Adaptive trajectory filtering method and system based on unmanned aerial vehicle motion state, and medium

Abstract

This invention discloses an adaptive trajectory filtering method, system, and medium based on the motion state of a UAV, relating to the field of UAV navigation and control technology. The method includes: calculating dynamic indicators of the UAV using real-time collected sensor data; determining the UAV to be in a hovering state when all dynamic indicators are less than separately set hovering indicator thresholds; otherwise, determining it to be in a moving state; dynamically adjusting the horizontal distance threshold of the UAV using the dynamic indicators and the determined UAV state; applying a sliding window weighted average filtering algorithm to the trajectory of the hovering UAV to obtain optimized trajectory points; and applying a Kalman filter algorithm to the trajectory of the moving UAV to obtain optimized trajectory points. This adaptive trajectory filtering method, system, and medium improves the filtering effect while reducing unnecessary computational burden.

Description

Technical Field

[0001] This invention relates to the field of unmanned aerial vehicle (UAV) navigation and control technology, and in particular to an adaptive trajectory filtering method, system, and medium based on the motion state of the UAV. Background Technology

[0002] When recording and displaying a drone's trajectory, it's necessary to record the drone's real-time reported location information (longitude, latitude, and altitude). When the drone hovers, due to GPS errors, sensor noise, or environmental interference (such as wind), the returned location may fluctuate within a very small range (e.g., ±0.5 meters). Directly using these points for trajectory display would cause unnecessary jitter in the trajectory, affecting the visualization effect. Removing invalid points can optimize trajectory smoothness, reduce the storage space occupied by invalid data, and improve trajectory loading efficiency to some extent.

[0003] The existing technical problems are as follows:

[0004] 1) Drawbacks of fixed threshold filtering: Existing UAV trajectory filtering methods mostly use fixed thresholds, which cannot adapt to the needs of different UAV motion states. When hovering, too many noise points may be retained, while when moving, effective points may be mistakenly filtered out.

[0005] 2) Waste of computational resources: High-performance filtering algorithms (such as Kalman filtering) cause unnecessary computational burden when hovering.

[0006] 3) Inaccurate state recognition: Existing methods are not accurate enough in judging the motion state of UAVs, which affects the filtering effect. Summary of the Invention

[0007] Based on the technical problems existing in the background technology, this invention proposes an adaptive trajectory filtering method, system and medium based on the motion state of UAVs, which not only adapts to the needs of different motion states of UAVs and improves the filtering effect, but also reduces unnecessary computational burden.

[0008] The adaptive trajectory filtering method based on the motion state of a UAV proposed in this invention includes:

[0009] The dynamic indicators of the UAV are calculated using real-time sensor data. When all dynamic indicators are less than the respective set hovering indicator thresholds, the UAV is determined to be in hovering state; otherwise, it is in motion state. The dynamic indicators include the current real-time speed, acceleration, GPS positioning accuracy, and IMU data confidence.

[0010] The horizontal distance threshold of the drone is dynamically adjusted using the aforementioned dynamic indicators and the determined drone status.

[0011] The sliding window weighted mean filtering algorithm is used to filter the trajectory of a hovering drone. The horizontal distance between the weighted mean of the current window and the weighted mean of the previous output is calculated. If the horizontal distance is greater than or equal to the adjusted horizontal distance threshold, the weighted mean of the sliding window is output as the optimized trajectory point.

[0012] By using the Kalman filter algorithm to fuse observation data and system model predictions, trajectory filtering is applied to the moving UAV to generate posterior state estimates, which are then output as optimized trajectory points.

[0013] Furthermore, before calculating the dynamic indicators of the UAV using real-time acquired sensor data, the system is initialized and initial parameters are loaded. These initial parameters include hovering indicator threshold, sliding window capacity, and process noise covariance matrix of Kalman filter.

[0014] Furthermore, in the process of calculating the dynamic indicators of the UAV using real-time acquired sensor data, the process noise covariance matrix of the Kalman filter is updated as follows:

[0015] ;

[0016] ;

[0017] in, For the current real-time speed, The updated process noise covariance matrix, The process noise covariance matrix before the update. For the updated number Noise of each observation This is the horizontal accuracy factor of GPS. Confidence level of IMU data.

