operational scenario to construct a linkage model to determine bird control decisions. This invention accurately identifies targets and predicts effective paths, thereby enabling efficient linkage of bird-repelling drones to achieve bird control behavior in the scenario. This invention effectively improves the low efficiency and low precision of traditional bird control methods and the single bird control mode of current bird-repelling drones, enhancing the bird control efficiency and drone linkage capabilities in operational scenarios. Attached Figure Description
[0035] Figure 1 This is a schematic diagram of the structure of a real-time monitoring system for an intelligent unmanned aerial vehicle (UAV) according to the present invention. Detailed Implementation
[0036] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0037] Example: Figure 1 As shown, the present invention provides a technical solution:
[0038] A real-time monitoring system for intelligent bird-repelling drones is provided, the target of which is birds;
[0039] The system includes an environmental perception module, an intelligent positioning and control module, a linkage decision-making module, and a data feedback module.
[0040] The environmental perception module acquires real-time environmental data of the operational scenario through an environmental acquisition drone, and formulates the flight trajectory of the bird-repelling drone based on the operational scenario environmental data. The intelligent positioning and control module collects real-time spatial image data corresponding to the flight trajectory of the bird-repelling drone, constructs a visual recognition model to identify and retrieve bird targets in the spatial image data, and performs real-time positioning analysis of bird targets based on the recognition results. The linkage decision module identifies bird target positioning data based on the visual recognition model, records the real-time movement trajectory of bird targets, and performs real-time prediction and simulation of bird target movement trajectory. It performs anomaly probability distribution analysis based on the real-time prediction and simulation of bird target movement trajectory, determines the drone linkage model based on the analysis results, and plans the drone bird-repelling decision based on the drone linkage model.
[0041] Furthermore, the environmental perception module includes a work scene environmental data acquisition unit and a flight trajectory planning unit;
[0042] The work scenario environmental data acquisition unit acquires the safe working area in the work scenario and collects real-time environmental data based on the safe working area; it achieves environmental data collection through an environmental acquisition drone combined with sensing devices; the real-time environmental data includes temperature data, wind speed data, humidity data, pressure data, etc.
[0043] The flight trajectory planning unit determines the operational parameter settings of the bird-repelling drone based on real-time operational environment data and plans the operational inspection route of the bird-repelling drone. The operational parameter settings of the bird-repelling drone refer to determining the flight parameters of the bird-repelling drone based on the real-time environment, including speed parameters, battery parameters, turning angle, etc.
[0044] Furthermore, the safe operating range is the safe hemispherical space in the operating scenario where equipment or personnel are performing operations; the operating inspection route refers to the inspection flight route of the bird-repelling drone within the safe space of the operating scenario; the bird-repelling drone detects disturbing birds along the inspection flight route.
[0045] Furthermore, the intelligent positioning control module includes a visual retrieval unit and a positioning analysis unit;
[0046] The visual retrieval unit collects real-time spatial image data along the patrol route of the bird-scaring drone and transmits it remotely to the monitoring terminal; at the monitoring terminal, a visual recognition model is constructed to continuously identify bird targets in the spatial image data.
[0047] The positioning analysis unit performs positioning analysis on the bird target identification results of real-time spatial image data on the bird-repelling drone's operation and inspection route based on the visual recognition model. If the identification result is no bird target, the flight positioning data of the bird-repelling drone is continuously recorded and transmitted to the monitoring terminal. If the identification result is that a bird target exists, the positioning of both the bird-repelling drone and the bird target is recorded simultaneously, and the spatial relative position between the bird-repelling drone and the bird target is determined. The spatial relative position is determined by performing spatial point processing on the bird-repelling drone and the bird target respectively, taking the spatial outer sphere of the bird-repelling drone and the bird target respectively, taking the center of the sphere as the spatial positioning node of the bird-repelling drone and the bird target respectively, and connecting the two nodes to obtain the spatial relative position data between the bird-repelling drone and the bird target. The positioning record of the bird-repelling drone and the bird target and the spatial relative position data are transmitted to the monitoring terminal.
