Multi-source target trajectory fusion method based on geographic coordinate system

Abstract

The invention discloses a multi-source target trajectory fusion method based on a geographic coordinate system. The method comprises the following steps: receiving original target trajectory data collected by a plurality of detection devices, analyzing and extracting core parameters, constructing a standardized trajectory point structural body, and removing abnormal trajectory points; performing time synchronization processing on the standardized track data of each device, and performing track point traversal, timestamp sorting and linear interpolation point supplementation to enable a multi-source track to achieve unified alignment in a time dimension; performing pairwise matching on the multi-source trajectories after time synchronization, calculating distance difference, speed difference and course difference of corresponding points of the trajectories, and combining a preset threshold system to judge whether the trajectories are the same target or not; and carrying out weighted average fusion on the multi-source trajectories which are judged to be the same target based on a preset weight distribution mechanism, carrying out smoothing and prediction optimization on the fused trajectory through a Kalman filtering algorithm adaptive to the characteristics of a geographic coordinate system, and outputting a final integrated target trajectory.

Description

Technical Field

[0001] This application relates to the field of radar data fusion technology, and in particular to a method for fusing multi-source target trajectories based on geographic coordinate systems. Background Technology

[0002] In complex monitoring scenarios, accurate perception of target trajectories is the core foundation for situational assessment and command and control. Currently, the industry relies on equipment such as radar, spectrum detectors, and photoelectric sensors to acquire trajectory data, but there are prominent problems such as large blind spots in single-device monitoring, asynchronous time of multi-source data, target matching relying on a single dimension which is prone to misjudgment, trajectory filtering not adapted to the characteristics of geographic coordinate systems, and unreasonable weight allocation in data fusion.

[0003] Current mainstream solutions include simple data stitching, fixed-interval interpolation, single-dimensional matching, equal-weighted average fusion, and directly applying Cartesian coordinate system filtering models. These solutions generally suffer from core defects such as merely processing data superficially, failing to achieve true time synchronization, high mismatch rates, ignoring differences in device accuracy, and large prediction errors due to not considering the curvature of the Earth.

[0004] Therefore, there is currently a lack of a multi-source trajectory fusion strategy that balances accuracy and efficiency for complex scenarios to meet real-time monitoring needs. Summary of the Invention

[0005] This application provides a multi-source target trajectory fusion method based on geographic coordinate system to solve the above problems.

[0006] On the one hand, this application provides a multi-source target trajectory fusion method based on a geographic coordinate system. The method includes the following steps: Step S1: Receive raw target trajectory data collected by multiple detection devices, parse and extract core parameters, construct a standardized trajectory point structure containing preset parameter items, timestamps, device identifiers, and device accuracy levels, and remove abnormal trajectory points that exceed the physical reasonable range; Step S2: Perform time synchronization processing on the standardized trajectory data of each device, and through trajectory point traversal, timestamp sorting, and linear interpolation to supplement points, achieve unified alignment of multi-source trajectories in the time dimension, and ensure the physical consistency of trajectory parameters after interpolation; Step S3: Perform pairwise matching on the time-synchronized multi-source trajectories, and determine whether they are the same target by calculating the distance difference, speed difference, and heading difference of corresponding points on the trajectories, combined with a preset threshold system; Step S4: For multi-source trajectories determined to be the same target, perform weighted average fusion based on a preset weight allocation mechanism, and then use a Kalman filter algorithm adapted to the characteristics of the geographic coordinate system to smooth and predict the fused trajectory, and output the final integrated target trajectory.

[0007] In one implementation of this application, in step S1, the core parameters include longitude, latitude, altitude, speed, and heading. The construction process of the standardized trajectory point structure includes unifying the data accuracy of each core parameter. The criteria for eliminating abnormal trajectory points are the judgment standards set by the equipment accuracy level, and only trajectory points that conform to the laws of physical motion are retained.

[0008] In one implementation of this application, step S2 specifically includes: initializing a synchronization trajectory point container; using two pointers to traverse the two trajectories to be synchronized; adding trajectory points to the container sequentially according to timestamp size rules; selecting and retaining trajectory points with the same timestamp according to weight priority; sorting all trajectory points in the container in ascending order according to the timestamp dimension to eliminate time disorder issues; performing linear interpolation to supplement points at preset time intervals; calculating the time difference between adjacent trajectory points; generating interpolated time points and calculating the interpolation results of each core parameter based on the interpolation coefficients; adding the generated interpolated trajectory points to the interpolation container; and appending the end trajectory points of the synchronized trajectory to the interpolation container to ensure the integrity of the trajectory data.

