CN112782696A - Sequence ISAR image scattering center multi-hypothesis tracking track correlation method - Google Patents

Sequence ISAR image scattering center multi-hypothesis tracking track correlation method Download PDF

Info

Publication number
CN112782696A
CN112782696A CN202110117517.9A CN202110117517A CN112782696A CN 112782696 A CN112782696 A CN 112782696A CN 202110117517 A CN202110117517 A CN 202110117517A CN 112782696 A CN112782696 A CN 112782696A
Authority
CN
China
Prior art keywords
track
hypothesis
observation
scattering center
frame
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN202110117517.9A
Other languages
Chinese (zh)
Other versions
CN112782696B (en
Inventor
周峰
杜荣震
刘磊
白雪茹
石晓然
王常龙
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Xidian University
Original Assignee
Xidian University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Xidian University filed Critical Xidian University
Priority to CN202110117517.9A priority Critical patent/CN112782696B/en
Publication of CN112782696A publication Critical patent/CN112782696A/en
Application granted granted Critical
Publication of CN112782696B publication Critical patent/CN112782696B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
    • G01S13/00—Systems using the reflection or reradiation of radio waves, e.g. radar systems; Analogous systems using reflection or reradiation of waves whose nature or wavelength is irrelevant or unspecified
    • G01S13/88—Radar or analogous systems specially adapted for specific applications
    • G01S13/89—Radar or analogous systems specially adapted for specific applications for mapping or imaging
    • G01S13/90—Radar or analogous systems specially adapted for specific applications for mapping or imaging using synthetic aperture techniques, e.g. synthetic aperture radar [SAR] techniques
    • G01S13/9021—SAR image post-processing techniques
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
    • G01S13/00—Systems using the reflection or reradiation of radio waves, e.g. radar systems; Analogous systems using reflection or reradiation of waves whose nature or wavelength is irrelevant or unspecified
    • G01S13/88—Radar or analogous systems specially adapted for specific applications
    • G01S13/89—Radar or analogous systems specially adapted for specific applications for mapping or imaging
    • G01S13/90—Radar or analogous systems specially adapted for specific applications for mapping or imaging using synthetic aperture techniques, e.g. synthetic aperture radar [SAR] techniques
    • G01S13/9004—SAR image acquisition techniques
    • G01S13/9019—Auto-focussing of the SAR signals
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
    • G01S13/00—Systems using the reflection or reradiation of radio waves, e.g. radar systems; Analogous systems using reflection or reradiation of waves whose nature or wavelength is irrelevant or unspecified
    • G01S13/88—Radar or analogous systems specially adapted for specific applications
    • G01S13/89—Radar or analogous systems specially adapted for specific applications for mapping or imaging
    • G01S13/90—Radar or analogous systems specially adapted for specific applications for mapping or imaging using synthetic aperture techniques, e.g. synthetic aperture radar [SAR] techniques
    • G01S13/904—SAR modes
    • G01S13/9064—Inverse SAR [ISAR]
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
    • Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
    • Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation

Landscapes

  • Engineering & Computer Science (AREA)
  • Remote Sensing (AREA)
  • Radar, Positioning & Navigation (AREA)
  • Physics & Mathematics (AREA)
  • Electromagnetism (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • General Physics & Mathematics (AREA)
  • Radar Systems Or Details Thereof (AREA)

Abstract

The invention discloses a sequence ISAR image scattering center multi-hypothesis tracking track correlation method, which comprises the steps of firstly, obtaining a high-resolution ISAR image sequence of an observed target and extracting two-dimensional projection coordinates of a scattering center in the image sequence on the basis of an ISAR target observation imaging model and a state transfer equation and an observation equation of a scattering center track position; then, using multi-hypothesis tracking method, the scattering center track association is performed from both forward and backward directions. Determining the only hypothesis of track association through N backtracking pruning hypothesis tree management, and generating an associated track; and finally, obtaining the final associated flight path through flight path fusion and Kalman filtering. The assumed tree framework adopted by the method has good response capability to false alarm, new generation and interruption situations in the association process, so that the robustness of the method is ensured.

Description

Sequence ISAR image scattering center multi-hypothesis tracking track correlation method
Technical Field
The invention relates to the technical field of radar image processing, in particular to a sequence ISAR image scattering center multi-hypothesis tracking track correlation method.
