Aerial target tracking method based on stationary orbit staring satellite remote sensing image
By applying the Kalman filtering model and the Marshall distance correlation algorithm in the gaze satellite remote sensing image of the stationary orbit, the problem of difficulty in trajectory correlation in aerial target tracking is solved, and efficient tracking of air targets with strong maneuverability is achieved.
Patent Information
- Application Number
- CN202411603029.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-11
- Publication Date
- 2025-04-01
AI Technical Summary
It is difficult for the prior art to effectively track aerial targets with strong maneuverability, especially when the remote sensing image resolution of satellites is low in stationary orbit gaze satellites, and traditional tracking algorithms are difficult to achieve accurate trajectory correlation.
The aerial target tracking method based on the Kalman filtering model is adopted. By obtaining the trajectory list of the aerial target and predicting the current moment, combining the aerial target detection results of the ground processing, the Mahayana distance is used to correlate the target with the trajectory to achieve the establishment and tracking of the trajectory of the aerial target.
It improves the accuracy and reliability of air target tracking, can effectively process the fast motion and low-resolution remote sensing images of air targets, and achieves effective tracking of air targets with strong maneuverability.
Smart Images
Figure CN119741344A_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the technical field of aerial target tracking, and in particular to an aerial target tracking method based on geostationary orbit staring satellite remote sensing images. Background Art
[0002] With the development of my country's geostationary orbit staring remote sensing satellite technology and capabilities, continuous detection and tracking of aerial targets has become possible. At present, most of the methods for tracking targets based on geostationary orbit staring satellite remote sensing images are aimed at targets with weak mobility, and mainly use the detection results of visible spectrum remote sensing images with relatively good resolution. The target movement direction can be estimated through the detection results of a single scene image. For aerial target tracking, since the resolution of space-based infrared spectrum remote sensing images used to detect aerial targets is low, generally a target corresponds to only one point in a single scene image, and aerial targets are more mobile than ship targets, traditional tracking algorithms are difficult to track well, and detection results are difficult to accurately associate with trajectories. At present, there is no aerial target tracking algorithm for geostationary orbit staring satellite remote sensing images.
[0003] Therefore, it is urgent to design a new aerial target tracking method based on geostationary orbit staring satellite remote sensing images. Summary of the invention
[0004] In order to solve the technical problems existing in the above-mentioned prior art, the purpose of the present invention is to provide an aerial target tracking method based on geostationary orbit staring satellite remote sensing images, which can achieve trajectory establishment and tracking of aerial targets.
[0005] To achieve the above-mentioned object of the invention, the present invention provides an aerial target tracking method based on geostationary orbit staring satellite remote sensing images, comprising the following steps:
[0006] Step S1, obtaining a trajectory list of all aerial targets, and predicting all trajectories in the trajectory list to the current moment;
[0007] Step S2, obtaining ground processing to form an aerial target detection result, and associating the updated trajectory with the obtained aerial target detection result;
[0008] Step S3: Process the trajectory with successful target association and the trajectory with failed target association separately.
[0009] According to a technical solution of the present invention, the step S1 specifically includes:
[0010] Step S11: Obtaining a list of aerial target trajectories Tra represents the current air target trajectory list set, T nrepresents a certain trajectory, N represents the number of all aerial target trajectories at the current moment, the current moment is moment k, and the current state information of the trajectory is the state of the previous frame, i.e. moment k-1, i.e. The state is (x k-1 ,vx k-1 ,y k-1 ,vy k-1 ) T , where x k-1 ,y k-1 are the longitude and latitude at time k-1, vx k-1 ,vy k-1 are the velocity components in longitude and latitude, respectively. is the covariance matrix of trajectory k-1;
[0011] Step S12: According to the Kalman filter model, all trajectory states are predicted to the current k moment:
[0012]
[0013] Among them, f k-1 is the state transfer function;
[0014] According to the mid-latitude navigation method, under the constant direction and constant speed model, the longitude and latitude update formula is as follows:
[0015]
[0016] Among them, t is the time difference between time k-1 and time k, and a is the number of kilometers corresponding to the arc length of 1° longitude at the equator, then:
[0017]
[0018] Step S13: predict all trajectory covariance matrices to the current k moment.
