A laser early interception bird repelling method and system based on trajectory prediction

Abstract

The application discloses a kind of laser early interception bird repelling method and system based on trajectory prediction, it is related to bird control technical field.The method first acquires bird video and ranging data, obtains three-dimensional time series observation dataset by solving, iteratively outputs bird posterior motion state vector by Kalman filtering;Generate multiple typical flight candidate trajectories covering straight line, left turn, right turn, dive, climb, dynamically calculate weight based on historical matching degree and weighted fusion, obtain predicted flight trajectory;Construct device protection area and dangerous area, solve optimal front intercept point by multidimensional repelling effect evaluation function;Pan-tilt adopts S-shaped smooth planning and front start aiming, during irradiation process, through the double-layer strategy closed-loop correction of spatial deviation filtering and angle deviation grading.The application effectively eliminates pan-tilt mechanical hysteresis, improves the prediction accuracy of mobile scene, reduces equipment loss and bird adaptability, and is suitable for active bird repelling protection of power transformation station power equipment.

Description

Technical Field

[0001] This invention relates to the field of bird control technology, specifically to a laser-based method and system for early interception and bird deterrence based on trajectory prediction. Background Technology

[0002] As a core node in the power system, the operational safety of substations directly affects the reliability of power supply. Bird activity is one of the major hidden dangers threatening the safe operation of substations. Birds nesting, crossing, and staying around equipment can easily cause faults such as insulator flashover, line short circuits, and equipment contamination, which can lead to large-scale power outages in severe cases.

[0003] The current laser bird deterrence technology used in substations and transmission lines mainly adopts the following solutions: Passive tracking laser bird deterrence: After identifying a bird using a pan-tilt camera, the pan-tilt motor is controlled to rotate, directing the laser beam to illuminate the bird's current location. This type of solution employs a serial "detection-tracking-illumination" working mode, requiring the pan-tilt to follow the bird's movement in real time.

[0004] Fixed-scan laser bird deterrence: The laser scans periodically along a preset trajectory, without targeting any specific object, relying on probability coverage to deter birds.

[0005] Radar-guided bird deterrence: The radar detects the location of birds and guides the gimbal to turn towards the target, but it is still necessary to continuously track the current position of the target.

[0006] Existing passive tracking laser bird deterrence technology has the following technical drawbacks: 1. Traditional solutions employ a "current position tracking" strategy, where the laser beam is always aimed at the bird's current location. Due to the inherent delay in the gimbal's mechanical response (motor start-up, deceleration, and stabilization), by the time the laser beam reaches the target location, the bird has already flown away from its original position, creating a passive situation of "chasing but not catching up."

[0007] 2. Because the laser beam always shines from behind or the side rear, birds easily develop the perception that "flying away will get them away," and gradually become adapted to the laser bird deterrence, with the bird deterrence effect decreasing over time.

[0008] 3. For critical equipment such as substation main transformers and transmission line insulator strings, it is necessary to intercept birds before they enter the danger zone. Traditional solutions cannot intercept birds in advance along their flight path, and often only initiate deflection when the birds have already approached or reached the equipment, resulting in a protection vacuum period.

[0009] 4. To compensate for the lag, some solutions increase the rotation speed of the gimbal, which leads to frequent acceleration and deceleration of the motor, increasing energy consumption and accelerating mechanical wear, thus reducing the equipment's lifespan.

[0010] The above-mentioned problems urgently need to be solved. Summary of the Invention

[0011] The purpose of this invention is to overcome at least one technical problem existing in the prior art and to provide a laser-based method and system for early interception and bird deterrence based on trajectory prediction.

[0012] On one hand, this invention provides a laser-based early interception and bird deterrence method based on trajectory prediction. The method includes: Step S1, continuously acquiring bird flight video sequences using a PTZ camera, extracting the pixel coordinates of birds from N consecutive frames, converting the pixel coordinates into three-dimensional observation coordinates in the camera coordinate system, and generating a time-series observation dataset sorted by time; Step S2, based on the time-series observation dataset, using a Kalman filter algorithm to iteratively estimate the bird's motion state, and outputting a posterior optimal state vector containing position, velocity, and acceleration at the current moment; Step S3, using the posterior optimal state vector as the initial motion condition, generating candidate trajectories corresponding to at least two typical bird flight behaviors; calculating corresponding weight coefficients based on the historical matching degree between historical observation data and each candidate trajectory; weighted fusion of each candidate trajectory to obtain a predicted flight trajectory within a future preset time window; Step S4, constructing a protection zone for the device body and extending the safety boundary outward to obtain a danger zone. The process involves: selecting candidate interception points that meet the pre-constraints of dangerous areas and hardware reachability on the predicted flight trajectory; calculating the deterrence effect score of each candidate interception point using the deterrence effect evaluation function, and selecting the point with the highest score as the optimal laser interception point; calculating the target azimuth and pitch angles of the gimbal based on the three-dimensional coordinates of the optimal laser interception point; calculating the gimbal activation timing based on the expected time of the bird's arrival at the interception point, and controlling the gimbal to rotate to the target angle and complete laser aiming before the bird arrives; updating the motion state and predicted flight trajectory for each frame of bird observation data acquired during laser irradiation, and recalculating the optimal laser interception point; calculating the spatial deviation between the old and new optimal laser interception points, and triggering an interception point update when the spatial deviation exceeds a preset spatial deviation threshold; calculating the gimbal angle deviation based on the updated interception point, and executing corresponding gimbal adjustment strategies according to the magnitude of the angle deviation; monitoring the deterrence status, and terminating laser irradiation and performing corresponding operations when any preset termination condition is met.

[0013] Further, step S1 includes: Step S11, using a target detection algorithm to identify bird targets in a single frame image, outputting the pixel coordinates and detection confidence of the center point of each bird target, and filtering targets with a detection confidence lower than a preset confidence threshold; Step S12, using the ByteTrack tracking algorithm to assign a unique tracking ID to each bird target, completing the association matching of the same target across frames, and generating a time-series sequence of pixel coordinates corresponding to each target; Step S13, based on the pinhole camera projection model, using the camera intrinsic parameter matrix to perform inverse projection calculation on the pixel coordinates of each target, obtaining the three-dimensional direction vector of the corresponding target in the camera coordinate system; Step S14, combining the target straight-line distance obtained by the ranging unit with the three-dimensional direction vector, calculating the three-dimensional observation coordinates of the bird target in the camera coordinate system; Step S15, grouping according to the tracking ID, sorting the three-dimensional observation coordinates of each bird target by timestamp, and generating a time-series observation dataset of the corresponding target.

[0014] Furthermore, step S2 includes: Step S21: Define the motion state vector ,in The true three-dimensional observation coordinates of the target in the camera coordinate system, The instantaneous three-axis velocity of the target in the camera coordinate system, The instantaneous acceleration of the target along three axes in the camera coordinate system; Step S22: Construct the state transition matrix F, observation matrix H, process noise covariance matrix Q, and observation noise covariance matrix R corresponding to the uniformly accelerated motion model; Step S23: Based on the posterior state estimate of the previous time step, calculate the prior state estimate and prior covariance matrix of the current time step, and execute the prediction step of Kalman filtering. Step S24: Introduce the current frame observation vector from the time-series observation dataset, calculate the Kalman gain, correct the prior state, obtain the posterior optimal motion state vector and posterior covariance matrix at the current time, and execute the Kalman filter update step. Step S25: During the cold start of filtering, the initial state vector and the initial covariance matrix are calculated using the observation data of the first three frames through difference operations.