[0018] Furthermore, in dynamically adjusting the horizontal distance threshold of the drone using the dynamic indicators and the determined drone status, the process includes: dynamically adjusting the horizontal distance threshold of the drone in a hovering state using the GPS positioning accuracy, and dynamically adjusting the horizontal distance threshold of the drone in a moving state using the current real-time speed.

[0019] Furthermore, the dynamic adjustment of the horizontal distance threshold of the hovering UAV using the GPS positioning accuracy specifically involves:

[0020] ;

[0021] in, The horizontal distance threshold. This is the horizontal accuracy factor of GPS. This is the GPS accuracy attenuation factor.

[0022] Furthermore, the dynamic adjustment of the horizontal distance threshold of the moving drone using the current real-time speed specifically involves:

[0023] ;

[0024] in, The horizontal distance threshold. This represents the current real-time speed.

[0025] Furthermore, the method of using a sliding window weighted mean filtering algorithm to filter the hovering drone specifically involves:

[0026] Add the drone's new location to the tail of the sliding window queue until the current window is full;

[0027] Calculate the weighted average for the full-load window :

[0028] ;

[0029] ;

[0030] in, The position of the data point in the current window. This represents the total number of points in the current window. The time decay factor, For the current window The weight of the sequence number For the current window The position point corresponding to each serial number.

[0031] Furthermore, after obtaining the optimized trajectory points, the optimized trajectory points, timestamps, UAV status labels, and the type of filtering algorithm used are encapsulated into standardized trajectory points, and the WGS84 coordinates of the standardized trajectory points are converted into ECEF coordinates, specifically:

[0032] ;

[0033] in, For standardized trajectory points, It was a moment. These are the optimized longitude, latitude, and altitude of the drone. The optimized trajectory points are formed. For drone status labels, The filtering algorithm used is, specifically, for a hovering drone. For a moving drone, a sliding window weighted mean filtering algorithm is used. This is the Kalman filter algorithm.

[0034] A computer system includes a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the method described above.

[0035] A computer-readable storage medium storing a plurality of classification programs, the plurality of classification programs being invoked by a processor to execute the method described above.

[0036] Those skilled in the art will understand that all or part of the steps of the above method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it performs the steps of the above method embodiments. The aforementioned storage medium includes various media that can store program code, such as ROM, RAM, magnetic disk, or optical disk.

[0037] The advantages of the adaptive trajectory filtering method, system, and medium based on the motion state of the UAV provided by this invention are as follows: It adopts a dynamic adjustment of the horizontal distance threshold of the UAV to adapt to the needs of different motion states of the UAV. In addition, it uses a sliding window weighted mean filtering algorithm to filter the UAV in the hovering state, avoiding the unnecessary computational burden caused by the existing Kalman filtering algorithm for both hovering and moving UAVs. Attached Figure Description

[0038] Figure 1 This is a schematic diagram of the process of the present invention;

[0039] Figure 2 This is a flowchart illustrating the adaptive trajectory filtering method.

[0040] Figure 3 This is a flowchart illustrating the status recognition module.

[0041] Figure 4 The switching logic diagram for the filtering algorithm. Detailed Implementation

[0042] The technical solution of the present invention will now be described in detail through specific embodiments. Many specific details are set forth in the following description to provide a thorough understanding of the invention. However, the present invention can be implemented in many other ways different from those described herein, and those skilled in the art can make similar modifications without departing from the spirit of the invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0043] like Figures 1 to 4 As shown, the adaptive trajectory filtering method based on the motion state of a UAV proposed in this invention includes:

[0044] Step 1: Calculate the dynamic indicators of the UAV using real-time sensor data. When all dynamic indicators are less than the set hovering indicator thresholds, the UAV is determined to be in hovering state; otherwise, it is in motion state. The dynamic indicators include the current real-time speed, acceleration, GPS positioning accuracy, and IMU data confidence.

[0045] Step 2: Dynamically adjust the horizontal distance threshold of the drone using the dynamic indicators and the determined drone status;

[0046] Step 3: Use the sliding window weighted mean filtering algorithm to filter the hovering drone. Calculate the horizontal distance between the current window weighted mean and the previous output weighted mean. If the horizontal distance is greater than or equal to the adjusted horizontal distance threshold, output the mean of the sliding window as the optimized trajectory point.

[0047] Step 4: Use the Kalman filter algorithm to fuse the observation data and the system model predictions to perform trajectory filtering on the UAV in motion, generate posterior state estimates, and output them as optimized trajectory points.