[0019] According to a technical solution of the present invention, in step S13, based on step S12, the state transfer function f k-1 The state transfer matrix F is nonlinear and is used to update the covariance matrix k-1 is the state transition function f k-1 The Jacobian matrix of is:
[0020]
[0021] in, Then the covariance matrix update formula is:
[0022]
[0023] Among them, Q k-1is the system process noise matrix, and the process noise influencing factors are set based on experience.
[0024] According to a technical solution of the present invention, the step S2 specifically includes:
[0025] Step S21: Obtain all aerial target detection results at the current k moment Tar represents the set of all test results at time k, Z m represents one of the measurement results, namely lon k ,lat k Respectively represent the latitude and longitude results of the target measurement;
[0026] Step S22: for each predicted trajectory in the Tra set With each measurement result Z in the Tar set m Make an association.
[0027] According to a technical solution of the present invention, the step S22 specifically includes:
[0028] Step S221, using Mahalanobis distance as the basis for association, that is, the target with the closest Mahalanobis distance of each trajectory is used as the associated target;
[0029] According to the trajectory Updated covariance matrix Extract the trajectory latitude and longitude covariance matrix:
[0030]
[0031] According to Cov(x,y) and each measurement point Z m Coordinates of the trajectory and calculate the Mahalanobis distance between each measurement point:
[0032]
[0033] At the same time, set the target association maximum distance threshold D max , if a detection point Z m With trajectory The Mahalanobis distance is the shortest among all the detection points and satisfies It is considered that the detection point Z m With trajectory The association is successful;
[0034] If the distance between a trajectory and all detection points is greater than D max , then the corresponding trajectory association is considered to have failed;
[0035] If the distance between a detection point and all trajectories is greater than D max, then the corresponding detection point is considered to have no associated trajectory.
[0036] According to a technical solution of the present invention, step S3 specifically includes:
[0037] Update the status of the trajectory that is successfully associated with the target, and count the number of successful trajectory associations. The specific update steps are as follows:
[0038] According to the test results The observation matrix The measurement residual is then:
[0039]
[0040] According to the Kalman filter formula, the new information covariance S k for:
[0041]
[0042] Among them, R k is the measurement noise matrix. Since there are only longitude and latitude for the target detection result, R k Generally expressed as:
[0043] Then the Kalman gain is:
[0044]
[0045] The trajectory state is updated as follows:
[0046]
[0047] The trajectory covariance matrix is updated as:
[0048]
[0049] According to a technical solution of the present invention, for a target that is not associated with any trajectory, a new trajectory is established, and the initial state of the new trajectory is set as: T new =(lon k ,vx k-1 ,lat k ,vy k-1 ) T , the initial covariance matrix is set as: Among them, V max The maximum speed at which an aerial target can fly.
[0050] According to a technical solution of the present invention, for the trajectory of failed association, the number of consecutive association failures is recorded;
[0051] Based on the preset continuous association failure number threshold as the trajectory deletion judgment threshold judgment condition, the trajectories that exceed the continuous association failure number threshold are deleted, and the undeleted trajectories are used as the trajectories for the next round of trajectory prediction association.
[0052] According to a technical solution of the present invention, based on a preset continuous association success number threshold as a trajectory output judgment threshold judgment condition, the trajectory that meets the continuous association success number threshold is output.
[0053] Compared with the prior art, the present invention has the following beneficial effects:
[0054] The present invention proposes an aerial target tracking method based on geostationary orbit staring satellite remote sensing images. Aiming at the characteristics of small number of aerial target points and fast target speed detected by geostationary orbit staring satellite, the method establishes and tracks the aerial target by modeling the aerial target motion, designs a target and trajectory association algorithm, and builds a target tracking framework based on the extended Kalman filter tracking model.
[0055] The present invention adopts Mahalanobis distance to match the target and the trajectory, thereby improving the accuracy and reliability of target association. BRIEF DESCRIPTION OF THE DRAWINGS
[0056] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention, and for ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.
