Target tracking method and system based on multi-source data fusion correlation

Abstract

The application provides a target tracking method and system based on multi-source data fusion correlation, comprising: inputting a single-frame target detection result, updating a track set according to a last time state of a formed track for a null time; filtering a non-null target detection result; judging whether there is a formed track in the current; if not, taking each target point as a track starting point; otherwise, calculating a matching measurement matrix between the formed track and target information detected in the current frame; two-dimensionally distributing the matching measurement matrix; updating a current formed track state according to a matching result; judging whether there is a target point not associated in the current frame; comparing corresponding values of the associated target in the matching measurement matrix with a threshold value, judging whether to add a new track, otherwise taking the target point not associated as a starting point of a new track; and updating the track set according to a last time state of each track.

Description

Technical Field

[0001] This invention relates to the field of spatial target recognition technology, and more specifically, to a target tracking method and system based on multi-source data fusion and correlation. Background Technology

[0002] Due to the limitations of their orbital cycles, reconnaissance satellites visit monitored areas intermittently, with long intervals between visits, typically on the order of hours. Therefore, the revisit rate of satellite reconnaissance information is low. Furthermore, influenced by the raw observation and processing algorithms, satellite reconnaissance information often possesses a degree of uncertainty and incompleteness. Under these conditions, it is difficult to reliably and effectively track and identify every target within a monitored area based solely on satellite reconnaissance information.

[0003] Patent document CN1O7704508A (application number: CN201710770466.3) discloses a data fusion and data mining method for massive amounts of multi-type space target data, including the following steps: Step 1: Collect multi-source space target data, perform data fusion on the obtained multi-source space target data to form a space target information database, construct a space target data management system based on the space target information database, and connect the space target data management system with simulation analysis software; Step 2: Perform data mining on the multi-source space target data in the space target information database, and at the same time provide data input for satellite Earth observation mission scheduling and planning simulation, and perform satellite Earth observation mission scheduling and planning simulation analysis.

[0004] Patent document CN107341220A (application number: CN201710506696.9) discloses a multi-source data fusion method and apparatus, comprising the following steps: Step 1: unifying the data format of the data in the dataset; Step 2: grouping the data of different entities that meet predetermined conditions into the same dataset; Step 3: for any entity, obtaining at least one associated attribute of each entity attribute; Step 4: obtaining the attribute similarity of the associated attributes of two entities; Step 5: if the attribute similarity is greater than the similarity threshold, then determining that the two entities are the same entity, and associating the entity attributes of the two entities with the same entity.

[0005] In real-world ship observations in actual sea environments, the satellite-borne sensors' access to the monitored area is intermittent due to the influence of satellite orbital periods, with long intervals between visits. Furthermore, sensor observations are inherently uncertain, exhibiting random positioning accuracy errors, and the errors of different sensors are not entirely consistent. This significantly increases the difficulty of maritime target identification and trajectory tracking. Traditional target identification and tracking algorithms have significant limitations when dealing with sparse target location information and large positional accuracy errors, failing to effectively correlate and maintain trajectories.

[0006] This invention provides a target tracking method and system based on multi-source data fusion and association. By projecting the position information obtained by spaceborne sensors into a planar coordinate system, and then fusing and filtering the position information of the same target observed by different sensors, the interference of sensor positioning accuracy error is effectively reduced, thereby further improving the accuracy of target trajectory association. Summary of the Invention

[0007] To address the shortcomings of existing technologies, the purpose of this invention is to provide a target tracking method and system based on multi-source data fusion and correlation.

[0008] A target tracking method based on multi-source data fusion and correlation provided by the present invention includes:

[0009] Step S1: Obtain multi-payload target detection information at the same time and get the target detection result of a single frame; when the detection result is empty, traverse each existing trajectory to obtain the target state at the last moment of the trajectory, and update the trajectory set according to the target state at the last moment of the formed trajectory.

[0010] Step S2: If the detection result is not empty, filter the current detection result to remove duplicate position information provided by different loads for the same target;

[0011] Step S3: Determine if there is currently an associated trajectory;

[0012] Step S4: If it does not exist, then each target point is used as the starting point of the trajectory, assembled into a defined format and stored;

[0013] Step S5: If it exists, calculate the matching metric matrix between the associated trajectory and the target information detected in the current frame; perform two-dimensional allocation on the calculated matching metric matrix to obtain the optimal matching result; based on the optimal matching result, apply the prediction method based on Kalman filtering to update the state of the currently associated trajectory.

[0014] Step S6: Determine whether there are any unassociated target points in the current frame; for associated targets, compare their corresponding values ​​in the matching metric matrix with the threshold value to determine whether to add a new trajectory; otherwise, use the unassociated target points as the starting points of the new trajectory.

[0015] Step S7: Assemble and store the new trajectory information according to the defined format;

[0016] Step S8: Update the trajectory set according to the state of each trajectory at the last moment.

[0017] Preferably, step S2 employs:

[0018] Step S2.1: Based on the load source of the target point, find the corresponding positioning longitude error of the single-frame target detection result;

[0019] Step S2.2: Iterate through each pair of targets sequentially to determine whether they are from the same load source;

[0020] Step S2.3: If the loads originate from the same source, project the positions of the two target points being compared onto a planar coordinate system;

[0021] Step S2.4: Calculate the positional error between two target points projected into the planar coordinate system;

[0022] Step S2.5: Determine whether the position error is greater than the threshold;

[0023] Step S2.6: For target point pairs whose position error does not exceed the threshold, update the position of the target point.