[0048] Step 5: Encapsulate the optimized trajectory points, timestamps, UAV status tags, and the type of filtering algorithm used into standardized trajectory points, and convert the WGS84 coordinates of the standardized trajectory points into ECEF coordinates (Geocentric Cartesian coordinate system).

[0049] The observation data consists of the raw output of the sensors (GPS / IMU / barometer), including latitude and longitude output by the GPS module, triaxial acceleration and angular velocity measured by the IMU, and relative altitude provided by the barometer.

[0050] The WGS84 coordinate system, short for World Geodetic System, is a geodetic system standard used in cartography, geodesy, and navigation (including the Global Positioning System). WGS84 coordinates are coordinates in the WGS84 coordinate system. The ECEF coordinate system (Earth-Centered, Earth-Fixed) is a geodetic coordinate system with the Earth's center of mass as its origin. It is widely used in the Global Positioning System (GPS) and geographic information fields. ECEF coordinates are coordinates in the ECEF coordinate system.

[0051] This embodiment adopts a dynamic adjustment of the horizontal distance threshold of the UAV to adapt to the needs of different UAV motion states. In addition, a sliding window weighted mean filtering algorithm is used to filter the UAV in hovering state, avoiding the unnecessary computational burden caused by the existing Kalman filtering algorithm for both hovering and moving UAVs.

[0052] In one embodiment, before step one, system initialization is performed and initial parameters are loaded. These initial parameters include a hovering metric threshold, a sliding window capacity, and the process noise covariance matrix of the Kalman filter. The hovering metric threshold includes a hovering speed threshold. Hovering acceleration threshold GPS positioning accuracy threshold and IMU data confidence threshold.

[0053] As an example of initial parameter setting:

[0054] Hovering speed threshold ;

[0055] Hovering acceleration threshold ;

[0056] GPS positioning accuracy threshold ;

[0057] IMU data confidence level ,in The confidence level of the calculated IMU data;

[0058] Sliding window capacity: N = 10;

[0059] Kalman filter process noise covariance matrix:

[0060] ;

[0061] in, The original process noise covariance matrix describes the uncertainty of the system model prediction and reflects the errors in the UAV motion model (such as sudden wind disturbances and inertial errors). To observe the noise covariance matrix, we describe the uncertainty of sensor measurements and reflect the measurement errors of GPS / IMU (such as signal drift and noise). , The initial value is set based on the sensor calibration data and is dynamically adjusted during operation through the status recognition module.

[0062] It is understandable that IMU data is information about the motion state of an object collected by an Inertial Measurement Unit. An IMU consists of a three-axis accelerometer and a three-axis gyroscope, which measure the linear acceleration and angular velocity of an object in three-dimensional space, respectively.

[0063] In one embodiment, step one involves calculating the dynamic indicators of the drone using real-time sensor data. When all dynamic indicators are less than the respective set hovering indicator thresholds, the drone is determined to be in a hovering state; otherwise, it is in a moving state. Specifically:

[0064] The GPS module outputs longitude via the sensor module. ,latitude ,high HDOP value (i.e., GPS positioning accuracy), where longitude ,latitude The current location of the drone was limited;

[0065] The triaxial acceleration is output through the IMU module of the sensor module. Triaxial angular velocity .

[0066] It should be noted that, due to the fundamental differences in sensor accuracy and control characteristics between the vertical and horizontal channels of the UAV, this embodiment decouples the horizontal and vertical motion and performs state recognition only based on the horizontal XY plane velocity, which can avoid the interference of altitude measurement noise on the horizontal trajectory filtering.

[0067] Therefore, the dynamic index is calculated as follows:

[0068] Current real-time speed ,in, These are the velocities along the X and Y axes, respectively.

[0069] Growth Acceleration ,in, These are the accelerations along the X, Y, and Z axes, respectively.

[0070] The GPS positioning accuracy is achieved by directly obtaining the HDOP value through the aforementioned GPS module.

[0071] IMU data confidence level ,in, The average angular velocity within a sliding window over a set time period (e.g., 1 second). The standard deviation of the triaxial angular velocities within the sliding window. For the first The angular velocity measurement of the shaft, where Corresponding to X, Y, and Z respectively, therefore The maximum value is 3. This represents the absolute deviation of the angular velocity of each axis from the mean, and is expressed in terms of standard deviation normalization. The purpose is to eliminate the influence of dimensions and unify the deviation of each axis to the standard deviation unit.