[0048] Furthermore, the specific steps for continuous-time bird target recognition using a visual recognition model are as follows:
[0049] By filtering the scene background pixels of the real-time spatial image data collected by the bird-scaring drone, dynamic flying targets in the spatial image data can be extracted.
[0050] By decomposing the dynamic flight target image data into pixels, the contour pixels of the dynamic flight target image are obtained. The contour pixels are then fused, and adjacent contour pixels are connected to determine the slope of each line segment. The difference between the slopes of adjacent line segments is calculated, and a slope judgment threshold is set. If the absolute value of the slope difference between adjacent line segments is less than or equal to the slope judgment threshold, the pixels on the adjacent line segments are filtered out; otherwise, they are retained. This process yields the contour attitude image data of the dynamic flight target.
[0051] Based on the fusion processing results of the contour pixels of dynamic flight target images, the remaining contour pixels in the dynamic flight target image data are labeled. By mapping the attitude image data of dynamic flight targets in continuous temporal and spatial image data to the same coordinate system, the spatial position changes of contour pixels with the same label in the continuous temporal and spatial image data are recorded, and the motion trajectories of pixels with the same label within a continuous time period are obtained through connection processing. By comparing the curvature values of the continuous time motion trajectories of adjacent labeled contour pixels, combined with the contour attitude image data of dynamic flight targets within a continuous time period, a matching search is performed in the database to identify bird targets. During the matching search, an initial search is performed first using the contour attitude image data of dynamic flight targets within a continuous time period; secondly, a further search is performed based on the curvature values of the continuous time motion trajectories of adjacent labeled contour pixels, by setting a curvature threshold. The system filters out database search targets with curvature values below a certain threshold. It then calculates the type distribution of the remaining database search targets and determines the weight of bird targets. If a bird target's weight is higher than a set threshold, it is identified as a bird target; otherwise, the data is fed back to the monitoring station for manual judgment. Specifically, it records that at least one complete trajectory movement exists for the spatial position changes of pixels with the same outline in continuous temporal and spatial image data. This means that recording stops when the spatial position changes of pixels with the same outline begin to repeat. Conversely, if the spatial position changes of pixels with the same outline are all at the same point within a continuous time period, a response time is set, and recording continues until a complete trajectory movement is captured. If no correct data is recorded within the response time, recording stops, and the data is fed back to the monitoring station for manual judgment.
[0052] Furthermore, the linkage decision-making module includes a target trajectory tracking and analysis unit and a linkage decision-making unit;
[0053] The target trajectory tracking and analysis unit determines the real-time moving trajectory of the bird target based on the real-time positioning data of the bird target and through the recorded positioning data of the bird target; and predicts and simulates the moving trajectory of the bird target based on the real-time moving trajectory of the bird target.
[0054] The linked bird deterrence decision unit performs anomaly analysis of the operation scenario based on the real-time bird movement trajectory prediction simulation, divides the safe operation range of the operation scenario into levels, determines the probability distribution analysis of the impact of the real-time bird movement trajectory prediction simulation generated route on the anomalies of different levels of space in the safe operation range of the operation scenario, and constructs a drone-linked bird deterrence model based on the analysis results to determine the bird deterrence decision.
[0055] Furthermore, the specific steps for predicting and simulating the movement trajectory of the bird target based on its real-time movement trajectory are as follows:
[0056] By collecting location data of bird targets and recording their movement trajectory, and using the bird target's spatial location point recorded at the current real-time point as the starting node of the predicted route, the movement trajectory of the bird target in the subsequent prediction period is predicted; the prediction period is the time required for the bird-repelling drone with the shortest spatial relative distance to the bird target's spatial location node to reach the bird target's spatial location node at the current time point.