Background
The motion restoration structure algorithm is initially applied to three-dimensional imaging of an optical target, and the algorithm reconstructs a three-dimensional profile of the observed target based on a sequence of optical images acquired by a camera under multiple viewing angles. The algorithm comprises 4 main steps: feature extraction, feature association, three-dimensional estimation and error optimization. The feature association is used for realizing correct matching of the same feature in different images, and is a key point and difficulty for realizing accurate three-dimensional reconstruction. The ISAR image mainly comprises strong scattering points which are similar to the corner point characteristics in the optical image, so that a plurality of researchers transfer the motion recovery structure algorithm to ISAR three-dimensional imaging, and the three-dimensional reconstruction of the observed target is realized based on an image sequence obtained by single-base ISAR multi-view echo data imaging. The ISAR target three-dimensional reconstruction result can visually and clearly reflect the three-dimensional structure of the target, and has important application value in the aspects of battlefield target investigation, target automatic identification and the like.
The difficulty associated with scattering centers of an ISAR image sequence is that Radar backscattering coefficient (RCS) of the scattering centers are significantly anisotropic, and the occlusion of different scattering centers causes the interruption of the scattering center track in the image sequence. The correlation algorithms proposed by the existing research can be roughly divided into two types: an image feature based correlation method and a motion model based correlation method. The correlation method based on the image features is characterized in that the strong scattering center in the ISAR image is used as a feature point to be extracted of the image, and matching correlation of the same scattering center in different images is realized by designing a reasonable feature descriptor and a feature matching criterion. The KLT (Kanade Lucas Tomasi, KLT) feature tracking algorithm is a typical image feature-based correlation algorithm, but the algorithm is based on the assumption that the brightness of a target is constant, and the condition is difficult to guarantee in an ISAR image sequence, so that the application of the algorithm in ISAR images is fundamentally limited. The Speeded Up Robust Features (SURF) algorithm is another commonly used correlation algorithm based on image Features, and correlation of scattering centers can be completed under the condition that a small rotation angle is satisfied between two adjacent frames of images. However, the scattering center characteristics of the ISAR images are similar, and when the rotation angle between the two images is slightly larger, the algorithm performance is poor.
In view of the fact that ISAR images have obvious anisotropy among images and single features in the images compared with optical images, and the correlation method based on the image features is difficult to obtain ideal correlation effects, some scholars propose a motion model based on a scattering center track to solve the correlation problem from the research angle of multi-target tracking. Unlike the correlation method based on image features, the method needs to complete the coordinate extraction of the scattering center in advance. In the research field of solving the association problem based on a scattering center motion model, on the basis of a model-built track form, a scholars proposes to adopt a Markov Chain Monte Carlo (MCMC) data association algorithm and design a prior and likelihood model of Bayesian inference according to the motion characteristics of a scattering center track to realize the scattering center track association. However, the performance of the algorithm depends heavily on the design of the prior and likelihood probability models, and once the models are mismatched, the correlation effect is poor. Another scholars proposes a correlation method based on plane homography, which assumes that a target surface consists of several rectangular planes according to the same characteristic of flight path state transfer equations of scattering centers between two frames of ISAR images on the same plane, manually marks the positions of four corner points of each rectangular plane in an image sequence and forms coordinate pairs through matching, and then solves the state transfer equations of the scattering centers of each plane between the image sequences by adopting a least square method, thereby realizing the flight path correlation of the scattering centers. However, in practice, the ISAR target surface is usually not only formed by rectangular planes, and therefore, the method has limited application in the actual sequential ISAR image track correlation.
According to the analysis, the ISAR image characteristics have the characteristics of single characteristic type and severe intensity change compared with the optical image characteristics, so that the correlation method with good performance in the optical image is limited to be applied to the ISAR image; due to the fact that ISAR observation environment and a target structure are complex, the existing correlation algorithm based on the scattering center track state transition motion model lacks robustness. In summary, for the present time, the problem of ISAR image sequence scattering center track correlation still lacks an accurate and robust solution.
Disclosure of Invention
Aiming at the problems in the prior art, the invention aims to provide a sequence ISAR image scattering center multi-hypothesis tracking track association method, and the adopted hypothesis tree framework has good response capability to false alarms, new generations and interruption situations in the association process, so that the robustness of the method is ensured; the forward and reverse correlation fusion operation solves the problem of inaccurate correlation caused by lack of prior information during initial correlation, eliminates the influence of the correlation direction on the correlation result, and further improves the accuracy of the method.