[0015] Furthermore, the candidate trajectories generated in step S3, which correspond to at least two typical bird flight behaviors, include: straight trajectory, left-turn trajectory, right-turn trajectory, dive trajectory, and / or climb trajectory. The mathematical expression for the straight line trajectory is: ; The mathematical expression for the left-turn trajectory is: ; The mathematical expression for the right-turn trajectory is: ; The mathematical expression for the dive trajectory is: ; The mathematical expression for the climbing trajectory is: ; ; In the formula, Let be the bird's initial three-dimensional position vector at the current moment. The initial three-dimensional velocity vector, The velocity component of the bird at the current moment. The turning radius is The net velocity of the birds at the current moment. For the maximum steering angular rate, The heading angle of the bird at the current moment. For equivalent gravitational acceleration, This represents the bird's current altitude.

[0016] Furthermore, in step S3, the corresponding weight coefficients are calculated based on the historical matching degree between historical observation data and the historical data of each candidate trajectory; the candidate trajectories are weighted and fused to obtain the predicted flight trajectory within a future preset time window, including: Calculate the matching degree between each candidate trajectory and historical observation trajectories: ; In the formula, Let be the actual observed position vector at the k-th historical moment. Let be the predicted position vector of the i-th candidate trajectory at the k-th historical moment. For historical frames, The historical matching score for the i-th candidate trajectory. For error sensitivity coefficient, It is a natural exponential function; Calculate the weights corresponding to each candidate trajectory model based on the matching degree: ; In the formula, Let be the weight coefficient of the i-th candidate trajectory; The weighted fusion of the candidate trajectory models yields the predicted flight trajectory within a future preset time window, including: ; In the formula, Let be the i-th candidate trajectory.

[0017] Furthermore, step S4 includes: Step S41: Define the electrical equipment to be protected as a cylinder, cuboid, or combined geometric model, and extend it outward by a preset safety boundary distance to form a danger zone; Step S42: Calculate the time of intersection between the predicted flight trajectory and the boundary of the danger zone, set the total lead time, and take the trajectory point at the time corresponding to the difference between the total lead time and the time of intersection as the initial interception point. Step S43: Verify whether the initial interception point is within the effective range of the laser and the rotation angle of the gimbal; Step S44: In response to the initial interception point not being within the effective range of the laser and the rotation angle of the gimbal, search backward along the trajectory for the first reachable point as the optimal laser interception point; Step S45: In response to the initial interception point being within the effective range of the laser and the rotation angle of the gimbal, sample candidate interception points with a preset step size within a preset time range before and after the initial interception point, substitute the candidate interception points into the expulsion effect evaluation function to calculate the score, and select the point with the highest score as the optimal laser interception point.

[0018] Furthermore, the mathematical expression for the repulsion effect evaluation function is: ; ; In the formula, The overall score for the deterrence effect of candidate interception point I. , , These are the weighting coefficients. d is the straight-line propagation distance from the laser emitter to the candidate interception point. This represents the maximum effective range of the laser. The standard deviation of the Gaussian kernel for distance decay. The angle between the laser incident direction and the bird's flight path. The remaining time for birds to reach the danger zone. This is the minimum safe time threshold.

[0019] Furthermore, step S5 includes: Step S51: Using the gimbal base as the spatial origin, calculate the gimbal target azimuth angle and target elevation angle based on the three-dimensional coordinates of the optimal laser interception point; Step S52: Compare the gimbal target azimuth angle and target pitch angle with the gimbal physical limit range, and perform amplitude limiting and clipping processing to obtain the gimbal target angle; Step S53: Use a seven-segment S-shaped acceleration planning method to generate the gimbal motion speed curve to ensure no mechanical impact during the motion process; Step S54: Calculate the gimbal activation time based on the time the bird arrives at the interception point and the total duration of the gimbal movement, and control the gimbal to complete its positioning before the bird arrives at the interception point.

[0020] Furthermore, step S6 includes: Step S61: During the laser irradiation process, update the observation data for each frame, re-execute the Kalman filter operation and trajectory prediction, and calculate the new optimal laser interception point; Step S62: Calculate the spatial deviation between the old and new interception points. When the spatial deviation exceeds a preset spatial deviation threshold, trigger the interception point update. Step S63: When the interception point is updated, repeat step S5 to calculate the new gimbal target angle based on the updated optimal laser interception point. Step S64: Based on the new gimbal target angle and the gimbal target angle before the update, calculate the gimbal azimuth axis angle deviation and pitch axis angle deviation respectively, and compare the gimbal azimuth axis angle deviation and pitch axis angle deviation with the preset angle deviation threshold respectively. Step S65: When the angle deviation of any axis is less than the preset angle deviation threshold, the axis is directly fine-tuned in speed mode; when the angle deviation of any axis is greater than or equal to the preset angle deviation threshold, the axis is re-planned with a seven-segment S-shaped smooth motion trajectory, and the two axes execute their respective motion control commands synchronously. Step S66: The termination conditions include: successful deflection condition, target flight condition, interception failure condition, timeout protection condition, and / or trajectory deviation condition. When any termination condition is met, the laser irradiation is terminated and the corresponding action is executed.

[0021] Secondly, embodiments of the present invention provide a substation laser pre-interception bird deterrence system based on multi-hypothesis trajectory prediction. The system employs the aforementioned laser pre-interception bird deterrence method based on trajectory prediction. The system includes: a time-series observation dataset generation module, suitable for continuously acquiring bird flight video sequences via a PTZ camera, extracting the pixel coordinates of birds from N consecutive frames, converting the pixel coordinates into three-dimensional observation coordinates in the camera coordinate system, and generating a time-series observation dataset ordered by time; a Kalman filtering module, suitable for iteratively estimating the bird's motion state using a Kalman filtering algorithm based on the time-series observation dataset, outputting a posterior optimal state vector containing position, velocity, and acceleration; a predicted flight trajectory generation module, suitable for generating candidate trajectories corresponding to at least two typical bird flight behaviors using the posterior optimal state vector as the initial motion condition; calculating corresponding weight coefficients based on the historical matching degree between historical observation data and each candidate trajectory; weighted fusion of each candidate trajectory to obtain a predicted flight trajectory within a future preset time window; and an optimal laser interception point generation module, suitable for... The system is used to construct the protection zone of the equipment body and extend the safety boundary outward to obtain the danger zone; to screen candidate interception points that meet the pre-constraints of the danger zone and the hardware reachability constraints on the predicted flight trajectory; to calculate the expulsion effect score of each candidate interception point through the expulsion effect evaluation function, and to select the point with the highest score as the optimal laser interception point; the laser aiming module is suitable for calculating the target azimuth and pitch angles of the gimbal based on the three-dimensional coordinates of the optimal laser interception point; to calculate the gimbal activation timing based on the expected time when the bird arrives at the interception point, and to control the gimbal to rotate to the target angle and complete the laser aiming before the bird arrives; the dynamic correction and closed-loop control module is suitable for updating the motion state and predicted flight trajectory for each frame of bird observation data acquired during the laser irradiation process, and recalculating the optimal laser interception point; to calculate the spatial deviation between the old and new optimal laser interception points, and to trigger the interception point update when the spatial deviation exceeds the preset spatial deviation threshold; to calculate the gimbal angle deviation based on the updated interception point, and to execute the corresponding gimbal adjustment strategy according to the graded angle deviation amplitude; to monitor the expulsion status, and to terminate the laser irradiation and execute the corresponding operation when any preset termination condition is met.

[0022] Thirdly, embodiments of the present invention also provide an electronic device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program, when executed by the processor, implements the above-described laser-based early interception and bird deterrence method based on trajectory prediction.

[0023] Fourthly, embodiments of the present invention also provide a readable storage medium, which, when the instructions in the storage medium are executed by the processor of an electronic device, enables the electronic device to execute the above-described laser-based early interception and bird-repelling method based on trajectory prediction.

[0024] The advantages of this invention compared to the prior art are as follows: (1) A multi-hypothesis trajectory weighted fusion mechanism is adopted to cover five types of typical bird flight behaviors: straight line, left turn, right turn, dive, and climb. The weight of each trajectory is dynamically and adaptively adjusted through historical matching degree. Compared with a single motion model, the prediction accuracy in maneuvering scenarios such as turning and ascending is significantly improved, and the prediction error caused by bird active maneuvering is reduced, providing a reliable trajectory basis for forward interception.