[0057] By obtaining the tangents of corresponding spatial nodes on the trajectory of the bird target, the tangent angles of the corresponding spatial nodes are recorded. The maximum value A of the tangent angles corresponding to each spatial node on the trajectory of the bird target is determined. The tangent angle is the angle between the tangent of the corresponding spatial node and the horizontal line. The extreme points of the trajectory of the bird target are determined, and the vertical distance D between the maximum and minimum points is calculated, with D as the fluctuation distance of the trajectory. A random trajectory is generated by fully traversing the tangent angle variation range of [0, A] from the starting node of the predicted route, with [0, D] as the fluctuation distance range of the predicted trajectory. The full traversal generation of the random route means that, from the starting node of the predicted route towards the work scene, the tangent angle variation range of [0, A] is fully traversed, and a route is generated for each tangent angle in the range. On this basis, during the route generation process for each tangent angle, the fluctuation distance of the generated route is limited to the full traversal between [0, D].
[0058] Furthermore, the safe working range of the work scenario is hierarchically divided by determining the center of the safe working range within the work scenario, and then dividing the hemispherical space of the safe working range into a first safety layer, a second safety layer, and a third safety layer by planning a first safety distance, a second safety distance, and a third safety distance with the center of the range as the radii of the spheres; wherein the first safety distance is greater than the second safety distance, and the second safety distance is greater than the third safety distance;
[0059] Based on the hierarchical division of the safe working area of the work scenario, and using the random motion trajectory routes generated from the starting nodes of the predicted routes of real-time bird targets, a probability distribution analysis of the abnormal impact of different spatial levels within the safe working area of the work scenario is conducted. This analysis is performed by comprehensively considering the total number of random motion trajectory routes generated from the starting nodes of the predicted routes of real-time bird targets, and then comprehensively considering the distribution of each generated random motion trajectory route within the different spatial levels of the safe working area of the work scenario. The percentage of random motion trajectory routes involving the third safety layer, the second safety layer, and the first safety layer at the current time point is calculated. The percentage thresholds for each level are then compared to determine... A drone-based bird control model is established, defining bird control decisions as follows: Level 3, Level 4, Level 5, and Level 6. Based on priority, Level 3 is higher than Level 2, and Level 2 is higher than Level 1. Therefore, if the proportion of random motion trajectories involving Level 3 safety exceeds the corresponding level threshold, Level 3 bird control is executed directly. Conversely, if the proportion of random motion trajectories involving Level 2 safety exceeds the threshold, Level 2 bird control is executed. Similarly, if the proportion of random motion trajectories involving Level 1 safety exceeds the threshold, Level 1 bird control is executed. Otherwise, a Level 3 bird control decision is executed.
[0060] The conventional bird deterrence decision-making process involves identifying the bird deterrence drone with the shortest relative distance to the bird target's spatial location at the current time point, and then using it to track and drive the bird target away from the safe operating area of the work scene in real time.
[0061] The first-level bird deterrence decision-making process involves linking two bird deterrence drones to construct a drone-linked bird deterrence model. The drone with the shortest relative distance to the bird target's spatial location at the current time is used to track and deter the bird target in real time. The first-level spatial point with the shortest vertical distance to the bird target's spatial location at the current time is determined, and the drone with the shortest distance to the first-level spatial point at the current time is activated to fly parallel to the bird target on the first level. When the bird target enters its deterrence range, it will take deterrence action. The first level is the outer boundary curved surface on the first safety layer.
[0062] The second-level bird deterrence decision-making involves constructing a drone-based bird deterrence linkage model using three bird deterrence drones. Based on the first-level bird deterrence decision-making, the second-level spatial point with the shortest vertical distance to the bird target's spatial location at the current time is determined. The bird deterrence drone with the shortest distance to the second-level spatial point at the current time is then activated and ordered to fly parallel to the bird target on the second level. When the bird target enters its deterrence range, it will be deterred. The second level is the outer boundary curved surface of the second safety layer.
[0063] The third-level bird deterrence decision-making involves constructing a drone-based bird deterrence model by linking four bird deterrence drones. Based on the second-level bird deterrence decision-making, the third-level spatial point with the shortest vertical distance to the bird target's spatial location at the current time is determined. The bird deterrence drone with the shortest distance to the third-level spatial point at the current time is then mobilized to fly parallel to the bird target on the third level and to drive away the bird target when it enters its deterrence range. The third level is the outer boundary curved surface on the third safety layer.