[0072] IMU data consistency verification, i.e., IMU data confidence level: (5 consecutive frames).

[0073] like Figure 3 As shown, the rules for determining whether a drone is in a hovering or moving state in this embodiment are as follows:

[0074] ;

[0075] HOVERING status

[0076] else:

[0077] Status: MOVING;

[0078] HOVERING indicates the hovering state, while MOVING indicates the moving state. This is the acceleration due to gravity.

[0079] Finally, update the noise covariance matrix of the Kalman filter process:

[0080] ;

[0081] ;

[0082] in, For the current real-time speed, The updated process noise covariance matrix, The process noise covariance matrix before the update. For the updated number The noise of each observation, i.e. It is diagonal matrix ( (for the dimension of observation), its diagonal elements Indicates the first Noise of each observation For GPS positioning accuracy, Confidence level of IMU data.

[0083] This embodiment uses dynamic indicators to determine whether the UAV is hovering or moving, which facilitates the subsequent selection of a sliding window weighted mean filtering algorithm or a Kalman filtering algorithm to perform trajectory filtering on the UAV, thereby saving unnecessary computational burden.

[0084] In one embodiment, step two, dynamically adjusting the horizontal distance threshold of the drone using the dynamic indicators and the determined drone status, specifically involves:

[0085] Based on the two states of the drone (hovering state and moving state) obtained in step one, dynamic threshold adjustments are made for these two states respectively.

[0086] Threshold parameters for hovering state:

[0087] ;

[0088] Threshold parameters for motion state:

[0089] ;

[0090] in, The horizontal distance threshold. This is the horizontal accuracy factor of GPS. This is the GPS accuracy attenuation factor. This represents the current real-time speed.

[0091] As an example of dynamic threshold adjustment: (1) In hovering state, the horizontal position threshold is 0.1 meters, and the time duration threshold is 3 seconds or more in the same position as valid hovering. (2) In motion state, the horizontal position threshold is 0.5 meters; the speed change threshold is: This is considered a valid exercise.

[0092] It should be noted that the horizontal position threshold serves the following purpose: when determining whether a drone is hovering, the sliding window filter checks whether the horizontal displacement is greater than or equal to the horizontal distance threshold. If the value is greater than the target value, output the current sliding window weighted average as the optimized trajectory point; otherwise, discard the current trajectory point. Additionally, the horizontal threshold is... It is not a fixed value; for example, the horizontal position threshold in hovering mode is... It is not 0.1.

[0093] This embodiment employs dynamic threshold adjustment, which can adapt to the needs of different movement states of the UAV. It avoids retaining excessive noise points during hovering and prevents the filtering of valid points during movement. Specifically, it offers: 1) Enhanced environmental adaptability: Fixed thresholds cannot handle complex environmental changes (such as sudden wind speed changes or GPS signal attenuation). In hovering mode, a small threshold is used to capture minute valid displacements; in movement mode, a larger threshold is switched to avoid filtering maneuver trajectories. 2) Optimized computational resources: Traditional methods suffer from wasted computational power due to high-precision filtering throughout; dynamic optimization utilizes lightweight sliding window filtering in hovering mode and switches to Kalman filtering in movement mode.

[0094] In one embodiment, step three involves using a sliding window weighted mean filtering algorithm to filter the trajectory of the hovering drone, calculating the horizontal distance between the current window's weighted mean and the previous output weighted mean, and if the horizontal distance is greater than or equal to the adjusted horizontal distance threshold, then the average value of the sliding window is output as the optimized trajectory point.

[0095] like Figure 4 As shown, the adaptive trajectory filtering process in the hovering state is as shown in (a1) to (a4):

[0096] (a1) Add the new location of the UAV to the tail of the sliding window queue until the current window is full;

[0097] First, initialize an empty window queue with a window capacity of N (i.e., the maximum number of storage points); that is... New position at any moment Add a point to the tail of the sliding window queue, check if the window is full. If not, continue waiting for a new point to be added; if the window is full, the number of points equals N. They are respectively The longitude, latitude, and altitude of the moment.

[0098] When the window is full, adaptive trajectory filtering will be executed;

[0099] For window updates: If the number of points in the current window is greater than N, remove the oldest position point at the head of the queue, keeping the current window with N position points so that the next adaptive trajectory filtering can be performed.