[0009] In one implementation of this application, the interpolation coefficient in the linear interpolation point supplementation is calculated by the proportion of the timestamp difference between the interpolation time point and the adjacent trajectory points, and the interpolation result of each core parameter is derived based on the interpolation coefficient and the corresponding parameter value of the adjacent trajectory points.

[0010] In one implementation of this application, step S3 involves pairwise matching of the time-synchronized multi-source trajectories. Specifically, this includes: performing independent time-continuous interpolation on each trajectory participating in the matching to ensure the continuity of the time dimension of a single trajectory; using a two-pointer traversal strategy to perform timestamp matching on the two interpolated trajectories, and determining a valid matching point pair when the timestamp difference between the two trajectory points meets a preset compatibility condition; calculating the geographical distance difference, absolute speed difference, and absolute heading difference of the valid matching point pair, and accumulating the total difference value of each dimension; counting the number of valid matching point pairs, and determining different targets if the number is zero; if the number is not zero, calculating the average difference value of each dimension, and determining the same target when the average difference value of all dimensions is less than the corresponding preset threshold.

[0011] In one implementation of this application, the preset threshold system is a dynamically adjustable multi-dimensional threshold set, and the adjustment of the threshold set is based on the characteristics of the monitoring scene, the target motion state, and the detection accuracy of the device.

[0012] In one implementation of this application, step S4, the weighted average fusion, specifically includes: constructing a weight evaluation mechanism based on the hardware performance, environmental adaptability, and historical data accuracy of the detection equipment; assigning corresponding weights to different source trajectory data of the same target; and performing weighted average calculation on the core parameters of each source trajectory to obtain preliminary fused trajectory data.

[0013] In one implementation of this application, the weight evaluation mechanism supports real-time dynamic updates, automatically adjusting the weight allocation ratio of the corresponding trajectory data when the working status of the detection device, environmental conditions, or data reliability changes.

[0014] In one implementation of this application, step S4, the Kalman filter prediction process includes: defining a state vector containing longitude, latitude, altitude, eastward velocity, and northward velocity; constructing a 5×5 state transition matrix adapted to the geographic coordinate system based on a uniform motion model; wherein the row vector corresponding to longitude in the state transition matrix contains correlation coefficients related to eastward velocity, and the row vector corresponding to latitude contains correlation coefficients related to northward velocity; calculating the predicted state vector for the next moment using the current state vector, the state transition matrix, and the time difference between adjacent prediction moments; and updating the predicted covariance matrix according to matrix operation rules, combining the state transition matrix, the covariance matrix of the previous moment, and the process noise matrix.

[0015] In one implementation of this application, the Kalman filter update process includes: constructing a measurement matrix that only relates to the position dimension of the state vector, wherein the position dimension includes longitude, latitude, and altitude; combining the measurement matrix, the prediction covariance matrix, and the measurement noise matrix, calculating the residual covariance matrix and solving for its inverse matrix; calculating the Kalman gain based on the prediction covariance matrix, the measurement matrix and its transpose, and the inverse matrix of the residual covariance matrix; and using the Kalman gain and the residual between the measured and predicted values ​​to correct the predicted state vector and update the covariance matrix simultaneously.

[0016] This application provides a multi-source target trajectory fusion method based on geographic coordinate systems, which has the following advantages: 1. By eliminating mismatches through multi-dimensional target matching, highlighting the value of high-precision equipment data through weighted fusion, and optimizing trajectory parameters through geographic adaptation Kalman filtering, accurate prediction and smoothing of target trajectories under geographic coordinates are achieved, significantly reducing positioning deviation and effectively overcoming the technical bottleneck of insufficient adaptation of traditional filtering models to geographic characteristics.

[0017] 2. By using time synchronization interpolation to achieve unified alignment of the time dimension of multi-source trajectories and data fusion of multiple devices to make up for the blind spots of single device monitoring, combined with anomaly point elimination to ensure data validity, the system achieves continuous coverage of target trajectories throughout the entire time period, reduces the impact of single device failures or errors, and significantly improves the reliability and completeness of monitoring results.