[0024] Preferably, step S2.3 includes:

[0025] Step S3.3.1: Project the meridians and parallels on the Earth's surface onto a cylindrical surface, and then unfold the cylindrical surface into a planar map, where the circumference is regarded as the X-axis;

[0026] Step S3.3.2: Convert longitude from degrees to radians;

[0027] x_radian = lon * PI / 180

[0028] Where x_radian represents the corresponding radian value, lon represents the longitude value, and PI represents pi;

[0029] Step S3.3.3: Convert latitude from degrees to radians;

[0030] y_radian = lat * PI / 180

[0031] Where y_radian represents the corresponding radian value, lat represents the latitude value, and PI represents pi;

[0032] Step S3.3.4: Perform Miller projection conversion;

[0033] y_miller=1.25*log(tan(0.25*PI+0.4*y_radian))

[0034] Where PI represents pi;

[0035] Step S3.3.5: Convert the radians into actual distances;

[0036] x=(W / 2)+(W / (2*PI))*x_radian

[0037] y=(H / 2)+(H / (2*mill))*y_radian

[0038] Where W represents the X-axis; PI represents pi; x_radian represents the radian value corresponding to longitude; H represents the Y-axis; mill represents a constant in Miller projection; and y_radian represents the radian value corresponding to latitude.

[0039] Preferably, step S5 employs the following methods:

[0040] Step S5.1: Input a single associated trajectory and predict the position at the next moment based on the position information of the trajectory at the last moment;

[0041] Step S5.2: Calculate the difference between the detected target position and the trajectory prediction position in each direction of the current processing frame;

[0042] Step S5.3: Determine whether the coarse judgment condition of non-association is met; for the detected target points that do not meet the coarse judgment condition of non-association, calculate the error between their position and the trajectory prediction position;

[0043] Step S5.4: Compare the position error with the threshold value and modify the correlation degree corresponding to the matching metric matrix; repeat steps S5.1 to S5.4.

[0044] Preferably, in step S5, the calculated matching metric matrix is ​​distributed in two dimensions to obtain the optimal matching result, using the following method:

[0045]

[0046] The optimal matching solution is

[0047] Where, m ij Represents the matching metric matrix The (i, j)th element; f ij Representing the allocation matrix The (i, j)th element is a binary variable; n1 represents the number of trajectory lines; n2 represents the number of detected target points; (z 1i , z 2j () indicates a successful match between the trajectory and the target.

[0048] Preferably, in step S5, based on the optimal matching result, a Kalman filter-based prediction method is applied to update the currently associated trajectory state, including:

[0049] Step S5.5: Input a single associated trajectory and determine whether the trajectory is successfully matched in the two-dimensional allocation;

[0050] Step S5.6: For a successfully matched trajectory, find its corresponding matching degree value in the matching metric matrix;

[0051] Step S5.7: Determine whether the corresponding matching degree value is greater than the threshold;

[0052] Step S5.8: Perform Kalman filtering on the target location information whose matching degree value does not exceed the threshold to obtain the predicted location information, according to the following formula:

[0053]

[0054] in, This represents the state prediction vector from time k-1 to time k; TN represents the state vector at time k-1; i represents the target number at time k-1; k-1 This represents the number of targets at time k-1; Let represent the estimation error covariance matrix from time k-1 to time k; F represents the posterior estimation error covariance matrix at time k-1; F is the state transition matrix; F T Represents the transpose of the state transition matrix; Q is the error covariance matrix corresponding to the process noise; c i，j This represents the predicted value of the target state; This represents the state vector at time k; σ represents the inverse of the error covariance matrix; T To track and predict the gate; j represents the target number at time k; This represents the number of targets at time k;

[0055] Step S5.9: Update based on the last position status of the trajectory.

[0056] A target tracking system based on multi-source data fusion and correlation provided by the present invention includes:

[0057] Module M1: Acquires multi-payload target detection information at the same time and obtains the target detection result of a single frame; when the detection result is empty, it traverses each existing trajectory to obtain the target state at the last moment of the trajectory, and updates the trajectory set according to the target state at the last moment of the formed trajectory.

[0058] Module M2: When the detection result is not empty, the current detection result is filtered to remove duplicate position information provided by different loads for the same target;

[0059] Module M3: Determines whether an associated trajectory already exists.

[0060] Module M4: If it does not exist, each target point will be used as the starting point of the trajectory, assembled into a defined format and stored.

[0061] Module M5: If it exists, calculate the matching metric matrix between the associated trajectory and the target information detected in the current frame; perform two-dimensional allocation on the calculated matching metric matrix to obtain the optimal matching result; based on the optimal matching result, apply a Kalman filter-based prediction method to update the state of the currently associated trajectory.

[0062] Module M6: Determines whether there are any unassociated target points in the current frame; for associated targets, compares their corresponding values ​​in the matching metric matrix with the threshold value to determine whether to add a new trajectory; otherwise, the unassociated target points are used as the starting points of the new trajectory.

[0063] Module M7: Assembles and stores new trajectory information according to the defined format;

[0064] Module M8: Updates the trajectory set based on the state of each trajectory at the last moment.