[0025] (2) By constructing equipment protection zones and dangerous zones, the interception point is solved in advance outside the dangerous zone, and the control PTZ is positioned and aimed in advance to realize the "birds arrive and are irradiated" forward interception mode, which moves the driving action to outside the dangerous zone, transforms passive defense into active intervention, eliminates the protection vacuum period, and improves the protection reliability of substation equipment.

[0026] (3) A two-layer control strategy of “spatial deviation filtering + angle deviation classification” is adopted: the first layer filters out small trajectory fluctuations through spatial deviation threshold to avoid invalid micro-movements of the gimbal; the second layer matches the corresponding control mode according to the angle deviation amplitude, with small deviations being quickly fine-tuned and large deviations being smoothed, which not only ensures the response speed but also greatly reduces the number of invalid gimbal movements. Combined with the seven-segment S-shaped acceleration and deceleration planning to eliminate mechanical shock, the service life of the gimbal is effectively extended.

[0027] (4) The entire process adopts a lightweight kinematic model and regularized calculation, the amount of calculation is controllable, and it can run in real time on embedded edge devices; the protection area can be flexibly adapted to different substation equipment forms such as transformers, switch cabinets, and towers. Attached Figure Description

[0028] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0029] Figure 1 This is a flowchart of a laser-based early interception and bird deterrence method based on trajectory prediction, provided in Embodiment 1 of this application.

[0030] Figure 2 This is a flowchart of generating a time-series observation dataset corresponding to a target, provided in Embodiment 1 of this application.

[0031] Figure 3 This is a Kalman filter flowchart provided in Embodiment 1 of this application.

[0032] Figure 4 This is a flowchart of generating the optimal laser interception point provided in Embodiment 1 of this application.

[0033] Figure 5 This is a gimbal control flowchart provided in Embodiment 1 of this application.

[0034] Figure 6This is a flowchart of dynamic correction and closed-loop control provided in Embodiment 1 of this application.

[0035] Figure 7 This is a schematic diagram of a laser-based early interception and bird deterrence system based on trajectory prediction, provided in Embodiment 2 of this application.

[0036] Figure 8 This is a partial block diagram of the electronic device provided in Embodiment 3 of the present invention. Detailed Implementation

[0037] 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, not all, of the embodiments of the present invention. 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.

[0038] Example 1

[0039] like Figure 1 The flowchart shown is a laser-based early interception and bird deterrence method based on trajectory prediction provided in this application embodiment.

[0040] As an example, the method includes: Step S1, continuously acquiring bird flight video sequences using a PTZ camera, extracting the pixel coordinates of birds from N consecutive frames, converting the pixel coordinates into three-dimensional observation coordinates in the camera coordinate system, and generating a time-series observation dataset sorted by time; Step S2, based on the time-series observation dataset, using a Kalman filter algorithm to iteratively estimate the bird's motion state, outputting a posterior optimal state vector containing position, velocity, and acceleration at the current moment; Step S3, using the posterior optimal state vector as the initial motion condition, generating candidate trajectories corresponding to at least two typical bird flight behaviors; calculating corresponding weight coefficients based on the historical matching degree between historical observation data and each candidate trajectory; weighted fusion of each candidate trajectory to obtain a predicted flight trajectory within a future preset time window; Step S4, constructing a protective area for the device body and extending the safety boundary outward to obtain a danger zone; filtering the predicted flight trajectory to obtain a safety zone; and selecting a suitable area from the predicted flight trajectory. Candidate interception points are selected based on pre-constraints and hardware reachability constraints in the hazardous area; the deterrence effect score of each candidate interception point is calculated using a deterrence effect evaluation function, and the point with the highest score is selected as the optimal laser interception point; Step S5: The azimuth and pitch angles of the gimbal are calculated based on the three-dimensional coordinates of the optimal laser interception point; the gimbal activation timing is calculated based on the expected time when the bird arrives at the interception point, and the gimbal is controlled to rotate to the target angle and complete laser aiming before the bird arrives; Step S6: During laser irradiation, the motion state and predicted flight trajectory are updated for each frame of bird observation data acquired, and the optimal laser interception point is recalculated; the spatial deviation between the old and new optimal laser interception points is calculated, and the interception point is updated when the spatial deviation exceeds a preset spatial deviation threshold; the gimbal angle deviation is calculated based on the updated interception point, and the corresponding gimbal adjustment strategy is executed according to the graded angle deviation amplitude; the deterrence status is monitored, and laser irradiation is terminated and the corresponding operation is executed when any preset termination condition is met.

[0041] In some feasible implementations, combined with Figure 2 As shown, step S1 includes: Step S11, using a target detection algorithm to identify bird targets in a single frame image, outputting the pixel coordinates and detection confidence of the center point of each bird target, and filtering targets with a detection confidence lower than a preset confidence threshold; Step S12, using the ByteTrack tracking algorithm to assign a unique tracking ID to each bird target, completing the association matching of the same target across frames, and generating a time-series sequence of pixel coordinates corresponding to each target; Step S13, based on the pinhole camera projection model, using the camera intrinsic parameter matrix to perform inverse projection calculation on the pixel coordinates of each target, obtaining the three-dimensional direction vector of the corresponding target in the camera coordinate system; Step S14, combining the target straight-line distance obtained by the ranging unit with the three-dimensional direction vector, calculating the three-dimensional observation coordinates of the bird target in the camera coordinate system; Step S15, grouping according to the tracking ID, sorting the three-dimensional observation coordinates of each bird target by timestamp, and generating a time-series observation dataset of the corresponding target.

[0042] Preferably, step S1 includes: performing bird detection on each frame of the image using the YOLOv8 object detection algorithm, and obtaining the coordinates of the bird's center point in the image coordinate system. Set the detection confidence threshold. Detection boxes below this threshold are discarded; the ByteTrack tracking algorithm is used to associate the IDs of the same bird in consecutive frames to ensure trajectory continuity.

[0043] Preferably, the coordinate system transformation process is as follows: pixel coordinates Convert to 3D direction vector in camera coordinate system ; Using camera intrinsic matrix Perform coordinate transformation: ; in , Focal length The coordinates of the main point.

[0044] Distance estimation: By combining radar ranging data or depth camera data, the target distance is obtained. Then the three-dimensional coordinates of the target in the camera coordinate system are: ; Obtain the true three-dimensional spatial coordinates of the target in the camera coordinate system ; The data is grouped according to tracking ID, and all 3D observation coordinates corresponding to each ID are sorted in ascending order of timestamp, generating independent time-series observation datasets for each target. Each data entry in the dataset contains two core pieces of information: a timestamp and 3D observation coordinates, which serve as input to the subsequent Kalman filter module. The length of the time-series dataset adopts a sliding window mechanism, always retaining the observation data from the most recent 30 frames, ensuring sufficient historical data support for filtering while controlling memory usage and computational load.

[0045] In some feasible implementations, combined with Figure 3 As shown, step S2 includes: step S21, defining the motion state vector. ,in The true three-dimensional observation coordinates of the target in the camera coordinate system, The instantaneous three-axis velocity of the target in the camera coordinate system, The target's three-axis instantaneous acceleration in the camera coordinate system is defined in step S22. The state transition matrix F, observation matrix H, process noise covariance matrix Q, and observation noise covariance matrix R corresponding to the uniformly accelerated motion model are constructed. Step S23. Based on the posterior state estimate from the previous moment, the prior state estimate and prior covariance matrix for the current moment are calculated, and the Kalman filter prediction step is executed. Step S24. The current frame observation vector from the time-series observation dataset is introduced, the Kalman gain is calculated, the prior state is corrected, and the posterior optimal motion state vector and posterior covariance matrix for the current moment are obtained. The Kalman filter update step is executed. Step S25. During the cold start of the filter, the initial state vector and initial covariance matrix are calculated using the first three frames of observation data through difference operations.