[0064] Furthermore, the data feedback module includes an alert unit and an event recording unit;
[0065] When the warning unit makes a bird-repelling decision for the bird-repelling drone in the operation scenario, it sends a warning prompt to the monitoring terminal.
[0066] The event recording unit logs bird-driving behavior events that occur in the work scenario;
[0067] The implementation example for bird deterrence decision-making is as follows:
[0068] In the current work scenario, considering the safety operation scope level of the work scenario and the random movement trajectory generated by the predicted route start node of the real-time bird target, the proportion of random movement trajectory generated by the predicted route start node of the real-time bird target in different safety operation scope levels of the work scenario is determined as follows: the proportion of random movement trajectory in the third safety layer (A3), the proportion of random movement trajectory in the second safety layer (A2), and the proportion of random movement trajectory in the first safety layer (A1) are determined respectively. The proportion of random movement trajectory in the corresponding level is compared with the proportion threshold of the corresponding level, and the bird deterrence decision to be executed in the current work scenario is determined based on the comparison result.
[0069] If A3 < a3, A2 < a2, and A1 < a1, then a conventional bird deterrence decision is executed. This decision determines the safe operating range for real-time tracking and deterring the bird target from the current spatial location point of the bird target. Here, a1 corresponds to the threshold for the proportion of random motion trajectory routes in the first safety layer; a2 corresponds to the threshold for the proportion of random motion trajectory routes in the second safety layer; and a3 corresponds to the threshold for the proportion of random motion trajectory routes in the third safety layer.
[0070] If A3 < a3, A2 < a2, and A1 ≥ a1, then the first-level bird deterrence decision is executed. This involves linking two bird deterrence drones to construct a drone-linked bird deterrence model. The bird deterrence drone with the shortest relative distance to the bird target's spatial location at the current time will track and deter the bird target in real time. The first-level spatial point with the shortest vertical distance to the bird target's spatial location at the current time will be determined. The bird deterrence drone with the shortest distance to the first-level spatial point at the current time will be mobilized and ordered to fly parallel to the bird target at the first level. When the bird target enters its deterrence range, it will take deterrence action.
[0071] If A3 < a3 and A2 ≥ a2, then according to the priority of hierarchical comparison, the third level is greater than the second level, and the second level is greater than the first level. Therefore, regardless of the comparison result between A1 and a1, the second-level bird deterrence decision is executed. This involves constructing a drone-linked bird deterrence model by linking three bird deterrence drones. Based on the first-level bird deterrence decision, the second-level spatial point with the shortest vertical distance to the bird target's spatial location point at the current time is determined. The bird deterrence drone with the shortest distance to the second-level spatial point at the current time is then mobilized and ordered to fly parallel to the bird target at the second level. When the bird target enters its deterrence range, it will perform deterrence actions. Since bird target deterrence actions are performed at the second level, the first level will be involved in the deterrence process. Therefore, the second-level bird deterrence decision and the first-level bird deterrence decision need to be executed sequentially during this process.
[0072] If A3 ≥ a3, then regardless of the comparison results between A2 and a2, or A1 and a1, the third-level bird deterrence decision is executed. This involves constructing a drone-linked bird deterrence model by linking four bird deterrence drones. Based on the second-level bird deterrence decision, the third-level spatial point with the shortest vertical distance to the bird target's spatial location at the current time is determined. The bird deterrence drone with the shortest distance to the third-level spatial point at the current time is then mobilized to fly parallel to the bird target on the third level. When the bird target enters its deterrence range, it will perform deterrence actions. Since bird target deterrence actions are performed on the third level, the second level and the lower level will be involved in the deterrence process. Therefore, the third-level bird deterrence decision, the second-level bird deterrence decision, and the first-level bird deterrence decision need to be executed sequentially during this process.
[0073] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims.