The technical principle of the invention is as follows: on the basis of an ISAR (inverse synthetic aperture radar) target observation imaging model and a state transfer equation and an observation equation of a scattering center track position, firstly, a distance-Doppler imaging algorithm is adopted to obtain a high-resolution ISAR image sequence of an observation target, and two-dimensional projection coordinates of a scattering center in the image sequence are extracted through an ESPRIT (rotation invariant subspace) algorithm; then, using multi-hypothesis tracking method, the scattering center track association is performed from both forward and backward directions. Continuously updating the flight path state transition parameters of the scattering center by using the information of the existing hypothesis trees in the association process to generate two-direction hypothesis trees; then, determining a unique hypothesis of track association through N backtracking pruning hypothesis tree management, and generating an associated track; and finally, realizing information fusion of the flight paths obtained by association in the positive direction and the negative direction through flight path fusion, and performing Kalman filtering on the fused flight paths to obtain the final associated flight paths.
In order to achieve the purpose, the invention is realized by adopting the following technical scheme.
A sequence ISAR image scattering center multi-hypothesis tracking track correlation method comprises the following steps:
step 1, obtaining target echoes of ISAR (inverse synthetic aperture radar) long-time continuous observation, and performing sub-aperture division on target echo data to obtain K-segment sub-aperture data; sequentially performing distance compression, envelope alignment and self-focusing operation on each section of sub-aperture data to obtain a target high-resolution two-dimensional ISAR image sequence;
step 2, establishing a target coordinate system and a radar observation imaging model after translation compensation; establishing a state transition equation of the scattering center track between two frames of images and a position state observation equation of the scattering center; extracting the scattering center projection coordinates in the target high-resolution two-dimensional ISAR image sequence by utilizing an ESPRIT algorithm to obtain the observation position of each scattering center track at the corresponding moment of each frame of image;
step 3, performing scattering center track association frame by frame from the positive direction and the negative direction by using a multi-hypothesis tracking algorithm, namely continuously updating scattering center track state transition parameters by using the generated hypothesis trees of the positive direction and the negative direction to generate hypothesis trees of the positive direction and the negative direction;
step 4, performing hypothesis tree management by using an N backtracking pruning method, and determining scattering center tracks generated by association in the positive and negative directions, namely association tracks in the positive and negative directions;
and 5, fusing the correlation tracks in the positive direction and the negative direction to realize the complementation of the correlation information in the positive direction and the negative direction to obtain a fused track, and performing Kalman filtering on the fused track to obtain a final track correlation result.
Compared with the prior art, the invention has the beneficial effects that:
the method of the invention provides an accurate robust track association scheme based on the characteristics and key problems of the ISAR image sequence scattering center track association: aiming at the characteristics of single characteristic type and severe intensity change of a scattering center imaging result, only depending on coordinate motion information projected by a scattering center to carry out correlation, and eliminating the influence of unreliable redundant information on the correlation result; aiming at the problem that the flight path state transition model parameters are unknown, parameter estimation is carried out by combining a flight path theoretical model, and the limitation that the flight path state cannot be predicted and associated due to the fact that the state transition model is unknown is broken through; aiming at the complex situations of track interruption, track regeneration and false alarm existing in the track association, multi-hypothesis association and delayed decision judgment are carried out on the basis of a hypothesis tree frame, and association accuracy is guaranteed; and the track fusion filtering is carried out aiming at the problem of inaccurate parameter estimation of the initial model, so that the accuracy of initial association is improved. The comprehensive analysis of the characteristics and the symptoms of the scattering center track correlation problem enables the method to have accurate and steady performance in the process of processing the scattering center track correlation problem.
Drawings
The invention is described in further detail below with reference to the figures and specific embodiments.