[0046] Preferably, step S2 includes: using a three-dimensional uniformly accelerated motion model, defining a 9-dimensional state vector to fully describe the bird's instantaneous motion state: ; Using a uniformly accelerated motion model, the state transition matrix is: ; in It is a 3×3 identity matrix. This is the sampling time interval (typical value 33ms, corresponding to 30fps video). The observation matrix is: ; Process noise covariance matrix Reflecting the uncertainty of the system model: ; in , The standard deviation of acceleration disturbance is taken as a typical value. ; Observation noise covariance matrix Reflecting measurement error: ; in The standard deviation of the position measurement is taken as a typical value. (Depending on the ranging accuracy); The Kalman filter iterative process includes: Prediction steps: ; ; This is the prior state estimate at time k. For the posterior state estimation at time k-1, Let be the prior covariance matrix at time k. Let be the posterior covariance matrix at time k-1; Update steps: ; ; ; The Kalman gain at time k is... Posterior state estimation at time k. The posterior covariance matrix at time k. Output This is the bird's current motion state vector obtained by optimal estimation through Kalman filtering at time k (current frame), which is a complete 9-dimensional motion state including position, velocity, and acceleration.

[0047] Preferably, when a new bird target is detected for the first time and the filter has not yet been established, the first three frames of observation data are used for cold start initialization: initial value of the state vector. The results were obtained by differential calculation from the first three frames of observation data: Initial position: The observation position of the first frame is directly taken as the initial position component of the state vector; Initial velocity: Divide the position difference between the first two frames by the time interval to obtain the average velocity during this period, which is used as the initial velocity estimate. Initial acceleration: ; Initial covariance matrix This reflects the uncertainty of the initial estimate.

[0048] This is the three-dimensional observation coordinate vector for frame 0 (the first frame in which the target is first detected); This is the three-dimensional observation coordinate vector for the first frame (the second frame after the target is detected); This is the three-dimensional observation coordinate vector for the second frame (the third frame in which the target was detected); This represents the time interval between two adjacent frames.

[0049] In some feasible implementations, the candidate trajectories generated in step S3 that correspond to at least two typical bird flight behaviors include: straight trajectory, left-turn trajectory, right-turn trajectory, dive trajectory and / or climb trajectory; The mathematical expression for the straight line trajectory is: ; The mathematical expression for the left-turn trajectory is: ; The mathematical expression for the right-turn trajectory is: ; The mathematical expression for the dive trajectory is: ; The mathematical expression for the climbing trajectory is: ; ; In the formula, Let be the bird's initial three-dimensional position vector at the current moment. The initial three-dimensional velocity vector, The velocity component of the bird at the current moment. The turning radius is The net velocity of the birds at the current moment. Maximum steering angular rate (typical value) ), The heading angle of the bird at the current moment. Equivalent gravitational acceleration (typical value) ), This represents the bird's current altitude.

[0050] In some feasible implementations, step S3 involves calculating corresponding weight coefficients based on the historical matching degree between historical observation data and each candidate trajectory; weighted fusion of each candidate trajectory yields the predicted flight trajectory within a future preset time window, including: Calculate the matching degree between each candidate trajectory and historical observation trajectories: ; In the formula, Let be the actual observed position vector at the k-th historical moment. Let be the predicted position vector of the i-th candidate trajectory at the k-th historical moment. For historical frames, The historical matching score for the i-th candidate trajectory. This is the error sensitivity coefficient (typically 1m). It is a natural exponential function; Calculate the weights corresponding to each candidate trajectory model based on the matching degree: ; In the formula, Let be the weight coefficient of the i-th candidate trajectory; The weighted fusion of the candidate trajectory models yields the predicted flight trajectory within a future preset time window, including: ; In the formula, Let be the i-th candidate trajectory.

[0051] Preferably, considering the actual physiological characteristics of birds during flight, all candidate trajectories satisfy the following boundary constraints: Speed ​​constraint: Bird flight speed ,in , ; Turning angle constraint: Bird heading change rate This reflects the limitations of birds' turning ability; Altitude constraint: Rate of change of bird flight altitude .

[0052] Preferably, the preset time window is dynamically adjusted according to the application scenario: general bird deterrence scenario: preset time window = 1.0~2.0s; critical equipment protection scenario: preset time window = 2.0~3.0s.

[0053] In some feasible implementations, combined with Figure 4 As shown, step S4 includes: Step S41, defining the power equipment to be protected as a cylinder, cuboid, or combined geometric model, and extending it outward by a preset safety boundary distance to form a danger zone; Step S42, calculating the intersection time of the predicted flight trajectory and the danger zone boundary, setting a total lead time, and taking the trajectory point corresponding to the difference between the total lead time and the intersection time as the initial interception point; Step S43, verifying whether the initial interception point is within the effective range of the laser and the gimbal rotation angle; Step S44, in response to the initial interception point not being within the effective range of the laser and the gimbal rotation angle, searching backward along the trajectory for the first reachable point as the optimal laser interception point; Step S45, in response to the initial interception point being within the effective range of the laser and the gimbal rotation angle, sampling candidate interception points within a preset time range before and after the initial interception point with a preset step size, substituting the candidate interception points into the drive-off effect evaluation function to calculate the score, and selecting the point with the highest score as the optimal laser interception point.

[0054] Preferably, step S41 includes: firstly, modeling the protection area of ​​the equipment body: abstracting the substation power equipment to be protected into a geometric model, supporting the following three types of basic models and combined models: The cylindrical model (applicable to main transformers, reactors, etc.) has the following mathematical expression: ; In the formula, Let r be the coordinates of the center of the base of the cylinder, r be the radius of the base, and h be the height of the equipment.

[0055] The rectangular prism model (suitable for control rooms, switch cabinets, etc.) has the following mathematical expression: ; Combinatorial model (complex equipment): The union of multiple basic geometries, suitable for complex equipment groups.

[0056] The equipment outline constitutes the protective area of ​​the equipment body. Based on the protected area of ​​the equipment itself, a preset safety boundary distance is extended outward at equal intervals. This creates a dangerous area. : ; In this embodiment The value is 5m, and the expansion method is three-dimensional Minkowski expansion, meaning that all distances from the protected area in space are less than or equal to... The set of points. The physical meaning of a danger zone is: birds entering this zone are considered a threat to the equipment and need to be driven away immediately; the goal of pre-emptive interception is to drive away birds before they reach the danger zone. Let P be the Euclidean distance from point P to point Q, where Q is any point on the equipment protection area and P is any point in three-dimensional space.

[0057] Preferably, step S42 includes: calculating the trajectory P(t) and the danger zone. The intersection point is used as the intrusion time. If the trajectory does not intersect with the danger zone, no interception is needed, and an empty value is returned. If the trajectory has entered the danger zone, the current time is taken as the threshold. Set lead time , The gimbal response time (measured at 0.2-0.4s, taken as 0.3s) is... The laser onset time (bird reaction time 0.3-0.8s, taken as 0.5s) is used. For a safety margin (0.2-0.5s, take 0.3s), the total is... Initial interception point .

[0058] Preferably, step S43 includes: verifying whether the initial interception point satisfies two hardware constraints: Range constraint: Calculate the straight-line distance from the initial interception point to the laser emitter and determine if it is less than or equal to the laser's maximum effective range. (In this embodiment, it is 100m); Angle constraints: Calculate the corresponding azimuth and pitch angles of the gimbal based on the initial interception point coordinates, and determine whether it is within the physical limits of the gimbal (azimuth angle). Pitch angle If both constraints are satisfied, the device is determined to be hardware reachable; if either constraint is not satisfied, the device is determined to be hardware unreachable.

[0059] Preferably, step S44 includes: if the initial interception point does not meet the hardware reachability constraint, then along the predicted trajectory towards the device (the direction of increasing time), traverse the trajectory points sequentially with a step size of 0.01s, and take the first trajectory point that simultaneously meets both the range constraint and the angle constraint as the optimal laser interception point. This mechanism ensures that an reachable interception position can be obtained even in extreme scenarios, avoiding algorithm failure.