[0065] Preferably, the module M2 adopts:

[0066] Module M2.1: The single-frame target detection result is used to find the corresponding positioning longitude error based on the load source of the target point;

[0067] Module M2.2: Iterate through each pair of targets sequentially to determine if they share the same load source;

[0068] Module M2.3: If the loads originate from the same source, the positions of the two target points being compared will be projected onto a planar coordinate system.

[0069] Module M2.4: Calculates the positional error between two target points projected into a planar coordinate system;

[0070] Module M2.5: Determines whether the position error is greater than a threshold;

[0071] Module M2.6: For target point pairs with position errors not exceeding a threshold, update the position of the target points;

[0072] The module M2.3 includes:

[0073] Module M3.3.1: Projects the meridians and parallels of the Earth's surface onto a cylindrical surface, and then unfolds the cylindrical surface into a planar map, where the perimeter is regarded as the X-axis;

[0074] Module M3.3.2: Converts longitude from degrees to radians;

[0075] x_radian = lon * PI / 180

[0076] Where x_radian represents the corresponding radian value, lon represents the longitude value, and PI represents pi;

[0077] Module M3.3.3: Converts latitude from degrees to radians;

[0078] y_radian = lat * PI / 180

[0079] Where y_radian represents the corresponding radian value, lat represents the latitude value, and PI represents pi;

[0080] Module M3.3.4: Performs Miller projection conversion;

[0081] y_miller=1.25*log(tan(0.25*PI+0.4*y_radian))

[0082] Where PI represents pi;

[0083] Module M3.3.5: Converts radians into actual distances;

[0084] x=(W / 2)+(W / (2*PI))*x_radian

[0085] y=(H / 2)+(H / (2*mill))*y_radian

[0086] Where W represents the X-axis; PI represents pi; x_radian represents the radian value corresponding to longitude; H represents the Y-axis; mill represents a constant in Miller projection; and y_radian represents the radian value corresponding to latitude.

[0087] Preferably, the module M5 adopts:

[0088] Module M5.1: Input a single associated trajectory and predict the position at the next moment based on the position information of the trajectory at the last moment;

[0089] Module M5.2: Calculates the difference between the detected target position and the trajectory prediction position in each direction of the current processing frame;

[0090] Module M5.3: Determines whether the coarse condition for non-association is met; for detected target points that do not meet the coarse condition for non-association, calculates the error between their position and the predicted trajectory.

[0091] Module M5.4: Compares the position error with the threshold value and modifies the correlation degree corresponding to the matching metric matrix; repeatedly triggers modules M5.1 to M5.4.

[0092] Preferably, in module M5, the calculated matching metric matrix is ​​distributed in two dimensions to obtain the optimal matching result, using the following method:

[0093]

[0094] The optimal matching solution is

[0095] Where, m ij Represents the matching metric matrix The (i, j)th element; f ij Representing the allocation matrix The (i, j)th element is a binary variable; n1 represents the number of trajectory lines; n2 represents the number of detected target points; (z 1i , z 2j () indicates a successful match between the trajectory and the target;

[0096] In module M5, based on the optimal matching result, a Kalman filter-based prediction method is applied to update the currently associated trajectory state, including:

[0097] Module M5.5: Input a single associated trajectory and determine whether the trajectory is successfully matched in the two-dimensional assignment;

[0098] Module M5.6: For a successfully matched trajectory, find its corresponding matching degree value in the matching metric matrix;

[0099] Module M5.7: Determines whether the corresponding matching degree value is greater than the threshold;

[0100] Module M5.8: For target location information with a matching degree value not exceeding a threshold, Kalman filtering is performed to obtain predicted location information, based on the following formula:

[0101]

[0102] in, This represents the state prediction vector from time k-1 to time k; TN represents the state vector at time k-1; i represents the target number at time k-1; k-1 This represents the number of targets at time k-1; Let represent the estimation error covariance matrix from time k-1 to time k; F represents the posterior estimation error covariance matrix at time k-1; F is the state transition matrix; F T Represents the transpose of the state transition matrix; Q is the error covariance matrix corresponding to the process noise; c i，j This represents the predicted value of the target state; This represents the state vector at time k; σ represents the inverse of the error covariance matrix; T To track and predict the gate; j represents the target number at time k; This represents the number of targets at time k;

[0103] Module M5.9: Updates based on the last position status of the trajectory.

[0104] Compared with the prior art, the present invention has the following beneficial effects:

[0105] 1. This invention projects the position information acquired by the spaceborne sensor into a planar coordinate system, and then fuses and filters the position information of the same target observed by different sensors. This effectively reduces the impact of the sensor's own positioning accuracy error, and avoids the interference of differences in position information acquired by multiple sensors for the same target, thereby further improving the accuracy of target trajectory association.

[0106] 2. This invention will be described in detail through specific technical features and solutions in specific embodiments. Those skilled in the art should be able to understand the beneficial technical effects brought about by these technical features and solutions through the introduction of these technical features and solutions. Attached Figure Description

[0107] Other features, objects, and advantages of the present invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings:

[0108] Figure 1 This is a schematic diagram of the method flow of the present invention.

[0109] Figure 2 This is a schematic diagram of the data association method based on two-dimensional allocation in this invention.

[0110] Figure 3 This is a schematic diagram of the process for calculating the matching metric matrix in this invention.