[0057] Figure 1 A schematic diagram schematically shows a flow chart of an aerial target tracking method based on geostationary orbit staring satellite remote sensing images according to an embodiment of the present invention;
[0058] Figure 2 The overall flow chart of an aerial target tracking method based on geostationary orbit staring satellite remote sensing images according to one embodiment of the present invention is schematically shown. DETAILED DESCRIPTION
[0059] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0060] like Figure 1and Figure 2 As shown, the present invention provides an aerial target tracking method based on geostationary orbit staring satellite remote sensing images, comprising the following steps:
[0061] Step S1, obtaining a trajectory list of all aerial targets, including all trajectories of all aerial targets, and predicting all trajectories in the trajectory list to the current moment;
[0062] Step S2, obtaining ground processing to form an aerial target detection result, and associating the updated trajectory with the obtained aerial target detection result;
[0063] Step S3: processing the trajectory with successful target association and the trajectory with failed target association separately, specifically including:
[0064] The trajectory state of the successfully associated trajectory is estimated based on Kalman filtering, and the number of failed associations is counted for the trajectory that failed to be associated; the target that is not associated with any trajectory is created as a new trajectory;
[0065] Determine whether the trajectory state meets the output conditions, and output the trajectory that meets the output conditions. If it does not meet the output conditions, determine whether it meets the trajectory deletion conditions, and delete the trajectory that meets the conditions.
[0066] In some embodiments of the present invention, the step S1 specifically includes:
[0067] Step S11: Obtaining a list of aerial target trajectories Tra represents the current air target trajectory list set, T n represents a certain trajectory, N represents the number of all aerial target trajectories at the current moment, assuming that the current moment is moment k, then the current state information of the trajectory is the state of the previous frame, i.e. moment k-1, that is, The state is (x k-1 ,vx k-1 ,y k-1 ,vy k-1 ) T , where x k-1 ,y k-1 are the longitude and latitude at time k-1, vx k-1 ,vy k-1 are the velocity components in longitude and latitude, respectively. is the covariance matrix of trajectory k-1;
[0068] Step S12: According to the Kalman filter model, all trajectory states are predicted to the current k moment:
[0069]
[0070] Among them, f k-1is the state transfer function;
[0071] According to the mid-latitude navigation method, under the constant direction and constant speed model, the longitude and latitude update formula is as follows:
[0072]
[0073]
[0074] Among them, t is the time difference between time k-1 and time k, a is the number of kilometers corresponding to the arc length of 1° longitude at the equator, a=2π(6378+h) / 360 kilometers, h is the trajectory height, then:
[0075]
[0076] Step S13: predict all trajectory covariance matrices to the current k moment.
[0077] In some embodiments of the present invention, in step S13, based on step S12, the state transfer function f k-1 The state transfer matrix F is nonlinear and is used to update the covariance matrix k-1 is the state transition function f k-1 The Jacobian matrix of is:
[0078]
[0079] in, Then the covariance matrix update formula is:
[0080]
[0081] Among them, Q k-1 is the system process noise matrix, and the process noise influencing factors are set based on experience.
[0082] In some embodiments of the present invention, the step S2 specifically includes:
[0083] Step S21: Obtain all aerial target detection results at the current k moment Tar represents the set of all test results at time k, Z m represents one of the measurement results, namely lon k ,lat k Respectively represent the latitude and longitude results of the target measurement;
[0084] Step S22: for each predicted trajectory in the Tra set (the current aerial target trajectory list set), With each measurement result Z in the Tar setm Make an association.
[0085] In some embodiments of the present invention, the step S22 specifically includes:
[0086] Step S221: Since the aerial target moves at a high speed, the Mahalanobis distance is used as the basis for association when associating the targets, that is, the target with the closest Mahalanobis distance of each trajectory is used as the associated target;
[0087] According to the trajectory Updated covariance matrix Extract the trajectory latitude and longitude covariance matrix:
[0088]
[0089] According to Cov(x,y) and each measurement point Z m Coordinates of the trajectory and calculate the Mahalanobis distance between each measurement point:
[0090]
[0091] At the same time, set the target association maximum distance threshold D max , if a detection point Z m With trajectory The Mahalanobis distance is the shortest among all the detection points and satisfies It is considered that the detection point Z m With trajectory The association is successful;
[0092] If the distance between a trajectory and all detection points is greater than D max , then the corresponding trajectory association is considered to have failed;
[0093] If the distance between a detection point and all trajectories is greater than D max , then the corresponding detection point is considered to have no associated trajectory.