[0060] Preferably, step S45 includes: if the initial interception point meets the hardware constraints, then within a preset time range (0.3s) before and after the initial interception point, several candidate interception points are sampled with a step size of 0.05s; each candidate point is substituted into the deflection effect evaluation function to calculate the comprehensive score, and the point with the highest score is selected as the optimal laser interception point.

[0061] The deterrence effectiveness evaluation function comprehensively evaluates the deterrence effect from three dimensions: distance, angle, and time. Its mathematical expression is: ; ; In the formula, The overall score for the deterrence effect of candidate interception point I. , , These are the weighting coefficients. d is the straight-line propagation distance from the laser emitter to the candidate interception point. This represents the maximum effective range of the laser. The standard deviation of the Gaussian kernel for distance decay. The angle between the laser incident direction and the bird's flight path. The remaining time for birds to reach the danger zone. This is the minimum safe time threshold.

[0062] Preferably, when multiple bird targets are present in the image simultaneously, the targets are selected for interception according to the following priority order: Priority 1: Birds that have already entered the danger zone, with the highest threat level; Priority 2: Birds closest to the boundary of the danger zone, with the highest probability of imminent intrusion; Priority 3: Birds with the fastest flight speed, most likely to quickly enter the danger zone. The highest priority targets are intercepted first, and subsequent targets are processed in sequence after they are driven away, ensuring maximum protection effectiveness with limited hardware resources.

[0063] In some feasible implementations, combined with Figure 5As shown, step S5 includes: Step S51, using the gimbal base as the spatial origin, calculating the gimbal target azimuth angle and target pitch angle based on the three-dimensional coordinates of the optimal laser interception point; Step S52, comparing the gimbal target azimuth angle and target pitch angle with the physical limit range of the gimbal, performing amplitude limiting and clipping processing to obtain the gimbal target angle; Step S53, using a seven-segment S-shaped acceleration planning to generate the gimbal movement speed curve to ensure no mechanical impact during the movement process; Step S54, calculating the gimbal start-up time based on the time when the bird arrives at the interception point and the total movement time of the gimbal, controlling the gimbal to complete its positioning before the bird arrives at the interception point.

[0064] Preferably, step S51 includes: taking the gimbal rotation center as the spatial coordinate origin, and based on the three-dimensional coordinates of the optimal laser interception point... Calculate the azimuth angle that the gimbal needs to rotate. and pitch angle : ; .

[0065] Preferably, step S52 includes: comparing the calculated target azimuth angle and target pitch angle with the physical limit range of the gimbal, respectively. ; ; If the target angle is within the limit range, it is retained directly; if it exceeds the limit range, it is cropped to the corresponding boundary value to ensure that the final output target angle is within the reachable range of the gimbal and to avoid overtravel commands. If the angle is limited, it is marked as "partially reachable" and can be compensated for later by adjusting the laser power or expanding the scanning range.

[0066] Preferably, step S53 includes: to avoid impact from gimbal movement, an S-shaped velocity planning curve (seven-segment acceleration planning) is adopted: Motion parameters: Maximum speed: (azimuth axis) (Pitch axis); Maximum acceleration: ; acceleration: ; Seven segments of the S-curve: Acceleration phase ( ): jerk is The acceleration increases linearly; Uniform acceleration segment ( ): The acceleration is constant ; deceleration phase ( ): jerk is The acceleration decreases linearly to 0; Uniform speed segment ( ): The speed is constant ; Acceleration and deceleration phases ( ): jerk is The acceleration decreases linearly; Uniform deceleration section ( ): The acceleration is constant ; Deceleration phase ( ): jerk is The acceleration increases linearly to 0.

[0067] Preferably, step S54 includes: total motion time Due to angle difference Determined by motion parameters: ; When rotating at a small angle, the maximum speed cannot be reached, so a triangular velocity curve is used; when rotating at a large angle, there is a uniform speed phase, so a trapezoidal velocity curve is used.

[0068] Preset trigger timing: Let the current time be The predicted arrival time at the interception point is The gimbal startup time is: ; in This is a buffer time to handle prediction errors. The estimated time when the birds will arrive at the interception point.

[0069] Triggering condition judgment: Only when Only when the time is right will the preset control of the PTZ be activated; if If the predicted time has passed, then move directly to the interception point at maximum speed. This refers to the current moment. When the system reaches... At any given moment, a motion control command is issued to control the pan-tilt unit to rotate according to an S-shaped speed curve, ensuring that... The timing is just right to reach the target angle and complete the aiming, achieving the effect of "shooting the bird as soon as it arrives" in advance, neither wasting time too early nor missing the best interception opportunity too late.

[0070] In some feasible implementations, combined with Figure 6As shown, step S6 includes: Step S61: During laser illumination, for each frame of updated observation data, Kalman filtering and trajectory prediction are re-executed to calculate a new optimal laser interception point; Step S62: The spatial deviation between the old and new interception points is calculated, and an interception point update is triggered when the spatial deviation exceeds a preset spatial deviation threshold; Step S63: When the interception point is updated, step S5 is repeated to calculate a new gimbal target angle based on the updated optimal laser interception point; Step S64: Based on the new gimbal target angle and the previous gimbal target angle, the azimuth axis angle deviation and pitch axis angle deviation of the gimbal are calculated respectively. Step S65: When the angle deviation of the gimbal's azimuth axis and pitch axis is less than the preset angle deviation threshold, the axis is directly fine-tuned using speed mode; when the angle deviation of any axis is greater than or equal to the preset angle deviation threshold, a seven-segment S-shaped smooth motion trajectory is replanned for the axis, and the two axes execute their respective motion control commands synchronously; Step S66: The termination conditions include: successful drive-away condition, target flight-out condition, interception failure condition, timeout protection condition, and / or trajectory deviation condition. When any termination condition is met, laser irradiation is terminated and the corresponding action is executed.

[0071] Preferably, step S6 achieves closed-loop dynamic tracking correction during laser irradiation, controls gimbal adjustment through a dual-layer deviation grading strategy, reduces mechanical wear while ensuring tracking accuracy, and monitors the drive-off status in real time, ending the drive-off when the termination conditions are met.

[0072] Preferably, step S61 includes: during the laser irradiation process, with video frames as the period, for each new frame of observation data input, that is, repeatedly performing S2 Kalman filter state update, S3 trajectory recalculation, and S4 interception point refresh to obtain a new optimal laser interception point. This forms a closed-loop tracking circuit, which adapts to changes in the bird's flight status in real time.

[0073] Preferably, step S62 includes: calculating the three-dimensional spatial deviation between the new interception point and the current irradiation point, i.e., the Euclidean distance between the two points: ; Set spatial deviation threshold In this embodiment, 0.5m is used.

[0074] like If the bird's position is determined to have changed substantially, the interception point will be updated, and the next level of angle deviation determination will begin. like If the fluctuation is determined to be minor, the current illumination position will remain unchanged, and the gimbal adjustment will not be triggered. The filter trajectory will experience slight jitter to avoid frequent micro-movements of the gimbal.

[0075] Preferably, step S63 includes: recalculating the corresponding gimbal target angle for the interception point that triggers the update, and calculating the angle deviations of the azimuth axis and pitch axis respectively. An angle deviation threshold is set; in this embodiment, it is set to 5°. Dual-axis independent judgment and independent selection of control modes are implemented. Small Deviation Fine-Tuning Mode: When the angular deviation of a single axis is less than 5°, the speed mode is used to directly fine-tune the axis. This mode has a fast response speed, short motion stroke, and low mechanical wear, making it suitable for small-range corrections.

[0076] Large Deviation Smoothing Mode: When the angular deviation of a single axis is greater than or equal to 5°, a seven-segment S-shaped smooth motion curve is redesigned for that axis to control the gimbal's smooth rotation. This mode has no mechanical shock and is suitable for large-angle rotation scenarios. The azimuth and pitch axes execute their respective control commands synchronously, balancing tracking response speed and motion stability.