[0111] Figure 4 This is a flowchart illustrating the trajectory update method based on Kalman filtering in this invention.

[0112] Figure 5 These are the feature diagrams corresponding to the nine types of loads simulated in this invention.

[0113] Figures 6a to 6f This is a set of processing effect diagrams without the addition of missed detection and error data in this invention.

[0114] Figures 7a to 7f This is a diagram illustrating the processing effect of adding random missed detection data in this invention.

[0115] Figures 8a to 8f This is a diagram showing the processing effect of adding random positioning error data in this invention.

[0116] Figures 9a to 9f This is a diagram showing the processing effect of adding false negative rate and random positioning error data in this invention. Detailed Implementation

[0117] The present invention will now be described in detail with reference to specific embodiments. These embodiments will help those skilled in the art to further understand the present invention, but do not limit the invention in any way. It should be noted that those skilled in the art can make several changes and improvements without departing from the concept of the present invention. These all fall within the protection scope of the present invention.

[0118] Example 1

[0119] According to the present invention, a target tracking method based on multi-source data fusion and correlation is provided, such as... Figure 1 As shown, the method specifically includes the following:

[0120] Step S1: Input the single-frame target detection result (i.e., the multi-payload target detection information acquired at the same time) and determine whether the detection result is empty.

[0121] Step S2: If the target detection result of the current frame is empty, update the trajectory set according to the state of the last moment of the formed trajectory.

[0122] Step S3: Filter the non-empty target detection results to remove duplicate position information provided by different payloads for the same target.

[0123] In step S3, the target location information obtained from multiple sensors needs to be filtered. The algorithm flow is as follows: Figure 2 As shown, it includes:

[0124] Step S3.1: Input the single-frame target detection result and find the corresponding positioning longitude error based on the load source of the target point;

[0125] Step S3.2: Iterate through each pair of targets sequentially to determine whether they are from the same load source;

[0126] Step S3.3: If the loads originate from the same source, project the positions of the two target points being compared onto a planar coordinate system;

[0127] Step S3.3 includes the following sub-steps:

[0128] Step S3.3.1: Project the meridians and parallels on the Earth's surface onto a cylindrical surface, and then unfold the cylindrical surface into a planar map, where the circumference is regarded as the X-axis;

[0129] Step S3.3.2: Convert longitude from degrees to radians, according to the following formula:

[0130] x_radian = lon * PI / 180

[0131] Where x_radian represents the corresponding radian value, lon represents the longitude value, and PI represents pi;

[0132] Step S3.3.3: Convert latitude from degrees to radians, according to the following formula:

[0133] y_radian = lat * PI / 180

[0134] Where y_radian represents the corresponding radian value, lat represents the latitude value, and PI represents pi;

[0135] Step S3.3.4: Perform the Miller projection conversion according to the following formula:

[0136] y_miller=1.25*log(tan(0.25*PI+0.4*y_radian))

[0137] Where PI represents pi, and y_radian represents the radian value corresponding to the latitude;

[0138] Step S3.3.5: Convert the radians into actual distances using the following formula:

[0139] x=(W / 2)+(W / (2*PI))*x_radian

[0140] y=(H / 2)+(H / (2*mill))*y_radian

[0141] Where W represents the X-axis (equal to the Earth's circumference), PI represents pi, x_radian represents the radian value corresponding to longitude, H represents the Y-axis (approximately equal to half the Earth's circumference), mill represents a constant in the Miller projection (ranging from approximately ±2.3), and y_radian represents the radian value corresponding to latitude.

[0142] Step S3.4: Calculate the positional error between two target points projected into the planar coordinate system;

[0143] Step S3.5: Determine whether the position error is greater than the threshold;

[0144] Step S3.6: For target point pairs with position errors not exceeding the threshold, update the position of the target points;

[0145] Step S3.7: Continue processing the target detection results of the next frame for iterative updates.

[0146] Step S4: Determine if there is currently an associated trajectory.

[0147] Step S5: If it does not exist, use each target point as the starting point of the trajectory, assemble it into the defined format and store it; otherwise, continue with the subsequent steps.

[0148] Step S6: Calculate the matching metric matrix between the associated trajectory and the target information detected in the current frame.

[0149] The steps for calculating the matching metric matrix in step S6 are as follows: Figure 3 As shown, it includes:

[0150] Step S6.1: Input a single associated trajectory and predict the position at the next moment based on the position information of the trajectory at the last moment;

[0151] Step S6.2: Calculate the difference between the detected target position and the trajectory prediction position in each direction of the current processing frame;

[0152] Step S6.3: Determine whether the coarse condition for non-association is met;

[0153] Step S6.4: For detected target points that do not meet the coarse judgment condition of non-association, calculate the error between their position and the predicted trajectory.

[0154] Step S6.5: Compare the position error with the threshold value and modify the correlation degree corresponding to the matching metric matrix;

[0155] Step S6.6: Continue processing the next formed trajectory and iteratively update it.

[0156] Step S7: Perform two-dimensional allocation based on the calculated matching metric matrix to obtain the optimal matching result.

[0157] In step S7, the two-dimensional allocation is performed based on the calculated matching metric matrix according to the following formula:

[0158]

[0159] The optimal matching solution is

[0160] Where, m ij Represents the matching metric matrix The (i, j)th element; f ij Representing the allocation matrix The (i, j)th element is a binary variable; n1 represents the number of trajectory lines; n2 represents the number of detected target points; (z 1i , z 2j () indicates a successful match between the trajectory and the target.