[0018] 3. By using a dynamically adjustable multi-dimensional threshold system and a real-time weight allocation mechanism based on equipment performance and environmental adaptability, the system balances fusion accuracy and algorithm time complexity, achieving flexible adaptation to different monitoring scenarios and equipment combinations. This meets the real-time monitoring needs in complex environments and enhances the engineering practicality and robustness of the solution. Attached Figure Description

[0019] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings: Figure 1 A flowchart of a multi-source target trajectory fusion method based on a geographic coordinate system is provided for an embodiment of this application. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0021] This application provides a method for multi-source target trajectory fusion based on a geographic coordinate system. The technical solution proposed in this application will be described in detail below with reference to the accompanying drawings.

[0022] Figure 1 A flowchart illustrating a multi-source target trajectory fusion method based on a geographic coordinate system, provided as an embodiment of this application. Figure 1 As shown, the method mainly includes the following steps: Step S1: Receive raw target trajectory data collected by multiple detection devices, parse and extract core parameters, construct a standardized trajectory point structure containing preset parameter items, timestamps, device identifiers and device accuracy levels, and remove abnormal trajectory points that exceed the physical reasonable range.

[0023] Specifically, raw trajectory data is acquired from multiple detection devices (radar, spectrum detectors, photoelectric sensors, etc.). Each trajectory consists of trajectory points with timestamps, containing information such as longitude (in degrees), latitude (in degrees), altitude (in meters), speed (in meters per second), and heading (in degrees, 0-360°). Standardization is achieved through the following steps: parsing the raw data formats of each device (such as the binary protocol of radar and the JSON format of spectrum devices) to extract core parameters; standardizing data precision (longitude and latitude are retained to 6 decimal places, and speed is retained to 1 decimal place), and constructing a standardized trajectory point structure (including parameter items, timestamps, device identifiers, and device precision levels); and removing outliers: setting thresholds based on device precision levels to remove trajectory points that exceed the physically reasonable range.

[0024] Step S2: Perform time synchronization processing on the standardized trajectory data of each device. By traversing trajectory points, sorting timestamps, and using linear interpolation to fill in the gaps, the multi-source trajectories achieve unified alignment in the time dimension and ensure the physical consistency of the interpolated trajectory parameters.

[0025] Specifically, because different radar and other detection devices may have different data acquisition frequencies, and the data may be discontinuous over time, it is necessary to interpolate the trajectory data of each detection device. The specific steps are as follows: 1. Traverse adjacent points in the trajectory and perform linear interpolation at fixed time intervals (e.g., 1 second) based on the timestamps and data values ​​of the two points.

[0026] 2. Calculate the longitude, latitude, altitude, speed, and heading of the interpolation point using linear interpolation formulas.

[0027] 3. Add interpolation points to the trajectory to make the trajectory data more dense and continuous in time, which facilitates subsequent matching and fusion operations.

[0028] Linear interpolation is a simple and effective data processing method used to estimate the value of unknown data points among known data points. In this scheme, linear interpolation makes the trajectory data more dense and continuous in time, facilitating subsequent matching and fusion operations. The formula for linear interpolation is as follows: y = y1 + (y2 - y1) / (x2 - x1) * (x - x1). Where (x1, y1) and (x2, y2) are known data points, x is the position to be interpolated, and y is the interpolation result.

[0029] To address the mismatch issue caused by different acquisition frequencies and asynchronous timestamps of trajectory data from multiple radar and spectrum detection devices, a trajectory point time synchronization and linear interpolation compensation mechanism was designed to achieve unified alignment of multi-source target trajectories in the time dimension. The specific process is as follows: 1. Trajectory point traversal and synchronization merging: Initialize the container storing synchronized trajectory points, and set two pointers to traverse the trajectory point sets of the two input radar and spectrum detection target trajectories (traj1, traj2) respectively. Perform preliminary merging of trajectory points according to the timestamp size rule: If the timestamp of the current trajectory point of traj1 is less than the timestamp of the current trajectory point of traj2, add the trajectory point of traj1 to the synchronization container, and move the traj1 traversal pointer forward; If the timestamp of the current trajectory point of traj1 is greater than the timestamp of the current trajectory point of traj2, add the trajectory point of traj2 to the synchronization container, and move the traj2 traversal pointer forward; If the timestamps of the two are exactly the same, compare the trajectory point weight values: If the weight of the trajectory point of traj1 is ≥ traj2, then add the trajectory point of traj1 to the container, otherwise select the trajectory point of traj2, and move the traversal pointers of both trajectories forward; After completing the core traversal, directly append the remaining trajectory points that have not been traversed in the two trajectories to the synchronization container to ensure that no trajectory points are missed.