[0077] Preferably, step S64 includes: real-time monitoring of the decoy status, and immediately terminating laser irradiation and resetting the gimbal to the standby position when any of the following termination conditions are met: Birds successfully driven away: If a bird's course changes by ≥45° and its distance from the danger zone continues to increase, the bird is deemed to have been driven away and the event is recorded as successful. Target flies out: The bird is beyond the effective range of the laser or flies out of the camera's field of view. The target is not detected for 5 consecutive frames and cannot be continuously tracked. Interception Failed: Birds break through the boundary of the danger zone and enter the protection zone of the equipment itself. The pre-interception fails, triggering a level 2 alarm and notifying maintenance personnel. Timeout protection: If the birds are not driven away after the laser has been continuously irradiated for more than the preset maximum time (10 seconds in this embodiment), the laser will be forcibly shut off to avoid damage to the birds from prolonged irradiation and to prevent the laser from overheating. Trajectory deviation: If the actual trajectory deviates from the predicted residual by more than 3 times the observation noise for 8 consecutive frames, it is determined to be a tracking failure, the illumination is terminated and the tracking is reinitialized.

[0078] Preferably, a tiered exception handling logic is also set up to deal with various abnormal situations during operation: Target loss: If no target is detected for 3 consecutive frames, illumination is paused and the gimbal search range is expanded; if the target is recaptured within 5 frames, illumination continues; otherwise, the current drive-away operation is terminated. Multi-target obfuscation: Real-time detection of tracking ID switching; if ID obfuscation or target identity mismatch is detected, tracking and filtering are re-initialized to avoid data crosstalk. System Fault: When hardware faults such as gimbal jamming, laser malfunction, or abnormal ranging are detected, laser irradiation will be immediately terminated and the fault will be reported to ensure equipment safety.

[0079] In some feasible implementations, specific application examples are provided here to facilitate understanding of the above embodiments: Example 1: Protection scenario for main transformer in substation Scene description: A bird (crow, body length 45cm) flies from 50m away towards the main transformer of a 220kV substation (size 8m×5m×6m).

[0080] Traditional solution: Detection delay: 100ms; Gimbal response: 300ms; Bird position at laser arrival time: already flown (0.1+0.3+0.5)×8=7.2m. Bird deterrence effect: With a laser delay of more than 7m, the crow has already entered within 5m of the transformer, resulting in a bird deterrence failure rate of approximately 40%.

[0081] Invention solution: Trajectory acquisition (t=0s~0.5s): Acquire 15 frames of images (30fps) to obtain the initial trajectory point sequence; Initial position: (45m, 10m, 15m); Initial velocity estimate: (-7.5m / s, -1.2m / s, -0.8m / s).

[0082] Motion state estimation: Kalman filter initialization: ; Covariance matrix: Process noise: Observation noise: .

[0083] Multi-hypothesis trajectory prediction (t=0.5s): Weights of the straight-line hypothesis: (Historical trajectory approximates a straight line); Weighting of left turn assumption: Right turn assumption weight: Dive Assumption Weights: Climbing Assumption Weights: ; Fusion predicted trajectory: will arrive at (22.5m, 6.4m, 12.6m) in 3 seconds.

[0084] Interception point calculation: Danger zone boundary: 5m beyond the transformer boundary, i.e., no more than 9m from the center point (0,0,3); Intersection of the trajectory and the danger zone: Location (24m, 6.6m, 12.8m); Lead time: (Response time 0.3s + onset time 0.5s + margin 0.3s); Initial interception point: position at t=1.7s (32.3m, 8.0m, 13.6m); Reachability check: distance 36.5m < 100m, angle (13.8°, 20.5°) within the limit range; Local optimization: search range [1.4s, 2.0s], optimal interception point (30.5m, 7.7m, 13.4m), drive-away effect score 0.82.

[0085] Gimbal preset control: Target angle: Azimuth 13.2°, Pitch 21.0°; Current angle: Azimuth 0°, Pitch 0°; Angle difference: 13.2° (Azimuth), 21.0° (Pitch); S-curve planning: Total motion time 0.45s; Start time: (Currently 0.5s, please wait 0.65s).

[0086] Laser Illumination and Dynamic Correction: t=1.15s: Start gimbal movement; t=1.6s: Gimbal reaches preset position, laser turns on; t=1.7s: Crow reaches interception point, laser illuminates from the front; t=1.9s: Crow changes course by 50°, flies away from transformer direction; t=2.0s: Successful deflection confirmed, laser turns off.

[0087] Results Comparison: This solution successfully intercepted and drove away birds at a distance of 24m from the transformer, with a success rate of 95% and a gimbal movement count of 1 (compared to approximately 5 tracking movements for the traditional solution).

[0088] Example 2: Multi-objective prioritization decision-making scenario

[0089] Scenario description: Three sparrows fly towards critical equipment in different directions at the same time. System resources are limited, and it is necessary to determine the interception priority.

[0090] The target parameters are shown in Table 1: Table 1:

[0091] Prioritization Decisions: Sparrow B is expected to arrive at the danger zone first (1.8s) and is currently the closest (5m), so it has the highest priority; Sparrow A is relatively close (8m), so it has the second highest priority; Sparrow C is farther away and takes longer to arrive, so it has the lowest priority.

[0092] Execution strategy: First, intercept sparrow B, calculate the interception point and execute the illumination; if sparrow B is driven away within 1.5 seconds, immediately switch to sparrow A; if the timeout for driving away sparrow B is exceeded, activate the second-level alarm and give up, focusing all efforts on intercepting sparrow A.

[0093] Results: Sparrow B: successfully driven away in 1.2s; Sparrow A: successfully driven away in 2.8s; Sparrow C: changed direction and flew away on its own without intervention.

[0094] The above implementation method effectively adapts to the uncertainty of bird maneuvering flight through a multi-hypothesis trajectory fusion mechanism for five typical flight behaviors, significantly improving trajectory prediction accuracy in complex scenarios such as turning and ascent / descendance. Through a dual-layer gimbal control strategy of "coarse filtering of spatial deviation + fine sub-level of angular deviation", the frequency of invalid gimbal movements and mechanical impacts are greatly reduced while ensuring tracking response speed and aiming accuracy, achieving a balance between driving-away effect and equipment lifespan. Through a full-process closed-loop correction and multi-dimensional termination protection mechanism, the robustness, safety and compliance of system operation are taken into account.

[0095] Example 2

[0096] Please see Figure 7 This embodiment provides a schematic diagram of a laser-based early interception and bird deterrence system based on trajectory prediction.

[0097] As an example, the system is implemented using the laser-based early interception and bird deterrence method based on trajectory prediction as described in Example 1. The system includes: The time-series observation dataset generation module 71 is suitable for continuously acquiring bird flight video sequences through a PTZ camera, extracting the pixel coordinates of birds in N consecutive frames of images, converting the pixel coordinates into three-dimensional observation coordinates in the camera coordinate system, and generating a time-series observation dataset sorted by time. Kalman filter module 72 is suitable for iteratively estimating the motion state of birds based on the time-series observation dataset using the Kalman filter algorithm, and outputting the current time posterior optimal state vector containing position, velocity, and acceleration. The predicted flight trajectory generation module 73 is suitable for generating candidate trajectories corresponding to at least two typical flight behaviors of birds, using the posterior optimal state vector as the initial condition for motion; calculating the corresponding weight coefficients based on the historical matching degree between historical observation data and the historical candidate trajectories; and performing weighted fusion on each candidate trajectory to obtain the predicted flight trajectory within a future preset time window. The optimal laser interception point generation module 74 is suitable for constructing the protection area of ​​the equipment body and expanding the safety boundary outward to obtain the danger area; it filters candidate interception points that meet the pre-constraints of the danger area and the hardware reachability constraints on the predicted flight trajectory; it calculates the drive-off effect score of each candidate interception point through the drive-off effect evaluation function, and selects the point with the highest score as the optimal laser interception point; The laser aiming module 75 is suitable for calculating the azimuth and elevation angles of the gimbal target based on the three-dimensional coordinates of the optimal laser interception point; calculating the gimbal activation timing based on the estimated time when the bird arrives at the interception point; and controlling the gimbal to rotate to the target angle and complete laser aiming before the bird arrives. The dynamic correction and closed-loop control module 76 is suitable for updating the motion state and predicted flight trajectory for each frame of bird observation data acquired during laser irradiation, recalculating the optimal laser interception point; calculating the spatial deviation between the old and new optimal laser interception points, triggering an interception point update when the spatial deviation exceeds a preset spatial deviation threshold; calculating the gimbal angle deviation based on the updated interception point, and executing corresponding gimbal adjustment strategies according to the magnitude of the angle deviation; monitoring the drive-away status, terminating laser irradiation and executing corresponding operations when any preset termination condition is met.