[0161] Step S8: Based on the matching results, apply a Kalman filter-based prediction method to update the current associated trajectory state.

[0162] The algorithm steps in step S8 are as follows: Figure 4 As shown, it includes:

[0163] Step S8.1: Input a single associated trajectory and determine whether the trajectory is successfully matched in the two-dimensional allocation;

[0164] Step S8.2: For a successfully matched trajectory, find its corresponding matching degree value in the matching metric matrix;

[0165] Step S8.3: Determine whether the corresponding matching degree value is greater than the threshold;

[0166] Step S8.4: Perform Kalman filtering on the target location information whose matching degree value does not exceed the threshold to obtain the predicted location information, according to the following formula:

[0167]

[0168] in, This represents the state prediction vector from time k-1 to time k; TN represents the state vector at time k-1; i represents the target number at time k-1; k-1 This represents the number of targets at time k-1; Let represent the estimation error covariance matrix from time k-1 to time k; F represents the posterior estimation error covariance matrix at time k-1; F is the state transition matrix; F T Represents the transpose of the state transition matrix; Q is the error covariance matrix corresponding to the process noise; c i，j This represents the predicted value of the target state; This represents the state vector at time k; σ represents the inverse of the error covariance matrix; T To track and predict the gate; j represents the target number at time k; This represents the number of targets at time k;

[0169] Step S8.5: Update the trajectory based on its final position status;

[0170] Step S8.6: Continue processing the next formed trajectory and iteratively update it.

[0171] Step S9: Determine whether there are any unassociated target points in the current frame.

[0172] Step S10: For the associated target, compare its corresponding value in the matching metric matrix with the threshold value to determine whether to add a new trajectory; otherwise, take the unassociated target point as the starting point of the new trajectory.

[0173] Step S11: Assemble and store the new trajectory information according to the defined format.

[0174] Step S12: Update the trajectory set according to the state of each trajectory at the last moment.

[0175] Step S13: Wait for the target detection results of the next frame to be processed iteratively and the trajectory set is updated.

[0176] A target tracking system based on multi-source data fusion and correlation provided by the present invention includes:

[0177] Module M1: Input the target detection result of a single frame (i.e., the target detection information of multiple payloads acquired at the same time) and determine whether the detection result is empty.

[0178] Module M2: If the target detection result of the current frame is empty, update the trajectory set according to the state of the last moment of the formed trajectory.

[0179] Module M3: Filters the detection results of non-empty targets, removing duplicate position information provided by different payloads for the same target.

[0180] This module M3 includes:

[0181] Module M3.1: Input single-frame target detection results and find the corresponding positioning longitude error based on the load source of the target point;

[0182] Module M3.2: Iterate through each pair of targets sequentially to determine if they share the same load source;

[0183] Module M3.3: If the loads originate from the same source, the positions of the two target points being compared will be projected onto a planar coordinate system.

[0184] The calculation steps for module M3.3 include:

[0185] Module M3.3.1: Projects the meridians and parallels of the Earth's surface onto a cylindrical surface, and then unfolds the cylindrical surface into a planar map, where the perimeter is regarded as the X-axis;

[0186] Module M3.3.2: Converts longitude from degrees to radians;

[0187] Module M3.3.3: Converts dimensions from degrees to radians;

[0188] Module M3.3.4: Performs Miller projection conversion;

[0189] Module M3.3.5: Converts radians into actual distances.

[0190] Module M3.4: Calculates the positional error between two target points projected into a planar coordinate system;

[0191] Module M3.5: Determines whether the position error is greater than a threshold;

[0192] Module M3.6: For target point pairs with position errors not exceeding a threshold, update the position of the target points;

[0193] Module M3.7: Continue processing the target detection results of the next frame for iterative updates.

[0194] Module M4: Determines whether there is a currently associated trajectory.

[0195] Module M5: If it does not exist, each target point is used as the starting point of the trajectory, assembled into a defined format and stored; otherwise, continue with the subsequent steps.

[0196] Module M6: Calculates the matching metric matrix between the associated trajectory and the target information detected in the current frame.

[0197] The steps for calculating the matching metric matrix in module M6 include:

[0198] Module M6.1: Input a single associated trajectory and predict the position at the next moment based on the position information of the trajectory at the last moment;

[0199] Module M6.2: Calculates the difference between the detected target position and the trajectory prediction position in each direction of the current processing frame;

[0200] Module M6.3: Determines whether the coarse condition for non-association is met;

[0201] Module M6.4: For detected target points that do not meet the coarse judgment condition of non-association, calculate the error between their position and the trajectory prediction position;

[0202] Module M6.5: Compares the position error with the threshold value and modifies the correlation degree corresponding to the matching metric matrix;

[0203] Module M6.6: Continue processing the next formed trajectory and iteratively update it.

[0204] Module M7: Performs two-dimensional allocation based on the calculated matching metric matrix to obtain the optimal matching result.

[0205] In module M7, the two-dimensional allocation is performed based on the calculated matching metric matrix according to the following formula:

[0206]

[0207] The optimal matching solution is

[0208] Where, m ij Represents the matching metric matrix The (i, j)th element; f ij Representing the allocation matrix The (i, j)th element is a binary variable; n1 represents the number of trajectory lines; n2 represents the number of detected target points; (z 1i , z 2j () indicates a successful match between the trajectory and the target.