FIG. 1 is a schematic flow chart of a sequential ISAR image scattering center multi-hypothesis tracking track association method provided by the present invention;
FIG. 2 is a schematic diagram of an ISAR target observation and imaging model according to an embodiment of the present invention;
FIG. 3 is a diagram of an approximate circular motion estimation model according to an embodiment of the present invention;
FIG. 4 is a hypothetical management process schematic of an embodiment of the present invention;
FIG. 5 is a diagram of an aircraft model composed of 16 ISAR target points according to an embodiment of the present invention;
FIG. 6 is a high-resolution two-dimensional ISAR image of a portion of an aircraft target point model obtained by the method of the present invention; wherein 6(a) is the 1 st sub-aperture imaging result; 6(b) is the 10 th sub-aperture imaging result; 6(c) is the 20 th sub-aperture imaging result;
FIG. 7 is a diagram of a track coordinate extraction result of an image sequence corresponding to different sub-apertures;
FIG. 8 is a result diagram of a track correlation performed on the extracted results of FIG. 7;
FIG. 9 is a result diagram of a target three-dimensional reconstruction of the track correlation result of FIG. 8; wherein, 9(a) is a comparison graph of a space target point model reconstruction result and a real structure thereof in a three-dimensional space; 9(b) is a comparison graph of the spatial target point model reconstruction result and the projection of the real structure on the XOY plane; FIG. 9(c) is a comparison graph of the reconstruction result of the spatial object point model and the projection of the real structure thereof on the XOZ plane; and 9(d) is a comparison graph of the space target point model reconstruction result and the projection of the real structure on the YOZ plane.
Detailed Description
Embodiments of the present invention will be described in detail below with reference to examples, but it will be understood by those skilled in the art that the following examples are only illustrative of the present invention and should not be construed as limiting the scope of the present invention.
Referring to fig. 1, the method for correlating a sequential ISAR image scattering center multi-hypothesis tracking track provided by the present invention includes the following steps:
step 1, obtaining target echoes of ISAR (inverse synthetic aperture radar) long-time continuous observation, and performing sub-aperture division on the target echoes to obtain K-segment sub-aperture target data; sequentially performing distance compression, envelope alignment and self-focusing operation on each section of sub-aperture target data to obtain a target high-resolution two-dimensional ISAR image sequence;
the method specifically comprises the following steps:
(1.1) continuously observing a target for a long time by an ISAR system to obtain echo data of the target in a wide observation angle range;
(1.2) carrying out sub-aperture division on the obtained target echo data to obtain K-segment sub-aperture data;
(1.3) constructing a reference signal, and performing pulse compression on each echo data in each sub-aperture to obtain a high-resolution one-dimensional range profile sequence;
the reference signal is generated as: taking the distance from the inverse synthetic aperture radar to the scene center as a reference distance, and constructing a linear frequency modulation signal with the same parameter as the transmitted signal as a reference signal;
(1.4) for each sub-aperture pulse pressure data, namely a high-resolution one-dimensional range profile sequence, firstly, envelope alignment is carried out on each sub-aperture pulse pressure data by adopting an envelope alignment method based on adjacent correlation, and envelope offset caused by target translation is compensated; then, compensating echo initial phase errors by adopting a self-focusing method based on a minimum entropy criterion; and finally, carrying out azimuth Fourier transform on the compensated data to obtain a two-dimensional high-resolution image sequence.
Step 2, establishing a target coordinate system and a radar observation imaging model after translation compensation; establishing a state transition equation of the scattering center track between two frames of images and position state observation of the scattering center; extracting the scattering center projection coordinates in the target high-resolution two-dimensional ISAR image sequence by utilizing an ESPRIT algorithm to obtain the observation position of each scattering center track at the corresponding moment of each frame of image;
as shown in fig. 1, a translation compensated target coordinate system O-XYZ is established, where O is the target center and LOS is the translation compensationThe latter equivalent radar line of sight, l, is the projection of the radar line of sight on the XOY plane. Azimuth angle theta and pitch angle of radar equivalent visual line direction
Figure BDA0002920896700000071
And (4) defining. The target motion is equivalent to a rotation with an angular velocity ω about the fixed axis OZ.
And dividing the whole observation period into K sub-apertures, and correspondingly, obtaining the image frame number of the ISAR image sequence as K. Let us note that the time interval between adjacent sub-apertures is Δ t, and the coherent processing time of each sub-aperture is tCPI. Consider the imaging situation in the kth sub-aperture, let tkRepresenting the instant of the sub-aperture centre for any nth scattering centre s on the targetn＝(xn，yn，zn)TThe theoretical projection coordinate on the k frame image is
Figure BDA0002920896700000072
The two satisfy the following projection relation:
Figure BDA0002920896700000073
scattering center snThe flight path is as follows:
Figure BDA0002920896700000074
the scattering center trajectory is in the form of an ellipse, where (x)n)2＝(xn)2+(yn)2Compensating for the equivalent radius of rotation of the back scattering center for translational motion. During the actual observation, the parameter rn，znAnd radar line of sight equivalent pitch angle
Figure BDA0002920896700000075
Estimation acquisition is usually required.