[0094] In some embodiments of the present invention, the step S3 specifically includes:
[0095] Update the status of the trajectory that is successfully associated with the target, and count the number of successful trajectory associations. The specific update steps are as follows:
[0096] According to the test results The observation matrix The measurement residual is then:
[0097]
[0098] According to the Kalman filter formula, the new information covariance S k for:
[0099]
[0100] Among them, R k is the measurement noise matrix. Since there are only longitude and latitude for the target detection result, R k Generally expressed as:
[0101] Then the Kalman gain is:
[0102]
[0103] The trajectory state is updated as follows:
[0104]
[0105] The trajectory covariance matrix is updated as:
[0106]
[0107] In some embodiments of the present invention, for a target that is not associated with any trajectory, a new trajectory is established, and the initial state of the new trajectory is set as: new =(lon k ,vx k-1 ,lat k ,vy k-1 ) T , the initial covariance matrix is set as: Among them, V max The maximum speed at which an aerial target can fly.
[0108] In some embodiments of the present invention, for the trajectory of failed association, the number of consecutive association failures is recorded;
[0109] Based on the preset continuous association failure number threshold as the trajectory deletion judgment threshold judgment condition, the trajectories that exceed the continuous association failure number threshold are deleted, and the undeleted trajectories are used as the trajectories for the next round of trajectory prediction association.
[0110] In some embodiments of the present invention, based on a preset threshold of consecutive successful association times as a trajectory output judgment threshold judgment condition, the trajectory that meets the threshold of consecutive successful association times is output.
[0111] The present invention provides an aerial target tracking method based on geostationary orbit staring satellite remote sensing images, comprising: obtaining a trajectory list of aerial targets, the trajectory list including all trajectories of the aerial targets, and predicting all trajectories in the trajectory list to the current moment; obtaining ground processing to form an aerial target detection result, and associating the updated trajectory with the acquired aerial target detection result; and processing the trajectory with successful target association and the trajectory with failed target association separately. In view of the characteristics of the small number of aerial target points and fast target speed detected by geostationary orbit staring satellites, the present invention models the motion of aerial targets, designs a target and trajectory association algorithm, and constructs a target tracking framework based on an extended Kalman filter tracking model to achieve trajectory establishment and tracking of aerial targets.
[0112] Furthermore, the present invention uses Mahalanobis distance to match the target and the trajectory, thereby improving the accuracy and reliability of target association.
[0113] In the present invention, the terms "include", "comprises" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or terminal device including a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, method, article or terminal device. In the absence of further restrictions, an element defined by the sentence "comprises a ..." does not exclude the existence of other identical elements in the process, method, article or terminal device including the element.
[0114] Finally, it should be noted that the above is a preferred embodiment of the present invention. It should be pointed out that although the preferred embodiment of the present invention has been described, for those skilled in the art, once the basic creative concept of the present invention is known, several improvements and modifications can be made without departing from the principles of the present invention, and these improvements and modifications should also be regarded as the protection scope of the present invention. Therefore, the attached claims are intended to be interpreted as including the preferred embodiment and all changes and modifications that fall within the scope of the embodiments of the present invention.
Claims
1. A method for tracking aerial targets based on geostationary orbit staring satellite remote sensing images, characterized in that: The following steps are involved: Step S1, obtaining a trajectory list of all aerial targets, and predicting all trajectories in the trajectory list to the current moment; Step S2, obtaining ground processing to form an aerial target detection result, and associating the updated trajectory with the obtained aerial target detection result; Step S3: Process the trajectory with successful target association and the trajectory with failed target association separately.