[0209] Module M8: Based on the matching results, it applies a Kalman filter-based prediction method to update the state of the currently associated trajectories.

[0210] The algorithm steps in module M8 include:

[0211] Module M8.1: Input a single associated trajectory and determine whether the trajectory is successfully matched in the two-dimensional assignment;

[0212] Module M8.2: For a successfully matched trajectory, find its corresponding matching degree value in the matching metric matrix;

[0213] Module M8.3: Determines whether the corresponding matching degree value is greater than the threshold;

[0214] Module M8.4: For target location information with a matching degree value not exceeding a threshold, Kalman filtering is performed to obtain predicted location information, based on the following formula:

[0215]

[0216]

[0217]

[0218] in, This represents the state prediction vector from time k-1 to time k; TN represents the state vector at time k-1; i represents the target number at time k-1; k-1 This represents the number of targets at time k-1; Let represent the estimation error covariance matrix from time k-1 to time k; F represents the posterior estimation error covariance matrix at time k-1; F is the state transition matrix; F T Represents the transpose of the state transition matrix; Q is the error covariance matrix corresponding to the process noise; c i，j This represents the predicted value of the target state; This represents the state vector at time k; σ represents the inverse of the error covariance matrix; T To track and predict the gate; j represents the target number at time k; This represents the number of targets at time k;

[0219] Module M8.5: Updates based on the final position status of the trajectory;

[0220] Module M8.6: Continue processing the next formed trajectory and iteratively update it.

[0221] Module M9: Determines whether there are any unassociated target points in the current frame.

[0222] Module M10: For the associated target, compare its corresponding value in the matching metric matrix with the threshold to determine whether to add a new trajectory; otherwise, use the unassociated target point as the starting point of the new trajectory.

[0223] Module M11: Assembles and stores new trajectory information according to the defined format.

[0224] Module M12: Updates the trajectory set based on the state of each trajectory at the last moment.

[0225] Module M13: Wait for the target detection result of the next frame to perform iterative processing and update the trajectory set. Next, the present invention will be described in more detail.

[0226] The effects of this invention can be further illustrated by the following simulation results.

[0227] 1. Simulation data:

[0228] The position information sequences of 22 moving ship targets at sea were observed using 9 loads simulated by STK, and the data processing was completed by simulation on the PyCharm 2017 platform.

[0229] 2. Simulation content and results:

[0230] To verify the effectiveness of the target tracking method based on multi-source data fusion and correlation provided by this invention, a test was conducted using position information sequences of 22 moving ship targets observed by 9 different payloads over a period of 23 minutes. The original experimental images and processed results from four different scenarios are compared below.

[0231] Figure 5 The data characteristics obtained from the nine observation payloads in the STK simulation include payload ID number, positioning accuracy error range (in kilometers), and detection rate.

[0232] Figures 6a to 6f The images show the processing results for a set of Scene 1 (original target points without added omissions and errors). Figures 6a to 6c This diagram illustrates the distribution of the actual target points, where... Figure 6a A schematic diagram showing the distribution of all targets. Figure 6b and Figure 6c They represent Figure 6a Enlarged schematic diagram of two dense regions in the middle; Figures 6d to 6f This represents the effect diagram of the target associated trajectory, where Figure 6d This represents the trajectory diagram formed by all the connections. Figure 6e and Figure 6f They represent Figure 6d Enlarged schematic diagram of two dense regions.

[0233] Figures 7a to 7f The image shows the processing results for Scenario 2 (adding random missed target points based on the detection rate). Figures 7a to 7c This diagram illustrates the distribution of the actual target points, where... Figure 7a A schematic diagram showing the distribution of all targets. Figure 7b and Figure 7c They represent Figure 7a Enlarged schematic diagram of two dense regions in the middle; Figures 7d to 7f This represents the effect diagram of the target associated trajectory, where, Figure 7d This represents the trajectory diagram formed by all the connections. Figure 7e and Figure 7f They represent Figure 7d Enlarged schematic diagram of two dense regions.

[0234] Figures 8a to 8f The image shows the processing results for Scenario 3 (adding target points with random errors within three kilometers based on positioning accuracy errors). Figures 8a to 8c This diagram illustrates the distribution of actual target points, where... Figure 8a A schematic diagram showing the distribution of all targets. Figure 8b and Figure 8c They represent Figure 8a Enlarged schematic diagram of two dense regions in the middle; Figures 8d to 8fThis represents the effect diagram of the target associated trajectory, where, Figure 8d This represents the trajectory diagram formed by all the connections. Figure 8e and Figure 8f They represent Figure 8d Enlarged schematic diagram of two dense regions.

[0235] Figures 9a to 9f The images show the processing results for Scenario 4 (adding target points with random missed detections and random errors within three kilometers). Figures 9a to 9c This diagram illustrates the distribution of actual target points, where... Figure 9a A schematic diagram showing the distribution of all targets. Figure 8b and Figure 8c They represent Figure 9a Enlarged schematic diagram of two dense regions in the middle; Figures 9d to 9f This represents the effect diagram of the target associated trajectory, where Figure 9d This represents the trajectory diagram formed by all the connections. Figure 9e and Figure 9f They represent Figure 9d Enlarged schematic diagram of two dense regions.