[0098] It is not difficult to see that this embodiment is a system implementation corresponding to the first embodiment, and this embodiment can be implemented in conjunction with the first embodiment. The relevant technical details mentioned in the first embodiment are still valid in this embodiment, and will not be repeated here to reduce repetition. Accordingly, the relevant technical details mentioned in this embodiment can also be applied to the first embodiment.

[0099] Example 3

[0100] Please see Figure 8 The present invention also provides an electronic device, including: a memory and a processor; the memory stores at least one program instruction; the processor loads and executes the at least one program instruction to implement the laser-based early interception and bird deterrence method based on trajectory prediction provided in Embodiment 1.

[0101] The memory 702 and processor 701 are connected via a bus, which may include any number of interconnecting buses and bridges, connecting various circuits of one or more processors 701 and memory 702 together. The bus may also connect various other circuits, such as peripheral devices, voltage regulators, and power management circuits, which are well known in the art and therefore will not be described further herein. A bus interface provides an interface between the bus and the transceiver. The transceiver may be a single element or multiple elements, such as multiple receivers and transmitters, providing a unit for communicating with various other devices over a transmission medium. Data processed by processor 701 is transmitted over a wireless medium via an antenna, which further receives data and transmits it to processor 701.

[0102] Processor 701 is responsible for managing the bus and general processing, and can also provide various functions, including timing, peripheral interfaces, voltage regulation, power management, and other control functions. Memory 702 can be used to store data used by processor 701 during operation.

[0103] Example 4

[0104] This invention also proposes a storage medium storing a laser-based early interception and bird-repelling method based on trajectory prediction. When the substation laser-based early interception and bird-repelling program based on multiple hypothesis trajectory prediction is executed by a processor, it implements the steps of the laser-based early interception and bird-repelling method based on trajectory prediction as described above. Since this storage medium adopts all the technical solutions of all the above embodiments, it has at least all the beneficial effects brought about by the technical solutions of the above embodiments, which will not be repeated here.

[0105] The above descriptions are merely embodiments of the present invention. Commonly known structures and characteristics are not described in detail here. Those skilled in the art are aware of all common technical knowledge in the field prior to the application date or priority date, are aware of all existing technologies in that field, and have the ability to apply conventional experimental methods prior to that date. Those skilled in the art can, under the guidance of this application, improve and implement this solution in combination with their own capabilities. Some typical known structures or methods should not be obstacles for those skilled in the art to implement this application. It should be noted that those skilled in the art can make several modifications and improvements without departing from the structure of the present invention. These should also be considered within the scope of protection of the present invention, and will not affect the effectiveness of the implementation of the present invention or the practicality of the patent. The scope of protection claimed in this application should be determined by the content of its claims, and the specific embodiments described in the specification can be used to interpret the content of the claims.

Claims

1. A laser-based early interception and bird-repelling method based on trajectory prediction, characterized in that, The method includes: Step S1: Continuously acquire bird flight video sequences using a PTZ camera, extract the pixel coordinates of birds from N consecutive frames, convert the pixel coordinates into three-dimensional observation coordinates in the camera coordinate system, and generate a time-series observation dataset sorted by time. Step S2: Based on the time-series observation dataset, the Kalman filter algorithm is used to iteratively estimate the motion state of the birds and output the posterior optimal state vector at the current time containing position, velocity, and acceleration. Step S3: Using the posterior optimal state vector as the initial condition for motion, generate candidate trajectories corresponding to at least two typical bird flight behaviors; calculate the corresponding weight coefficients based on the historical matching degree between historical observation data and the historical data of each candidate trajectory; perform weighted fusion on each candidate trajectory to obtain the predicted flight trajectory within a future preset time window; Step S4: Construct the protection zone of the device body and extend the safety boundary outward to obtain the danger zone; screen candidate interception points that meet the pre-constraints of the danger zone and the hardware reachability constraints on the predicted flight trajectory; calculate the expulsion effect score of each candidate interception point through the expulsion effect evaluation function, and select the point with the highest score as the optimal laser interception point; Step S5: Calculate the target azimuth and elevation angles of the gimbal based on the three-dimensional coordinates of the optimal laser interception point; calculate the gimbal activation timing based on the estimated time when the bird arrives at the interception point, and control the gimbal to rotate to the target angle and complete laser aiming before the bird arrives; Step S6: During laser irradiation, the motion state and predicted flight trajectory are updated for each frame of bird observation data acquired, and the optimal laser interception point is recalculated; the spatial deviation between the old and new optimal laser interception points is calculated, and the interception point is updated when the spatial deviation exceeds the preset spatial deviation threshold; the gimbal angle deviation is calculated based on the updated interception point, and the corresponding gimbal adjustment strategy is executed according to the magnitude of the angle deviation; the driving status is monitored, and the laser irradiation is terminated and the corresponding operation is executed when any preset termination condition is met.

2. The laser-based early interception and bird-repelling method based on trajectory prediction according to claim 1, characterized in that, Step S1 includes: Step S11: Use a target detection algorithm to identify bird targets in a single frame image, output the pixel coordinates and detection confidence of the center point of each bird target, and filter targets with a detection confidence lower than the preset confidence threshold. Step S12: Use the ByteTrack tracking algorithm to assign a unique tracking ID to each bird target, complete the association matching of the same target across frames, and generate the pixel coordinate time sequence corresponding to each target; Step S13: Based on the pinhole camera projection model, the pixel coordinates of each target are inversely projected using the camera intrinsic parameter matrix to obtain the three-dimensional direction vector of the corresponding target in the camera coordinate system. Step S14: Combine the target straight-line distance obtained by the ranging unit with the three-dimensional direction vector to calculate the three-dimensional observation coordinates of the bird target in the camera coordinate system; Step S15: Group the bird targets by tracking ID, sort the three-dimensional observation coordinates of each bird target by timestamp, and generate the time-series observation dataset of the corresponding target.

3. The laser-based early interception and bird-repelling method based on trajectory prediction according to claim 1, characterized in that, Step S2 includes: Step S21: Define the motion state vector ,in The true three-dimensional observation coordinates of the target in the camera coordinate system, The instantaneous three-axis velocity of the target in the camera coordinate system, The instantaneous acceleration of the target along three axes in the camera coordinate system; Step S22: Construct the state transition matrix F, observation matrix H, process noise covariance matrix Q, and observation noise covariance matrix R corresponding to the uniformly accelerated motion model; Step S23: Based on the posterior state estimate of the previous time step, calculate the prior state estimate and prior covariance matrix of the current time step, and execute the prediction step of Kalman filtering. Step S24: Introduce the current frame observation vector from the time-series observation dataset, calculate the Kalman gain, correct the prior state, obtain the posterior optimal motion state vector and posterior covariance matrix at the current time, and execute the Kalman filter update step. Step S25: During the cold start of filtering, the initial state vector and the initial covariance matrix are calculated using the observation data of the first three frames through difference operations.