[0030] 2. Synchronization Track Point Timestamp Sorting: For all track points within the synchronization container, an ascending sort is performed based on the timestamp dimension (by overloading the less than operator for track points, sorting is achieved only by timestamp), ensuring that all track points are arranged strictly in chronological order and eliminating the time disorder problem during track merging.

[0031] 3. Fixed-interval linear interpolation: Using a fixed time interval of 1 second, linear interpolation is performed on the sorted synchronized trajectory points to ensure the continuity of the trajectory in the time dimension. The interpolation trajectory point container is initialized, and the synchronized trajectory points are traversed. The currently traversed trajectory point is directly added to the interpolation container, and the timestamps of the current point (starting point) and the next point (ending point) are extracted, and the time difference is calculated. Starting from "starting point timestamp + 1 second", interpolation time points are generated cyclically (increasing by 1 second each time) until the interpolation time point is less than the ending point timestamp. For each interpolation time point, the interpolation coefficient t = (current interpolation time point - starting point timestamp) / (ending point timestamp - starting point timestamp) is calculated. Based on this coefficient, linear interpolation calculations are performed on the target parameters such as latitude, longitude, altitude, speed, and heading of the starting and ending points to generate new interpolated trajectory points and add them to the interpolation container. After the traversal is complete, the last synchronized trajectory point is manually appended to the interpolation container to ensure the integrity of the trajectory data.

[0032] 4. The output returns a set of trajectory points that have completed time synchronization and linear interpolation. The output trajectory data is continuously aligned at 1-second intervals, which solves the problem of data discontinuity caused by the difference in the acquisition frequency of multiple radars.

[0033] Step S3: Perform pairwise matching on the multi-source trajectories after time synchronization. Calculate the distance difference, speed difference, and heading difference between corresponding points on the trajectories, and determine whether they are the same target by combining the results with a preset threshold system.

[0034] Specifically, the trajectory data from different radars and spectrum detection devices are compared to determine which trajectories belong to the same target. By calculating the differences in distance, speed, and heading between the trajectories, it is determined whether they meet the criteria for the same target. The specific steps are as follows: 1. Interpolate the target trajectory of each detection device to ensure that the trajectories are aligned in time.

[0035] 2. Traverse the corresponding points in the two trajectories, calculate the differences in distance, speed and heading between them, and sum these differences.

[0036] 3. Calculate the average distance difference, average speed difference, and average heading difference.

[0037] 4. If these average differences are all less than a pre-set threshold, then the two trajectories are considered to belong to the same target.

[0038] This application targets multi-radar and spectrum target trajectory data. Through the core logic of "trajectory pre-interpolation - precise timestamp matching - multi-dimensional difference threshold determination", it achieves accurate determination of whether two radar trajectories point to the same target. The specific implementation technology is as follows: 1. Single-track time-continuous interpolation processing: Independent linear interpolation with a fixed interval of 1 second is performed on the two input radar tracks (traj1, traj2) to ensure the continuity of the time dimension of each track and eliminate the data discontinuity caused by the difference in the acquisition frequency within a single track. The interpolation yields time-continuous interpolated tracks syncedTraj2 and syncedTraj2.

[0039] 2. Timestamp matching and multi-dimensional difference calculation employ a two-pointer traversal strategy. Precise timestamp matching is performed on the two interpolated trajectories, and the multi-dimensional parameter differences between the matching point pairs are calculated: the total differences in distance, speed, and heading (totalDistDiff, totalSpeedDiff, totalHeadingDiff) are initialized to 0, the matching point pair count (matchCount) is initialized to 0, and the two pointers i and j are set to point to the first trajectory point of syncedTraj1 and syncedTraj2 respectively; the two interpolated trajectories are iterated repeatedly (until either trajectory has been traversed): if the current syncedTraj1[i] and syncedTraj2[i] are... If the absolute value of the timestamp difference of [j] is less than 1 second (compatible with minor timestamp errors), it is determined to be a valid matching point pair: Calculate the geographical distance difference of the point pair: Call the geographical distance calculation function, input the latitude and longitude coordinates of the two points, obtain the distance difference and accumulate it in totalDistDiff; Calculate the absolute speed difference of the point pair: Take the absolute difference of the speed values ​​of the two points and accumulate it in totalSpeedDiff; Calculate the absolute heading difference of the point pair: Take the absolute difference of the heading values ​​of the two points and accumulate it in totalHeadingDiff; Increment the matching point pair count matchCount by 1, and move both pointers i and j one position to the right; If the timestamp of syncedTraj1[i] is less than the timestamp of syncedTraj2[j] (no matching point), only move pointer i one position to the right; If the timestamp of syncedTraj1[i] is greater than the timestamp of syncedTraj2[j] (no matching point), only move pointer j one position to the right.