[0236] Those skilled in the art will understand that, besides implementing the system and its various devices, modules, and units provided by this invention in the form of purely computer-readable program code, the same functions can be achieved entirely through logical programming of the method steps, making the system and its various devices, modules, and units of this invention function in the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, and embedded microcontrollers. Therefore, the system and its various devices, modules, and units provided by this invention can be considered as a hardware component, and the devices, modules, and units included therein for implementing various functions can also be considered as structures within the hardware component; alternatively, the devices, modules, and units for implementing various functions can be considered as both software modules implementing the method and structures within the hardware component.

[0237] Specific embodiments of the present invention have been described above. It should be understood that the present invention is not limited to the specific embodiments described above, and those skilled in the art can make various changes or modifications within the scope of the claims, which do not affect the essence of the present invention. Unless otherwise specified, the embodiments and features described in this application can be arbitrarily combined with each other.

Claims

1. A target tracking method based on multi-source data fusion and correlation, characterized in that, include: Step S1: Obtain multi-payload target detection information at the same time to obtain the target detection result of a single frame; If the detection result is empty, then traverse each existing trajectory to obtain the target state at the last moment of the trajectory, and update the trajectory set according to the target state at the last moment of the formed trajectory. Step S2: If the detection result is not empty, filter the current detection result to remove duplicate position information provided by different loads for the same target; Step S3: Determine if there is currently an associated trajectory; Step S4: If it does not exist, then each target point is used as the starting point of the trajectory, assembled into a defined format and stored; Step S5: If it exists, calculate the matching metric matrix between the associated trajectory and the target information detected in the current frame; The calculated matching metric matrix is ​​distributed in two dimensions to obtain the optimal matching result; based on the optimal matching result, the prediction method based on Kalman filtering is applied to update the state of the currently associated trajectories. Step S6: Determine whether there are any unassociated target points in the current frame; for associated targets, compare their corresponding values ​​in the matching metric matrix with the threshold value to determine whether to add a new trajectory; otherwise, use the unassociated target points as the starting points of the new trajectory. Step S7: Assemble and store the new trajectory information according to the defined format; Step S8: Update the trajectory set according to the state of each trajectory at the last moment; Step S2 employs the following: Step S2.1: Based on the load source of the target point, find the corresponding positioning longitude error of the single-frame target detection result; Step S2.2: Iterate through each pair of targets sequentially to determine whether they are from the same load source; Step S2.3: If the loads originate from the same source, project the positions of the two target points being compared onto a planar coordinate system; Step S2.4: Calculate the positional error between two target points projected into the planar coordinate system; Step S2.5: Determine whether the position error is greater than the threshold; Step S2.6: For target point pairs whose position error does not exceed the threshold, update the position of the target point.

2. The target tracking method based on multi-source data fusion and correlation according to claim 1, characterized in that, Step S2.3 includes: Step S2.3.1: Project the meridians and parallels on the Earth's surface onto a cylindrical surface, and then unfold the cylindrical surface into a planar map, where the circumference is regarded as the X-axis; Step S2.3.2: Convert longitude from degrees to radians; in, This indicates the corresponding radian value. Indicates the longitude value. Represents pi; Step S2.3.3: Convert latitude from degrees to radians; in, This indicates the corresponding radian value. Indicates latitude value, Represents pi; Step S2.3.4: Perform Miller projection conversion; in, Represents pi; Step S2.3.5: Convert the radians into actual distances; in, express axis; Represents pi; This represents the radian value corresponding to longitude. express axis; This represents a constant in the Miller projection. This represents the radian value corresponding to latitude.

3. The target tracking method based on multi-source data fusion and correlation according to claim 1, characterized in that, Step S5 employs the following: Step S5.1: Input a single associated trajectory and predict the position at the next moment based on the position information of the trajectory at the last moment; Step S5.2: Calculate the difference between the detected target position and the trajectory prediction position in each direction of the current processing frame; Step S5.3: Determine whether the coarse judgment condition of non-association is met; for the detected target points that do not meet the coarse judgment condition of non-association, calculate the error between their position and the trajectory prediction position; Step S5.4: Compare the position error with the threshold value, and modify the correlation degree corresponding to the matching metric matrix; Repeat steps S5.1 to S5.

4.

4. The target tracking method based on multi-source data fusion and correlation according to claim 1, characterized in that, In step S5, the calculated matching metric matrix is ​​distributed in two dimensions to obtain the optimal matching result, using the following method: The optimal matching solution is ; in, Represents the matching metric matrix The One element; Representing the allocation matrix The Each element is a binary variable; Indicates the number of trajectories; Indicates the number of target points detected; This indicates a successful match between the trajectory and the target.

5. The target tracking method based on multi-source data fusion and correlation according to claim 1, characterized in that, In step S5, based on the optimal matching result, a Kalman filter-based prediction method is applied to update the currently associated trajectory state, including: Step S5.5: Input a single associated trajectory and determine whether the trajectory is successfully matched in the two-dimensional allocation; Step S5.6: For a successfully matched trajectory, find its corresponding matching degree value in the matching metric matrix; Step S5.7: Determine whether the corresponding matching degree value is greater than the threshold; Step S5.8: Perform Kalman filtering on the target location information whose matching degree value does not exceed the threshold to obtain the predicted location information, according to the following formula: in, express Always The state prediction vector at time t; express The state vector at any given time; express The target number at that moment; express Number of targets at any given time; express Time's up The covariance matrix of the estimation error at time; express The posterior estimation error covariance matrix at time t; This is the state transition matrix; This represents the transpose of the state transition matrix; The error covariance matrix corresponding to the process noise; This represents the predicted value of the target state; express The state vector at any given time; Represents the inverse of the error covariance matrix; To track and predict the gate; express The target number at that moment; express Number of targets at any given time; Step S5.9: Update based on the last position status of the trajectory.