[0100] (a2) Calculate the weighted average for the full-load window :

[0101] ;

[0102] ;

[0103] in, (Current point) The earliest point ), This represents the total number of points in the current window. As the time decay factor, this embodiment preferably uses... Used to control the rate at which historical data is forgotten. For the current window The weight of the sequence number For the current window The position point corresponding to each serial number.

[0104] (a3) Calculate the weighted mean of the current window using the Haversine formula. Weighted average of the previous output The horizontal distance between them;

[0105] ;

[0106] in, This is the Haversine formula, a mathematical method for calculating the spherical distance between two coordinates on Earth, based on the principles of spherical trigonometry.

[0107] (a4) If the horizontal distance is greater than or equal to the adjusted horizontal distance threshold, the weighted average of the output sliding window is used as the optimized trajectory point output.

[0108] Decision output:

[0109] ;

[0110] Understandably, if the horizontal distance is greater than or equal to the adjusted horizontal distance threshold, the weighted average of the output sliding window is used as the optimized trajectory point output. If the horizontal distance is less than the adjusted horizontal distance threshold, it is determined to be invalid jitter data, the current window data is discarded, and the filtering judgment is re-executed after new data is filled into the window.

[0111] In one embodiment, step four involves using a Kalman filter algorithm to fuse observation data with system model predictions to perform trajectory filtering on the moving UAV, generating a posterior state estimate, and outputting it as the optimized trajectory point. Figure 4 As shown, specifically:

[0112] (b1) State prediction: ;

[0113] Error covariance prediction : .

[0114] in This is the state transition matrix, which describes how the state evolves over time (based on a motion model). This represents the prior estimate obtained through the state transition matrix F, i.e., the predicted value of the system model. In the above equation, Q is initially... The Q value loaded subsequently is dynamically adjusted based on the drone's motion state (hovering / moving) and real-time speed. The adjustment is based on the real-time speed. .

[0115] Where the state transition matrix Dynamically select based on the drone's motion status:

[0116] The hovering state uses a constant velocity (FCV) model, which only predicts position and velocity;

[0117] The motion state adopts a uniform acceleration model (FCA), with an acceleration term added to cope with the changes in maneuverability.

[0118] Among them, CV (constant velocity), CA (constant acceleration), and CT (constant turning rate) are three common motion models.

[0119] (b2) Measurement update (observation correction):

[0120] Kalman gain calculation: ;

[0121] State correction: ;

[0122] Covariance correction: ;

[0123] in It is an identity matrix, with the same dimensions as the state vector, used to ensure... Dimensionality and stability of matrix operations. The state vector contains position (longitude, latitude) and velocity (eastward velocity, northward velocity).

[0124] (b3) Output filter value ;

[0125] in, for The state vector at time step describes the drone's state at time step. The state at a given time (position, velocity, etc.), where, This is a posterior estimate after fusing observational data; Let be the state estimation error covariance matrix, representing the uncertainty of the state estimation (a criterion for judging filter convergence); Q is the process noise covariance matrix, representing the uncertainty of the system model (such as sudden wind disturbances). In the initial stage, Q is... The following is the updated process noise covariance matrix; H is the Kalman gain matrix, which controls the weighting of predicted and observed values; H is the observation matrix, which maps the state vector to the observation space (which states can be directly measured); R is the observation noise covariance matrix, representing the sensor measurement error. The observation vector represents the actual measurement value of the sensor.

[0126] The Kalman filtering algorithm in this embodiment is as follows:

[0127] (1) Dynamic noise adjustment mechanism;

[0128] The process noise covariance matrix Q is dynamically adjusted based on the UAV's motion state (hovering / moving) and real-time speed, for example, during the moving state... Scaling addresses the filter divergence problem caused by sudden changes in maneuverability;

[0129] Observation noise covariance matrix R: combining GPS HDOP values ​​and IMU confidence. Dynamic calculation To suppress interference from low-quality sensors.

[0130] (2) State-driven algorithm switching:

[0131] Motion state recognition (hovering / movement) is used as the basis for selecting filtering algorithms: 1) Hovering state: Lightweight sliding window filtering (time decay weighted); 2) Motion state: Adaptive Kalman filtering;

[0132] Therefore, the Kalman filter in this embodiment has improved computational efficiency compared to the traditional fixed filtering mode.