Establishing a state transition equation of the projection position of the scattering center between two adjacent frames of images:
Figure BDA0002920896700000076
wherein the content of the first and second substances,
Figure BDA0002920896700000081
are respectively a parameter zn、
Figure BDA0002920896700000082
An estimate of ω. Target three-dimensional equivalent rotation angular velocity omega and imaging effective rotation angular velocity omegaeApproximately satisfy the relation
Figure BDA0002920896700000083
Obtaining an estimate of effective rotational angular velocity by image orientation scaling
Figure BDA0002920896700000084
Thereafter, the estimated values are further combined
Figure BDA0002920896700000085
Calculating an estimated value of equivalent rotational angular velocity
Figure BDA0002920896700000086
Figure BDA0002920896700000087
And
Figure BDA0002920896700000088
the estimation and updating of (b) is performed during the hypothesis tree growing process.
The one-step state transition equation of the scattering center track between two frames of images is as follows:
Figure BDA0002920896700000089
wherein the content of the first and second substances,
Figure BDA00029208967000000810
Figure BDA00029208967000000811
for noise in the state transition process, white Gaussian noise is usually assumed, and the correlation matrix is Qp＝E[χk(χk)T]＝diag{[δ1，δ2]}，δ1And delta2Representing the noise covariance of the distance and azimuth position during the transfer, respectively.
The position state observation of the scattering center is obtained as follows:
Figure BDA00029208967000000812
wherein the content of the first and second substances,
Figure BDA00029208967000000813
denotes the scattering center snAt the observation position in the image of the k-th frame,
Figure BDA00029208967000000814
a representation of an observation matrix is shown,
Figure BDA00029208967000000815
to observe noise; the observed noise is assumed to be white Gaussian noise
Figure BDA00029208967000000816
Figure BDA00029208967000000817
To know
Figure BDA00029208967000000818
Representing the noise covariance of the range and azimuth positions during the observation, respectively.
Adopting an ESPRIT algorithm to carry out scattering center extraction operation on each frame of image in the calibrated ISAR image sequence to obtain the observation position of each scattering center track at the corresponding moment of each frame of image
Figure BDA00029208967000000819
Step 3, performing scattering center track association frame by frame from the positive direction and the negative direction by using a multi-hypothesis tracking algorithm, namely continuously updating scattering center track state transition parameters by using the generated hypothesis trees of the positive direction and the negative direction to generate hypothesis trees of the positive direction and the negative direction;
the method specifically comprises the following steps:
(3.1) assume tree initialization:
for a chronological sequence of ISAR images, it is defined that the correlation is backward from the first frame as a forward correlation and forward from the last frame as a backward correlation. Selecting any one correlation direction, setting the optimal assumed nodes reserved in each layer of the assumed tree to be M, assuming the tree depth to be N, setting the maximum Life of the flight path to be k, wherein k is an integer larger than 0, and considering that the flight path is terminated if the Life of a certain flight path is equal to 0. When the hypothesis tree is initialized, all observations are initialized to new tracks, and the initial innovation covariance of each new track is P0. At this time, the tree is assumed to have only one root node, and the nodes are the only assumptions of the scattering center track association.
(3.2) state transition model estimation and state prediction:
suppose a new hypothesis is being generated based on a hypothesis node corresponding to the k-1 frame image, let
Figure BDA0002920896700000091
Representing the set of tracks for which an association has been assumed in the node,
Figure BDA0002920896700000092
representing the nth track, L, in the setk-1Representing the total number of tracks in the set. And defining the flight path length as the total number of observation points and prediction points of the flight path. From Tk-1Zhongren flight path
Figure BDA0002920896700000093
According to
Figure BDA0002920896700000094
The length of the path completes the state transition parameter estimation and state prediction of the path.
Figure BDA0002920896700000095
The length is divided into three cases:
(a)
Figure BDA0002920896700000096
the length is 1:
due to the lack of more track information, the state prediction is performed as follows:
Figure BDA0002920896700000097
wherein the content of the first and second substances,
Figure BDA0002920896700000098
to represent
Figure BDA0002920896700000099
The position state prediction result in the k-th frame image,
Figure BDA00029208967000000910
representing the theoretical observed coordinates of the flight path in the k-th frame image,
Figure BDA00029208967000000911
to represent
Figure BDA00029208967000000912
And observing the actual position in the k-1 frame image.