Claims
1. A real-time monitoring system for intelligent drones, characterized by: The system includes an environmental perception module, an intelligent positioning and control module, a linkage decision-making module, and a data feedback module. The environmental perception module acquires real-time environmental data of the operational scenario through the environmental acquisition drone, and formulates the drone's operational flight trajectory based on the operational scenario environmental data; the intelligent positioning and control module acquires real-time spatial image data corresponding to the drone's operational flight trajectory according to the drone's operational flight, identifies and retrieves targets in the spatial image data by constructing a visual recognition model, and performs real-time target positioning analysis based on the recognition results; the linkage decision module identifies target positioning data based on the visual recognition model, records the real-time target motion trajectory, and performs real-time target motion trajectory prediction simulation; it performs anomaly probability distribution analysis based on the real-time target motion trajectory prediction simulation, determines the drone linkage model based on the analysis results, and plans drone decisions based on the drone linkage model; The specific steps for identifying and retrieving targets from spatial image data by constructing a visual recognition model, and for performing real-time target localization analysis based on the recognition results, are as follows: By filtering the scene background pixels of the real-time spatial image data collected by the UAV, dynamic flying targets in the spatial image data can be extracted. By decomposing the dynamic flight target image data into pixels, the contour pixels of the dynamic flight target image are obtained. The contour pixels are then fused, which involves connecting adjacent contour pixels to determine the slope of each line segment. The difference between the slopes of the connecting lines of adjacent pixels is calculated, and a slope judgment threshold is set. If the absolute value of the slope difference between the connecting lines of adjacent pixels is less than or equal to the slope judgment threshold, the pixels on the connecting lines of adjacent pixels are filtered out; otherwise, they are retained. This process yields the contour attitude image data of the dynamic flight target. Based on the fusion processing results of the contour pixels of the dynamic flight target image, the remaining contour pixels of the dynamic flight target image data are labeled. By mapping the attitude image data of the dynamic flight target in the continuous time and space image data to coincide with the same coordinate system, the spatial position change data of the contour pixels with the same label in the continuous time and space image data are recorded, and the motion trajectory of the pixel with the same label in the continuous time is obtained by connecting the pixels. By comparing the curvature values of the continuous time motion trajectories of adjacent labeled contour pixels, and combining the contour attitude image data of the dynamic flight target in the continuous time, the target is identified by matching and searching the database. 2.The real-time monitoring system of an intelligent unmanned aerial vehicle according to claim 1, characterized in that: The environmental perception module includes an operational scenario environmental data acquisition unit and a flight trajectory planning unit; The work scenario environmental data acquisition unit acquires the safe working range in the work scenario and performs real-time environmental data collection based on the safe working range; it achieves environmental data collection by combining an environmental data acquisition drone with sensing devices. The flight trajectory planning unit determines the drone's operational parameter settings based on real-time operational scenario environmental data and plans the drone's operational inspection route. 3.The real-time monitoring system of an intelligent unmanned aerial vehicle according to claim 2, characterized in that: The safe operating range is the safe hemispherical space in the work scenario where equipment or personnel are performing operations; the work inspection route refers to the inspection flight route of the drone within the safe space of the work scenario; the bird-repelling drone is used to detect disturbing birds on the inspection flight route.
4. The real-time monitoring system of an intelligent unmanned aerial vehicle according to claim 3, characterized in that: The intelligent positioning control module includes a visual retrieval unit and a positioning analysis unit; The visual retrieval unit collects real-time spatial image data along the drone's operation and inspection route and transmits it remotely to the monitoring terminal; at the monitoring terminal, a visual recognition model is constructed to perform continuous time target recognition on the spatial image data. The positioning analysis unit performs positioning analysis on the target recognition results of real-time spatial image data on the UAV's operation and inspection route based on the visual recognition model. If the recognition result is no target, the UAV's flight positioning data is continuously recorded and transmitted to the monitoring terminal. If the recognition result is that a target exists, the positioning of both the UAV and the target is recorded simultaneously, and the spatial relative position between the UAV and the target is determined. The spatial relative position is obtained by performing spatial pointization processing on the UAV and the target respectively. The spatial circumscribing spheres of the UAV and the target are taken respectively, and the centers of the spheres are taken as the spatial positioning nodes of the UAV and the target. The two nodes are connected to obtain the spatial relative position data between the UAV and the target. The positioning record of the UAV and the target and the spatial relative position data are transmitted to the monitoring terminal.