6. A target tracking system based on multi-source data fusion and correlation, characterized in that, include: Module M1: Acquires multi-payload target detection information at the same time and obtains the corresponding target detection result for a single frame; If the detection result is empty, then traverse each existing trajectory to obtain the target state at the last moment of the trajectory, and update the trajectory set according to the target state at the last moment of the formed trajectory. Module M2: When the detection result is not empty, the current detection result is filtered to remove duplicate position information provided by different loads for the same target; Module M3: Determines whether an associated trajectory already exists. Module M4: If it does not exist, each target point will be used as the starting point of the trajectory, assembled into a defined format and stored. Module M5: If it exists, calculate the matching metric matrix between the associated trajectory and the target information detected in the current frame; The calculated matching metric matrix is ​​distributed in two dimensions to obtain the optimal matching result; based on the optimal matching result, the prediction method based on Kalman filtering is applied to update the state of the currently associated trajectories. Module M6: Determines whether there are any unassociated target points in the current frame; for associated targets, compares their corresponding values ​​in the matching metric matrix with the threshold value to determine whether to add a new trajectory; otherwise, the unassociated target points are used as the starting points of the new trajectory. Module M7: Assembles and stores new trajectory information according to the defined format; Module M8: Updates the trajectory set based on the final state of each trajectory; The module M2 adopts: Module M2.1: The single-frame target detection result is used to find the corresponding positioning longitude error based on the load source of the target point; Module M2.2: Iterate through each pair of targets sequentially to determine if they share the same load source; Module M2.3: If the loads originate from the same source, the positions of the two target points being compared will be projected onto a planar coordinate system. Module M2.4: Calculates the positional error between two target points projected into a planar coordinate system; Module M2.5: Determines whether the position error is greater than a threshold; Module M2.6: For target point pairs with position errors not exceeding a threshold, update the position of the target points; The module M2.3 includes: Module M2.3.1: Projects the meridians and parallels of the Earth's surface onto a cylindrical surface, and then unfolds the cylindrical surface into a planar map, where the perimeter is regarded as the X-axis; Module M2.3.2: Converts longitude from degrees to radians; in, This indicates the corresponding radian value. Indicates the longitude value. Represents pi; Module M2.3.3: Converts latitude from degrees to radians; in, This indicates the corresponding radian value. Indicates latitude value, Represents pi; Module M2.3.4: Performs Miller projection conversion; in, Represents pi; Module M2.3.5: Converts radians into actual distances; in, express axis; Represents pi; This represents the radian value corresponding to longitude. express axis; This represents a constant in the Miller projection. This represents the radian value corresponding to latitude.

7. The target tracking system based on multi-source data fusion and correlation according to claim 6, characterized in that, The module M5 adopts: Module M5.1: Input a single associated trajectory and predict the position at the next moment based on the position information of the trajectory at the last moment; Module M5.2: Calculates the difference between the detected target position and the trajectory prediction position in each direction of the current processing frame; Module M5.3: Determines whether the coarse condition for non-association is met; for detected target points that do not meet the coarse condition for non-association, calculates the error between their position and the predicted trajectory. Module M5.4: Compares the position error with the threshold value and modifies the correlation degree corresponding to the matching metric matrix; Repeatedly trigger modules M5.1 through M5.

4.

8. The target tracking system based on multi-source data fusion and correlation according to claim 6, characterized in that, In module M5, the calculated matching metric matrix is ​​distributed in two dimensions to obtain the optimal matching result, using the following method: The optimal matching solution is ; in, Represents the matching metric matrix The One element; Representing the allocation matrix The Each element is a binary variable; Indicates the number of trajectories; Indicates the number of target points detected; This indicates a successful match between the trajectory and the target. In module M5, based on the optimal matching result, a Kalman filter-based prediction method is applied to update the currently associated trajectory state, including: Module M5.5: Input a single associated trajectory and determine whether the trajectory is successfully matched in the two-dimensional assignment; Module M5.6: For a successfully matched trajectory, find its corresponding matching degree value in the matching metric matrix; Module M5.7: Determines whether the corresponding matching degree value is greater than the threshold; Module M5.8: For target location information with a matching degree value not exceeding a threshold, Kalman filtering is performed to obtain predicted location information, based on the following formula: in, express Always The state prediction vector at time t; express The state vector at any given time; express The target number at that moment; express Number of targets at any given time; express Time's up The covariance matrix of the estimation error at time; express The posterior estimation error covariance matrix at time t; This is the state transition matrix; This represents the transpose of the state transition matrix; The error covariance matrix corresponding to the process noise; This represents the predicted value of the target state; express The state vector at any given time; Represents the inverse of the error covariance matrix; To track and predict the gate; express The target number at that moment; express Number of targets at any given time; Module M5.9: Updates based on the last position status of the trajectory.

Images