(b)
Figure BDA00029208967000000913
Length greater than 1 and less than 4:
to wind the center o of the image (0, 0)TThe circular motion of (2) is approximate to track motion, and estimation of state transition parameters is carried out. As shown in FIG. 3, assume that
Figure BDA00029208967000000914
The observation in the image of the (k-1) th frame is an observation point 2, the observation in the image of the (k-2) th frame is an observation point 1, the motion from the observation point 1 to the observation point 2 is approximate to the circular motion with the center o of the image as the center and the radius of the distance r from the observation value 2 to the o as the radius, and the corner
Figure BDA0002920896700000101
With counterclockwise being positive. The motion of the flight path from the (k-1) th frame to the (k) th frame is still approximate to the circular motion and the corner with o as the center of a circle and r as the radius according to the stable and continuous motion characteristic of the flight path
Figure BDA0002920896700000102
Calculating the position prediction of the flight path in k frames as follows:
Figure BDA0002920896700000103
if it is
Figure BDA0002920896700000104
If there is no corresponding observation in the k frame image, then order
Figure BDA0002920896700000105
(c)
Figure BDA0002920896700000106
A length of 4 or more
And directly adopting least square fitting ellipse parameters to calculate a transfer equation of the track position state. The scattering center projection trajectory satisfies the elliptical motion equation,
Figure BDA0002920896700000107
the corresponding equation of motion is:
Figure BDA0002920896700000108
the elliptical trajectory is converted into a more general form:
Figure BDA0002920896700000109
Figure BDA00029208967000001010
the following relationship is further obtained:
Figure BDA00029208967000001011
Figure BDA00029208967000001012
Figure BDA00029208967000001013
e2＝e4＝0
in the formula
Figure BDA0002920896700000111
To know
Figure BDA0002920896700000112
Are the parameters that need to be estimated in the state transition equations.
Let e be [ e1，e3，e5，e6]TFormation matrix W [ u ], [ v ], [ 1 ]]Wherein u and v respectively represent an azimuth coordinate vector and a distance coordinate vector composed of all scattering center coordinates on the associated track, < > represents a Hadamard product, and 1 represents a column vector of all elements 1. Assuming the flight path length is xi, then
Figure BDA0002920896700000113
Figure BDA0002920896700000114
Defining an error function
F(e)＝(We)TWe＝eTWTWe
When the e parameter is the best fit parameter of the ellipse, the value of F (e) is minimized. Thus, the solution of the ellipse parameters is translated into the solution of the following constraint problem:
arg min F(e)＝eTWTWe
s.t.eTFe＝1
wherein the content of the first and second substances,
Figure BDA0002920896700000115
solving the equation constraint optimization problem by adopting a Lagrange multiplier method, introducing a Lagrange multiplier lambda and performing derivation to obtain the equation constraint optimization problem
2Fe-2λWTWe＝0
eTFe＝1
Let S be WTW, further obtaining
Fe＝λSe
eTFe＝1
Solving by adopting a generalized eigenvalue solving method to obtain an eigenvalue and eigenvector pair (lambda)i，ei) It is clear that for any non-zero real number μ, (μ λ)i，μei) For a set of solutions that still satisfy the constraints, there are
Figure BDA0002920896700000116
Further obtaining an ellipse parameter set
Figure BDA0002920896700000117
Then the elliptical optimal parameter estimate eoptIs composed of
Figure BDA0002920896700000118
By means of eoptEach element of (A) and
Figure BDA0002920896700000119
to know
Figure BDA00029208967000001110
Is calculated from the relationship of
Figure BDA00029208967000001111
And
Figure BDA00029208967000001112
the value of (c). At this time, A in the state transition equationk|k-1、bnAll estimates are obtained, and the state prediction of the target at the next moment is calculated according to the estimation:
Figure BDA0002920896700000121
traverse T as abovekL in (1)kAnd (4) tracking the flight path to complete the state prediction of all the flight paths. The covariance is predicted in one step:
Figure BDA0002920896700000122
(3.3) hypothesis generation:
(a) and calculating the probability of the old flight path continuation, the new flight path and the false alarm of any current observation.