5. The real-time monitoring system of an intelligent unmanned aerial vehicle according to claim 4, characterized in that: The linkage decision-making module includes a target trajectory tracking and analysis unit and a linkage decision-making unit; The target trajectory tracking and analysis unit determines the target's real-time moving trajectory based on the target's real-time positioning data and the recorded target positioning data; and predicts and simulates the target's movement trajectory based on the target's real-time moving trajectory. The linkage decision-making unit performs anomaly analysis of the operation scenario based on the real-time motion trajectory prediction simulation, divides the safe operation range of the operation scenario into levels, determines the probability distribution analysis of the impact of the target's real-time motion trajectory prediction simulation generated route on the anomalies of different levels of space in the safe operation range of the operation scenario, and constructs a UAV linkage model based on the analysis results to determine the decision. 6.The real-time monitoring system of an intelligent unmanned aerial vehicle according to claim 5, characterized in that: The specific steps for predicting and simulating the target's trajectory based on its real-time moving trajectory are as follows: By collecting positioning data of the target and recording its movement trajectory, and using the target spatial positioning point recorded at the current real-time point as the starting node of the predicted route, the movement trajectory of the target in the subsequent prediction period is predicted; the prediction period is the time required for the UAV with the shortest spatial relative distance to the target at the current time point to reach the target spatial position node. By obtaining the tangents of corresponding spatial nodes on the target's already moving trajectory, the tangent angles of the corresponding spatial nodes are recorded respectively; the maximum value A of the tangent angles corresponding to each spatial node on the target's already moving trajectory is determined; the tangent angle is the angle between the tangent of the corresponding spatial node and the horizontal line; the extreme points of the target's already moving trajectory are determined, and the vertical distance D between the maximum and minimum points is determined, with D as the fluctuation distance of the already moving trajectory; at the starting node of the predicted route, with [0,A] as the tangent angle variation range and [0,D] as the fluctuation distance range of the predicted motion trajectory, a full traversal is performed to generate a random motion trajectory.
7. The real-time monitoring system for an intelligent unmanned aerial vehicle according to claim 6, characterized in that: Specifically, the hierarchical division of the safe working area in the work scenario involves determining the center of the safe working area within the work scenario, using the center of the area as the sphere's center, and planning the first, second, and third safe distances as the sphere's radii, thus dividing the hemispherical space of the safe working area into the first, second, and third safe layers; wherein the first safe distance is greater than the second safe distance, and the second safe distance is greater than the third safe distance; Based on the hierarchical division of the safe operating range of the work scenario, and using the random motion trajectory generated by the starting node of the predicted route where the real-time target is located, a probability distribution analysis of the abnormal impact of different spatial levels within the safe operating range of the work scenario is conducted. This is achieved by comprehensively considering the total number of random motion trajectory routes generated by the starting node of the predicted route where the real-time target is located, and by comprehensively considering the distribution of each generated random motion trajectory route within different spatial levels of the safe operating range of the work scenario. The proportions of random motion trajectory routes involving the third safety layer, the second safety layer, and the first safety layer at the current time point are calculated respectively. By comparing the corresponding proportion thresholds for each level, the UAV linkage model is determined, and the decision-making processes are identified as conventional level decision, first level decision, second level decision, and third level decision.
8. The real-time monitoring system for an intelligent unmanned aerial vehicle according to claim 7, characterized in that: The data feedback module includes an alert unit and an event recording unit; When the warning unit makes a decision regarding the drone in the operational scenario, it sends a warning notification to the monitoring terminal. The event recording unit logs the drone decision-making events that occur in the operation scenario.