2. The method for tracking aerial targets based on geostationary orbit staring satellite remote sensing images according to claim 1, characterized in that: The step S1 specifically includes: Step S11: Obtaining a list of aerial target trajectories Tra represents the current air target trajectory list set, T n represents a certain trajectory, N represents the number of all aerial target trajectories at the current moment, the current moment is moment k, and the current state information of the trajectory is the state of the previous frame, i.e. moment k-1, i.e. The state is (x k-1 ,vx k-1 ,y k-1 ,vy k-1 ) T , where x k-1 ,y k-1 are the longitude and latitude at time k-1, vx k-1 ,vy k-1 are the velocity components in longitude and latitude, respectively. is the covariance matrix of trajectory k-1; Step S12: According to the Kalman filter model, all trajectory states are predicted to the current k moment: Among them, f k-1 is the state transfer function; According to the mid-latitude navigation method, under the constant direction and constant speed model, the longitude and latitude update formula is as follows: Among them, t is the time difference between time k-1 and time k, and a is the number of kilometers corresponding to the arc length of 1° longitude at the equator, then: Step S13: predict all trajectory covariance matrices to the current k moment.
3. The method for tracking aerial targets based on geostationary orbit staring satellite remote sensing images according to claim 2, characterized in that: In step S13, based on step S12, it can be known that the state transfer function f k-1 The state transfer matrix F is nonlinear and is used to update the covariance matrix k-1 is the state transition function f k-1 The Jacobian matrix of is: in, Then the covariance matrix update formula is: Among them, Q k-1 is the system process noise matrix, and the process noise influencing factors are set based on experience.
4. The method for tracking aerial targets based on geostationary orbit staring satellite remote sensing images according to claim 3 is characterized in that: The step S2 specifically includes: Step S21: Obtain all aerial target detection results at the current k moment Tar represents the set of all test results at time k, Z m represents one of the measurement results, namely lon k ,lat k Respectively represent the latitude and longitude results of the target measurement; Step S22: for each predicted trajectory in the Tra set With each measurement result Z in the Tar set m Make an association.
5. The method for tracking aerial targets based on geostationary orbit staring satellite remote sensing images according to claim 4 is characterized in that: The step S22 specifically includes: Step S221, using Mahalanobis distance as the basis for association, that is, the target with the closest Mahalanobis distance of each trajectory is used as the associated target; According to the trajectory Updated covariance matrix Extract the trajectory latitude and longitude covariance matrix: According to Cov(x,y) and each measurement point Z m Coordinates of the trajectory and calculate the Mahalanobis distance between each measurement point: At the same time, set the target association maximum distance threshold D max , if a detection point Z m With trajectory The Mahalanobis distance is the shortest among all the detection points and satisfies It is considered that the detection point Z m With trajectory The association is successful; If the distance between a trajectory and all detection points is greater than D max , then the corresponding trajectory association is considered to have failed; If the distance between a detection point and all trajectories is greater than D max , then the corresponding detection point is considered to have no associated trajectory.
6. The method for tracking aerial targets based on geostationary orbit staring satellite remote sensing images according to claim 5, characterized in that: The step S3 specifically includes: Update the status of the trajectory that is successfully associated with the target, and count the number of successful trajectory associations. The specific update steps are as follows: According to the test results The observation matrix The measurement residual is then: According to the Kalman filter formula, the new information covariance S k for: Among them, R k is the measurement noise matrix. Since there are only longitude and latitude for the target detection result, R k Generally expressed as: Then the Kalman gain is: The trajectory state is updated as follows: The trajectory covariance matrix is updated as:
7. The method for tracking aerial targets based on geostationary orbit staring satellite remote sensing images according to claim 5, characterized in that: For targets that are not associated with any trajectory, a new trajectory is established, and the initial state of the new trajectory is set to: T new =(lon k ,vx k-1 ,lat k ,vy k-1 ) T , the initial covariance matrix is set as: Among them, V max The maximum speed at which an aerial target can fly.
8. The method for tracking aerial targets based on geostationary orbit staring satellite remote sensing images according to claim 5, characterized in that: For the trajectory with failed association, the number of consecutive association failures is recorded; Based on the preset continuous association failure number threshold as the trajectory deletion judgment threshold judgment condition, the trajectories that exceed the continuous association failure number threshold are deleted, and the undeleted trajectories are used as the trajectories for the next round of trajectory prediction association.
9. The method for tracking aerial targets based on geostationary orbit staring satellite remote sensing images according to claim 7, characterized in that: Based on the preset continuous association success number threshold as the trajectory output judgment threshold judgment condition, the trajectory that meets the continuous association success number threshold is output.