4. The laser-based early interception and bird-repelling method based on trajectory prediction according to claim 1, characterized in that, The candidate trajectories generated in step S3, which correspond to at least two typical bird flight behaviors, include: straight trajectory, left-turn trajectory, right-turn trajectory, dive trajectory, and / or climb trajectory. The mathematical expression for the straight line trajectory is: ; The mathematical expression for the left-turn trajectory is: ; The mathematical expression for the right-turn trajectory is: ; The mathematical expression for the dive trajectory is: ; The mathematical expression for the climbing trajectory is: ; ； In the formula, Let be the bird's initial three-dimensional position vector at the current moment. The initial three-dimensional velocity vector, The velocity component of the bird at the current moment. The turning radius is The net velocity of the birds at the current moment. For the maximum steering angular rate, The heading angle of the bird at the current moment. For equivalent gravitational acceleration, This represents the bird's current altitude.

5. The laser-based early interception and bird-repelling method based on trajectory prediction according to claim 4, characterized in that, In step S3, the corresponding weight coefficients are calculated based on the historical matching degree between historical observation data and each candidate trajectory. The candidate trajectories are weighted and fused to obtain the predicted flight trajectory within a future preset time window, including: Calculate the matching degree between each candidate trajectory and historical observation trajectories: ； In the formula, Let be the actual observed position vector at the k-th historical moment. Let be the predicted position vector of the i-th candidate trajectory at the k-th historical moment. For historical frames, The historical matching score for the i-th candidate trajectory. For error sensitivity coefficient, It is a natural exponential function; Calculate the weights corresponding to each candidate trajectory model based on the matching degree: ； In the formula, Let be the weight coefficient of the i-th candidate trajectory; The weighted fusion of the candidate trajectory models yields the predicted flight trajectory within a future preset time window, including: ； In the formula, Let be the i-th candidate trajectory.

6. The laser-based early interception and bird-repelling method based on trajectory prediction according to claim 1, characterized in that, Step S4 includes: Step S41: Define the electrical equipment to be protected as a cylinder, cuboid, or combined geometric model, and extend it outward by a preset safety boundary distance to form a danger zone; Step S42: Calculate the time of intersection between the predicted flight trajectory and the boundary of the danger zone, set the total lead time, and take the trajectory point at the time corresponding to the difference between the total lead time and the time of intersection as the initial interception point. Step S43: Verify whether the initial interception point is within the effective range of the laser and the rotation angle of the gimbal; Step S44: In response to the initial interception point not being within the effective range of the laser and the rotation angle of the gimbal, search backward along the trajectory for the first reachable point as the optimal laser interception point; Step S45: In response to the initial interception point being within the effective range of the laser and the rotation angle of the gimbal, sample candidate interception points with a preset step size within a preset time range before and after the initial interception point, substitute the candidate interception points into the expulsion effect evaluation function to calculate the score, and select the point with the highest score as the optimal laser interception point.

7. The laser-based early interception and bird-repelling method based on trajectory prediction according to claim 6, characterized in that, The mathematical expression for the repulsion effect evaluation function is: ； ； In the formula, The overall score for the deterrence effect of candidate interception point I. , , These are the weighting coefficients. d is the straight-line propagation distance from the laser emitter to the candidate interception point. This represents the maximum effective range of the laser. The standard deviation of the Gaussian kernel for distance decay. The angle between the laser incident direction and the bird's flight path. The remaining time for birds to reach the danger zone. This is the minimum safe time threshold.

8. The laser-based early interception and bird-repelling method based on trajectory prediction according to claim 1, characterized in that, Step S5 includes: Step S51: Using the gimbal base as the spatial origin, calculate the gimbal target azimuth angle and target elevation angle based on the three-dimensional coordinates of the optimal laser interception point; Step S52: Compare the gimbal target azimuth angle and target pitch angle with the gimbal physical limit range, and perform amplitude limiting and clipping processing to obtain the gimbal target angle; Step S53: Use a seven-segment S-shaped acceleration planning method to generate the gimbal motion speed curve to ensure no mechanical impact during the motion process; Step S54: Calculate the gimbal activation time based on the time the bird arrives at the interception point and the total duration of the gimbal movement, and control the gimbal to complete its positioning before the bird arrives at the interception point.

9. The laser-based early interception and bird-repelling method based on trajectory prediction according to claim 1, characterized in that, Step S6 includes: Step S61: During the laser irradiation process, update the observation data for each frame, re-execute the Kalman filter operation and trajectory prediction, and calculate the new optimal laser interception point; Step S62: Calculate the spatial deviation between the old and new interception points. When the spatial deviation exceeds a preset spatial deviation threshold, trigger the interception point update. Step S63: When the interception point is updated, repeat step S5 to calculate the new gimbal target angle based on the updated optimal laser interception point. Step S64: Based on the new gimbal target angle and the gimbal target angle before the update, calculate the gimbal azimuth axis angle deviation and pitch axis angle deviation respectively, and compare the gimbal azimuth axis angle deviation and pitch axis angle deviation with the preset angle deviation threshold respectively. Step S65: When the angle deviation of any axis is less than the preset angle deviation threshold, the axis is directly fine-tuned in speed mode; when the angle deviation of any axis is greater than or equal to the preset angle deviation threshold, the axis is re-planned with a seven-segment S-shaped smooth motion trajectory, and the two axes execute their respective motion control commands synchronously. Step S66: The termination conditions include: successful deflection condition, target flight condition, interception failure condition, timeout protection condition, and / or trajectory deviation condition. When any termination condition is met, the laser irradiation is terminated and the corresponding action is executed.

10. A laser-based early interception bird deterrence system based on trajectory prediction, wherein the system is implemented using the laser-based early interception bird deterrence method based on trajectory prediction as described in any one of claims 1-9, characterized in that, The system includes: The time-series observation dataset generation module is suitable for continuously acquiring bird flight video sequences through a PTZ camera, extracting the pixel coordinates of birds in N consecutive frames of images, converting the pixel coordinates into three-dimensional observation coordinates in the camera coordinate system, and generating a time-series observation dataset sorted by time. The Kalman filter module is suitable for iteratively estimating the motion state of birds based on the aforementioned time-series observation dataset using the Kalman filter algorithm, and outputting the current time-posterior optimal state vector containing position, velocity, and acceleration. The predicted flight trajectory generation module is suitable for generating candidate trajectories corresponding to at least two typical flight behaviors of birds, using the posterior optimal state vector as the initial condition for motion; calculating the corresponding weight coefficients based on the historical matching degree between historical observation data and the historical candidate trajectories; and performing weighted fusion on each candidate trajectory to obtain the predicted flight trajectory within a future preset time window. The optimal laser interception point generation module is suitable for constructing the protection area of ​​the equipment body and expanding the safety boundary outward to obtain the danger area; it filters candidate interception points that meet the pre-constraints of the danger area and the hardware reachability constraints on the predicted flight trajectory; it calculates the drive-off effect score of each candidate interception point through the drive-off effect evaluation function, and selects the point with the highest score as the optimal laser interception point; The laser aiming module is suitable for calculating the azimuth and elevation angles of the gimbal target based on the three-dimensional coordinates of the optimal laser interception point; calculating the gimbal activation timing based on the estimated time when the bird arrives at the interception point; and controlling the gimbal to rotate to the target angle and complete laser aiming before the bird arrives. The dynamic correction and closed-loop control module is suitable for updating the motion state and predicted flight trajectory for each frame of bird observation data acquired during laser irradiation, recalculating the optimal laser interception point; calculating the spatial deviation between the old and new optimal laser interception points, and triggering an interception point update when the spatial deviation exceeds a preset spatial deviation threshold; calculating the gimbal angle deviation based on the updated interception point, and executing corresponding gimbal adjustment strategies according to the magnitude of the angle deviation; monitoring the drive-away status, and terminating laser irradiation and executing corresponding operations when any preset termination condition is met.

Images