[0040] 3. Average Difference Calculation and Threshold Determination: Based on the multi-dimensional total difference value of matched point pairs, the average difference is calculated and compared with a preset threshold to output the target identity determination result: If the matched point pair count (matchCount) is 0 (no valid matched points), the two trajectories are directly determined to be different targets, and false is returned; if there are valid matched points, the average difference of three types of parameters is calculated respectively: such as average distance difference: avgDistDiff = total distance difference totalDistDiff / number of matched point pairs (matchCount); if the average distance difference, average speed difference, and average heading difference are all less than the preset threshold, the two trajectories are determined to point to the same target. This application first interpolates to ensure the time continuity of a single trajectory, then accurately matches the timestamp to calculate the multi-dimensional average difference, and combines it with threshold determination to achieve quantitative judgment of the target identity of multiple radar trajectories, solving the problems of fuzzy target attribution determination and large influence of subjective factors in multiple radar trajectory determination.

[0041] Target matching algorithms are used to determine whether target trajectories detected by different radars and spectrums belong to the same target. By calculating the differences in distance, speed, and heading between trajectories and comparing them with pre-set thresholds, it is determined whether they meet the criteria for the same target. This method is simple and intuitive, and can accurately determine target consistency to a certain extent.

[0042] Step S4: For multi-source trajectories identified as the same target, perform weighted average fusion based on a preset weight allocation mechanism, and then use a Kalman filter algorithm adapted to the characteristics of the geographic coordinate system to smooth and predict the fused trajectory, and output the final integrated target trajectory.

[0043] Specifically, trajectories belonging to the same target are fused to obtain a more accurate target trajectory. A weighted averaging method is used, assigning different weights to the trajectory data from different radars based on the measurement accuracy and reliability of each radar. Then, the information from corresponding points is weighted and averaged to obtain the fused trajectory data. Kalman filtering is then used to further optimize the fused trajectory, improving its accuracy and stability. The specific steps are as follows: 1. Determine the weight of each detection device, which can be evaluated based on factors such as the type, performance, and accuracy of historical data of the detection devices, such as radar and spectrum.

[0044] 2. For trajectories belonging to the same target, traverse the corresponding points in the trajectory and perform a weighted average of information such as longitude, latitude, altitude, speed, and heading according to the weights.

[0045] 3. Add the weighted average points to the fused trajectory to obtain the final target trajectory.

[0046] 4. Kalman Filter Update: The Kalman filter is applied to the initially fused trajectory data, which consists of two steps: prediction and update.

[0047] The trajectory fusion algorithm employs a weighted averaging method. Based on the measurement accuracy and reliability of each detection device, different weights are assigned to the trajectory data from different devices. Then, the information at corresponding points is weighted and averaged to obtain the fused trajectory data. This method fully utilizes the advantages of multiple detection devices, improving the accuracy and reliability of the target trajectory. The Kalman filter state and covariance prediction method is as follows: Kalman filtering is a recursive optimal estimation algorithm used to estimate the state of a system in the presence of noise. In this scheme, Kalman filtering is used to remove noise from the trajectory data, improving the accuracy of the trajectory. The main steps of Kalman filtering include prediction and updating: 1. Prediction steps: Based on the dynamic model of the system, predict the state vector and covariance matrix at the next time step.

[0048]

[0049]

[0050] in It is the predicted state vector, F k It is the state transition matrix. It is the estimated state vector from the previous time step, P. K|K-1 It is the predicted covariance matrix, P k-1|k-1 Qk is the covariance matrix of the previous time step, and Qk is the process noise covariance matrix.