Recording probability P of track point detected in imagedThe nth observation in the kth frame image
Figure BDA0002920896700000123
The probability associated with the mth track is:
Figure BDA0002920896700000124
wherein, define
Figure BDA0002920896700000125
Respectively representing the innovation and innovation covariance corresponding to the association:
Figure BDA0002920896700000126
Figure BDA0002920896700000127
Figure BDA0002920896700000128
is Gaussian distributed and has
Figure BDA0002920896700000129
Furthermore, let PnewRepresenting the probability that a track point is a new track, PfThe probability that the track point is a false alarm is represented.
(b) Hypothesis assignments are made to all current observed sources, and probabilities of hypothesis assignments being established are calculated.
Suppose an observation set D of k frame imageskHas a total of NsAn observation, i group association hypothesis at previous time
Figure BDA00029208967000001210
The number of tracks in is NTIn a
Figure BDA00029208967000001211
On the basis of a priori assumptions on NsAssigning each observation, and constructing an assignment matrix omega:
Ω1(m，n)＝-ln(Pmn)
Ω2(m，n)＝-ln(Pnew)δ(m，n)
Ω3(m，n)＝-ln(Pf)δ(m，n)
Ω＝[Ω1，Ω2，Ω3]
and adopting a Murty algorithm to make an assignment hypothesis with the maximum M groups of possibilities for all current observations, wherein the j group of possibilitiesSuppose that
Figure BDA0002920896700000131
Middle to NsOne observation is assigned: old track continuation NtNew born target NnewFalse alarm NfAnd (4) respectively. Based on
Figure BDA0002920896700000132
Computing
Figure BDA0002920896700000133
Probability of being true:
Figure BDA0002920896700000134
the probability that the remaining M-1 groups are assigned is calculated in the same manner.
(c) Taking into account the probability that the previous assumption holds
Figure BDA0002920896700000135
Probability of being true:
Figure BDA0002920896700000136
wherein c is a normalization coefficient, D1：kSet of all observations up to the k-th frame, hypothesis set
Figure BDA0002920896700000137
Included
Figure BDA0002920896700000138
And the related assumptions made at all previous times IIk-1. The probability that the remaining M-1 groups are assigned to be established is calculated in the same way, and a new layer of hypothetical tree nodes is formed on the hypothetical tree.
(3.4) status update:
and updating the positions, the innovation covariance, the Life and other states of the targets at the previous moment to the current moment. If a certain observation is assumedIf the state is false alarm, the observation is removed from the state set; if a certain observation is assumed to be a new flight path, the position of the observation is taken as the initial position of the scattering center flight path, Life is set to be k, and innovation covariance is set to be P0And incorporating the observation into a track set at the current moment; if a certain observation hypothesis is associated with a certain track at the previous moment, updating the track position to be the current observed position, and updating the innovation covariance:
Figure BDA0002920896700000141
Figure BDA0002920896700000142
if the flight path at the previous moment is not associated with any observation hypothesis at the current moment, the flight path state classification is carried out according to the Life value of the flight path: if Life is greater than 0, considering that the flight path is temporarily interrupted, predicting and updating the position of the target at the current moment to be in the current position state, keeping the innovation covariance unchanged, and reducing Life by 1; and if Life is 0, the track is considered to be terminated, and the state of the track is not updated.
And (4) repeating the steps (3.1) to (3.4) and continuously generating new hypothesis nodes to form a hypothesis tree.
Step 4, performing hypothesis tree management by using an N backtracking pruning method, and determining scattering center tracks generated by association in the positive and negative directions, namely association tracks in the positive and negative directions;
the process is shown in fig. 4: assuming that the Murty algorithm selects M to be 2 optimal hypotheses each time, assuming that the tree depth N to be 3, when the depth of the assumed tree is greater than N, selecting a node (a node 11 in the figure) with the minimum assumed cost in all the current leaf nodes, tracing back to the N-1 layer upwards, taking the traced node as a new root hypothesis, and pruning the side branches (subtracting all the nodes connected with the node 3). The hypothetical case of track association in the newly generated root node is determined to be the scattering center track case.
And 5, fusing the correlation tracks in the positive direction and the negative direction to realize the complementation of the correlation information in the positive direction and the negative direction to obtain a fused track, and performing Kalman filtering on the fused track to obtain a final track correlation result.