[0133] In one embodiment, step five involves generating standardized trajectory points from the updated hovering index threshold and the filtered state vector, and then generating the final trajectory of the UAV state through coordinate transformation. Specifically:

[0134] (c1) Generate standardized trajectory points:

[0135] ;

[0136] in, For standardized trajectory points, It is a moment, as a timestamp. These are the optimized longitude, latitude, and altitude of the drone. The optimized trajectory points are formed. For drone status labels, The filtering algorithm used is, specifically, for a hovering drone. For a moving drone, a sliding window weighted mean filtering algorithm is used. This is the Kalman filter algorithm.

[0137] (c2) Coordinate system transformation (WGS84 to ECEF):

[0138] ;

[0139] in, The coordinates are in the ECEF coordinate system. , , The longitude, latitude, and altitude from the standardized trajectory points serve as the input for coordinate system transformation; N is the radius of curvature of the geoid. ; e is the semi-major axis of the WGS84 ellipsoid; e is the first eccentricity of the ellipsoid. b is the minor semi-axis of the WGS84 ellipsoid.

[0140] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. An adaptive trajectory filtering method based on the motion state of an unmanned aerial vehicle (UAV), characterized in that, include: The dynamic indicators of the UAV are calculated using real-time sensor data. When all dynamic indicators are less than the respective set hovering indicator thresholds, the UAV is determined to be in hovering state; otherwise, it is in motion state. The dynamic indicators include the current real-time speed, acceleration, GPS positioning accuracy, and IMU data confidence. The horizontal distance threshold of the drone is dynamically adjusted using the dynamic indicators and the determined drone status, including: dynamically adjusting the horizontal distance threshold of the drone in a hovering state using the GPS positioning accuracy, and dynamically adjusting the horizontal distance threshold of the drone in a moving state using the current real-time speed, specifically: ； ； in, The horizontal distance threshold. This is the horizontal accuracy factor of GPS. This is the GPS accuracy attenuation factor. This is the current real-time speed; The sliding window weighted mean filtering algorithm is used to filter the trajectory of a hovering drone. The horizontal distance between the weighted mean of the current window and the weighted mean of the previous output is calculated. If the horizontal distance is greater than or equal to the adjusted horizontal distance threshold, the weighted mean of the sliding window is output as the optimized trajectory point. By using the Kalman filter algorithm to fuse observation data and system model predictions, trajectory filtering is applied to the moving UAV to generate posterior state estimates, which are then output as optimized trajectory points.

2. The adaptive trajectory filtering method according to claim 1, characterized in that, Before calculating the dynamic indicators of the UAV using real-time acquired sensor data, the system is initialized and initial parameters are loaded. These initial parameters include hovering indicator threshold, sliding window capacity, and process noise covariance matrix of Kalman filter.

3. The adaptive trajectory filtering method according to claim 2, characterized in that, In the process of calculating the dynamic indicators of the UAV using real-time acquired sensor data, the noise covariance matrix of the Kalman filter is updated as follows: ； ； in, For the current real-time speed, The updated process noise covariance matrix, The process noise covariance matrix before the update. For the updated number Noise of each observation This is the horizontal accuracy factor of GPS. Confidence level of IMU data.

4. The adaptive trajectory filtering method according to claim 1, characterized in that, The method of using a sliding window weighted mean filtering algorithm to filter hovering drones is as follows: Add the drone's new location to the tail of the sliding window queue until the current window is full; Calculate the weighted average for the full-load window : in, The position of the data point in the current window. This represents the total number of points in the current window. The time decay factor, For the current window The weight of the sequence number For the current window The position point corresponding to each serial number.

5. The adaptive trajectory filtering method according to claim 1, characterized in that, After obtaining the optimized trajectory points, the optimized trajectory points, timestamps, UAV status labels, and the type of filtering algorithm used are encapsulated into standardized trajectory points. The WGS84 coordinates of these standardized trajectory points are then converted to ECEF coordinates, specifically: in, For standardized trajectory points, It was a moment. These are the optimized longitude, latitude, and altitude of the drone. The optimized trajectory points are formed. For drone status labels, The filtering algorithm used is, specifically, for a hovering drone. For a moving drone, a sliding window weighted mean filtering algorithm is used. This is the Kalman filter algorithm.

6. A computer system comprising a memory, a processor, and a computer program stored in the memory, characterized in that, The processor executes the computer program to implement the method as described in any one of claims 1-5.

7. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a plurality of classification programs, which are used by a processor to execute the method as described in any one of claims 1-5.

Images