[0051] This application addresses the target trajectory optimization requirements in a geographic coordinate system. Based on a uniform motion model, it implements the update of the state vector (latitude, longitude, and velocity) and covariance matrix during the Kalman filter prediction stage, adapting to accurate trajectory prediction in multi-radar-spectrum-photoelectric fusion scenarios. The specific process is as follows: (1) Function input and output description.

[0052] Input parameters: Kalman state structure (state), which includes the target's longitude (lon), latitude (lat), altitude, eastward velocity (v_east), northward velocity (v_north), 5×5 covariance matrix (P), 5×5 process noise matrix (Q); time interval (dt, i.e., the time difference between adjacent prediction times); Output format: No return value. It directly modifies the input Kalman state structure to complete the prediction update of the state vector and covariance matrix.

[0053] (2) State vector prediction (dynamic update of latitude and longitude).

[0054] Based on a uniform motion model (assuming the target has no acceleration and no vertical motion), the latitude and longitude of the next moment are predicted using the current velocity and time interval dt (altitude and east / north velocity remain unchanged).

[0055] (3) Construct a 5×5 state transition matrix (F matrix).

[0056] The state transition matrix F corresponds to the state vector dimensions of [longitude, latitude, altitude, eastward velocity, northward velocity]. Based on the uniform motion model, the matrix elements are defined to reflect the update rules of each state variable, as shown in the table below:

[0057] (4) Update of covariance matrix (P) (core formula: ).

[0058] The prediction and update of the covariance matrix are completed in two steps according to the matrix operation rules, and the process noise matrix Q is superimposed to reflect the system disturbance: Step 1: Calculate the product of F and the original covariance matrix P (F P).

[0059] Initialize a 5×5 temporary matrix temp, and iterate through each element temp[i][j]: (k ranges from 0 to 4), that is, calculating the dot product of the i-th row of matrix F and the j-th column of the original matrix P to obtain F. The result of P is stored in temp.

[0060] Step 2: Calculation Update the P matrix.

[0061] Iterate through each element of the new P matrix : First, calculate the dot product between the i-th row of the temp matrix and the j-th column of the F transpose matrix: (k ranges from 0 to 4), where F[j][k] is equivalent to the transpose of F. ,Finish The operation; The corresponding element state.Q[i][j] of the noise matrix Q during the superposition process is used to obtain the final prediction covariance matrix element state.P[i][j].

[0062] This method adapts to the characteristics of geographic coordinate systems (latitude and longitude), constructs a state transition matrix that fits reality through a uniform motion model, and accurately completes the state and covariance update in the Kalman filter prediction stage, thus solving the problem of geographic trajectory prediction error caused by directly applying the Cartesian coordinate system model in conventional Kalman filtering.

[0063] 2. Update steps: Based on the radar measurements, calculate the error (i.e., residual) between the measured and predicted values, and update the state vector and covariance matrix using Kalman gain.

[0064]

[0065]

[0066]

[0067] Among them, K k It is the Kalman gain, H k It is a measurement matrix, z k It is a measured value, R k It measures the noise covariance matrix. It is the updated state vector, P K|KIt is the updated covariance matrix.

[0068] This technology is the core implementation of the Kalman filter observation update stage. Based on target location data, it quickly corrects the predicted state and covariance, thereby improving the accuracy of the fused trajectory.

[0069] Core Input / Output - Input: Kalman state (latitude, longitude, altitude, velocity, covariance matrix P, measurement noise matrix R), multi-source observation positions (latitude, longitude, altitude); - Output: Corrected target state (position + velocity) and covariance matrix.

[0070] The key steps are as follows: (1) Construct a measurement matrix H as a 3×5 matrix, which only associates the position dimensions (latitude, longitude, and altitude) in the state.

[0071] (2) Calculate the residual covariance S according to the formula The calculation comprehensively reflects the uncertainty of prediction and measurement noise.

[0072] (3) Solve for the inverse matrix of S and calculate the determinant of S. If the determinant is close to 0 (not invertible), return directly; otherwise, solve for the inverse matrix S using the adjoint matrix method. -1 .

[0073] (4) Calculate the Kalman gain K according to the formula Calculate and balance the credibility of predictions and observations (the larger K is, the more confident the observations are).

[0074] (5) Correct the state vector.

[0075] (6) Update the covariance matrix P according to the formula Calculate (I is the identity matrix) to reduce the uncertainty of state estimation.

[0076] It can quickly fuse multi-source observation data, correct position and velocity status, and ensure robustness through irreversible judgment, significantly reducing positioning error and adapting to the real-time fusion requirements in complex scenarios.

[0077] The following is an example of a practical application scenario of this application. Low-altitude UAV target monitoring scenario: For multiple UAVs in urban low-altitude areas (radius of 10 kilometers), accurate trajectory tracking is achieved, providing data support for low-altitude defense and solving the problems of "easy to miss small UAV targets and easy to confuse the trajectories of multiple UAVs" in traditional monitoring.

[0078] The operating environment is as follows: Hardware configuration: Detection equipment combination: 1 ground phased array radar (acquisition interval 2 seconds), 1 portable spectrum detector (acquisition interval 5 seconds), 1 photoelectric device (for monitoring and identification). Data processing equipment: Lenovo Y9000P (Memory: 16GB, Hard drive: 1TB SSD), Operating system: Win11, Programming environment: C++, Qt; Application process: During the data acquisition phase, the UAV intrudes and departs, and radar, electro-optical, and spectrum sensors simultaneously detect the target. Data is received from these detection devices. The execution steps are as follows: Step 1: Data standardization. Analyze the raw data from multiple detection devices and standardize the format to "timestamp (milliseconds) + latitude and longitude (6 decimal places) + altitude (meters) + speed (meters / second) + heading (degrees)", removing outliers in speed and altitude (those exceeding reasonable ranges). Step 2: Time synchronization interpolation. Perform time synchronization interpolation at 1-second intervals on the trajectory of each device to achieve time alignment of the trajectories of more than 4 devices; Step 3: Multi-dimensional target matching. The trajectories of multiple devices are matched pairwise. If all trajectories are less than preset thresholds (30 meters, 3 meters / second, 10°), they are determined to be the same target. Step 4: Weighted fusion and filtering optimization. Weights are assigned based on device performance: phased array radar (0.4), spectrum detector (0.2), and photoelectric sensor (0.4). A weighted average is applied to the same target trajectory points from U1 to U4, and then corrected using a geographically adapted Kalman filter to obtain the preliminary fused trajectory. Step 5: Trajectory Integration and Output. Merge the trajectories of multiple devices targeting the same objective and output the final optimized trajectory.

[0079] In low-altitude UAV monitoring scenarios, this solution successfully achieves precise trajectories fusion of multi-source detection devices. Its positioning error, matching accuracy, and real-time performance are all superior to traditional solutions. It can effectively solve problems such as missed detections, confusion, and blind spots in low-altitude monitoring, and has clear engineering feasibility and practical application value.

[0080] The various embodiments in this application are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the device embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0081] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0082] The above description is merely an embodiment of this application and is not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A multi-source target trajectory fusion method based on a geographic coordinate system, characterized in that, The method includes the following steps: Step S1: Receive raw target trajectory data collected by multiple detection devices, parse and extract core parameters, construct a standardized trajectory point structure, and remove abnormal trajectory points; Step S2: Perform time synchronization processing on the standardized trajectory data by traversing trajectory points, sorting timestamps, and adding points using linear interpolation; Step S3: Perform pairwise matching on the multi-source trajectories after time synchronization, and determine whether they are the same target based on a preset threshold system; Step S4: For multi-source trajectories identified as the same target, perform weighted average fusion based on a preset weight allocation mechanism, and then use a Kalman filter algorithm adapted to the characteristics of the geographic coordinate system to smooth and predict the fused trajectory, and output the final integrated target trajectory.

2. The multi-source target trajectory fusion method based on geographic coordinate system according to claim 1, characterized in that, In step S1, the core parameters include longitude, latitude, altitude, speed, and heading. The construction process of the standardized trajectory point structure includes unifying the data accuracy of each core parameter. The criteria for eliminating abnormal trajectory points are the judgment standards set by the equipment accuracy level, and only trajectory points that conform to the laws of physical motion are retained.

3. The multi-source target trajectory fusion method based on geographic coordinate system according to claim 1, characterized in that, Step S2 specifically includes: Initialize the synchronization trajectory point container, use two pointers to traverse the two trajectories to be synchronized, add trajectory points to the container in sequence according to the timestamp size rule, and select and retain trajectory points with the same timestamp according to weight priority; Sort all trajectory points within the container in ascending order based on timestamp dimension to eliminate time disorder issues; Linear interpolation is performed at preset time intervals to supplement points, the time difference between adjacent trajectory points is calculated, interpolation time points are generated, and the interpolation results of each core parameter are calculated based on the interpolation coefficients. The generated interpolation trajectory points are then added to the interpolation container. The end trajectory points of the synchronized trajectory are appended to the interpolation container to ensure the integrity of the trajectory data.

4. The method for multi-source target trajectory fusion based on a geographic coordinate system according to claim 1, characterized in that, In the linear interpolation point supplementation, the interpolation coefficient is calculated by the proportion of the timestamp difference between the interpolation time point and the adjacent trajectory points. The interpolation results of each core parameter are derived based on the interpolation coefficient and the corresponding parameter values ​​of the adjacent trajectory points.

5. The multi-source target trajectory fusion method based on a geographic coordinate system according to claim 1, characterized in that, In step S3, pairwise matching is performed on the time-synchronized multi-source trajectories, specifically including: Independent time-continuous interpolation is performed on each trajectory involved in the matching process to ensure the continuity of the time dimension of each trajectory. A two-pointer traversal strategy is used to match the timestamps of the two interpolated trajectories. When the timestamp difference between the two trajectory points meets the preset compatibility condition, they are determined to be a valid matching point pair. Calculate the geographical distance difference, absolute speed difference, and absolute heading difference of valid matching point pairs, and sum the total difference values ​​for each dimension respectively; The system counts the number of valid matching pairs. If the count is zero, the pairs are considered different targets. If the count is not zero, the system calculates the average difference value for each dimension. When the average difference value for all dimensions is less than the corresponding preset threshold, the pairs are considered the same target.

6. The multi-source target trajectory fusion method based on a geographic coordinate system according to claim 1, characterized in that, The preset threshold system is a dynamically adjustable multi-dimensional threshold set, and the adjustment of the threshold set is based on the characteristics of the monitoring scene, the target motion state, and the detection accuracy of the equipment.

7. The multi-source target trajectory fusion method based on a geographic coordinate system according to claim 1, characterized in that, In step S4, the weighted average fusion specifically includes: constructing a weight evaluation mechanism based on the hardware performance, environmental adaptability, and historical data accuracy of the detection equipment; assigning corresponding weights to different source trajectory data of the same target; and performing weighted average calculation on the core parameters of each source trajectory to obtain preliminary fused trajectory data.

8. The multi-source target trajectory fusion method based on a geographic coordinate system according to claim 7, characterized in that, The weight evaluation mechanism supports real-time dynamic updates. When the working status of the detection device, environmental conditions, or data reliability changes, the weight allocation ratio of the corresponding trajectory data is automatically adjusted.

9. The multi-source target trajectory fusion method based on a geographic coordinate system according to claim 1, characterized in that, In step S4, the Kalman filter prediction process includes: Define a state vector containing longitude, latitude, altitude, eastward velocity, and northward velocity. Based on the uniform motion model, construct a 5×5 state transition matrix adapted to the geographic coordinate system. In the state transition matrix, the row vector corresponding to longitude contains the correlation coefficient related to the eastward velocity, and the row vector corresponding to latitude contains the correlation coefficient related to the northward velocity. Using the current state vector, the state transition matrix, and the time difference between adjacent prediction moments, calculate the predicted state vector for the next moment; According to the matrix operation rules, the predicted covariance matrix is ​​updated by combining the state transition matrix, the covariance matrix of the previous time step, and the process noise matrix.

10. The multi-source target trajectory fusion method based on a geographic coordinate system according to claim 9, characterized in that, The Kalman filter update process includes: Construct a measurement matrix that associates only the position dimension of the state vector, where the position dimension includes longitude, latitude, and altitude; By combining the measurement matrix, the prediction covariance matrix, and the measurement noise matrix, the residual covariance matrix is ​​calculated and its inverse matrix is ​​solved. The Kalman gain is calculated based on the prediction covariance matrix, the measurement matrix and its transpose, and the inverse of the residual covariance matrix. The predicted state vector is corrected by using Kalman gain and the residual between the measured and predicted values, and the covariance matrix is ​​updated at the same time.

Images