The method is implemented by the following steps:
(a) track matching
Assume that the forward correlated track results in
Figure BDA0002920896700000143
Figure BDA0002920896700000144
The number is the ith' track, and I is the total number of tracks; similarly, the inversely correlated track results in
Figure BDA0002920896700000145
Definition of
Figure BDA0002920896700000146
Representing flight paths
Figure BDA0002920896700000147
To know
Figure BDA0002920896700000148
And counting the number of the coincident tracks of the two tracks. Definition ai′j′Representing flight paths
Figure BDA0002920896700000149
To know
Figure BDA00029208967000001410
A matching relationship ofi′j′1 represents
Figure BDA00029208967000001411
Matching
Figure BDA00029208967000001412
ai′j′Is 0 represents
Figure BDA00029208967000001413
And
Figure BDA00029208967000001414
and not matched. The following optimization problem is solved by the hungian algorithm,
Figure BDA0002920896700000151
Figure BDA0002920896700000152
obtaining a flight path matching relation set A ═ { a ═ ai′j′＝1}；
(b) Taking any element a in the set A i′j′1, first rejecting
Figure BDA0002920896700000153
To know
Figure BDA0002920896700000154
The first four frame results of (1); then go through
Figure BDA0002920896700000155
To know
Figure BDA0002920896700000156
In the mutual exclusion frame number, the track point (including the predicted value) in the mutual exclusion frame number is used as the fusion track ti′j′Track points of; then go through
Figure BDA0002920896700000157
And
Figure BDA0002920896700000158
in the common frame number, the track points (including the predicted values) with the same coordinates in the corresponding frame number are taken as the fusion track ti′j′Taking the coordinate mean value of track points (including predicted values) with different coordinates in corresponding frame times as a fusion track ti′j′Predicting track points; wherein mutually exclusive frames are next to
Figure BDA0002920896700000159
And
Figure BDA00029208967000001510
of the frames appearing only once, the total number of frames being
Figure BDA00029208967000001511
To know
Figure BDA00029208967000001512
The number of frames present in each;
(c) repeating the step (b) until all the elements A are traversed to obtain a track fusion result Tfuse＝{ti′j′}；
(d) For TfusePerforming Kalman filtering along the forward direction and then along the reverse direction to obtain a filtering result T for each track in the systemfilter。
The method of the invention provides an accurate robust track association scheme based on the characteristics and key problems of the ISAR image sequence scattering center track association: aiming at the characteristics of single characteristic type and severe intensity change of a scattering center imaging result, only depending on motion information of a scattering center projection track to carry out correlation, and eliminating the influence of unreliable redundant information on the correlation result; aiming at the problem that the flight path state transition model parameters are unknown, parameter estimation is carried out by combining a flight path theoretical model, and the limitation that the flight path state cannot be predicted and associated due to the fact that the state transition model is unknown is broken through; aiming at the complex situations of track interruption, track regeneration and false alarm existing in the track association, multi-hypothesis association and delayed decision judgment are carried out on the basis of a hypothesis tree frame, and association accuracy is guaranteed; and the track fusion filtering is carried out aiming at the problem of inaccurate parameter estimation of the initial model, so that the accuracy of initial association is improved. The comprehensive analysis of the characteristics and the symptoms of the scattering center track correlation problem enables the method to have accurate and steady performance in the process of processing the scattering center track correlation problem.
Simulation experiment
The effectiveness of the present invention is further illustrated by point target simulation imaging experiments.
(1) Simulation conditions
The ISAR target point model is an airplane model composed of 16 points as shown in figure 5, the equivalent radar sight azimuth angle theta is 45 degrees after translational motion compensation, and the pitch angle is
Figure BDA0002920896700000161
(2) Simulation experiment content and result analysis
Based on the ISAR echo data of the space target with long time and large angle, the method of the invention can obtain a high-resolution two-dimensional ISAR image sequence. FIG. 6 shows a high-resolution two-dimensional ISAR image of a target point model of an airplane obtained by the method of the present invention. Wherein, fig. 6(a) is the 1 st sub-aperture imaging result; FIG. 6(b) shows the result of the 10 th sub-aperture imaging; fig. 6(c) shows the 20 th sub-aperture imaging result. And (3) extracting the track coordinates in the sequence image, wherein the extraction result is shown in FIG. 7, and the light and dark points in the graph represent the sub-aperture sequence of the observation.
Then, the extracted scattering center track observation is correlated, and the obtained track correlation result is shown in fig. 8. And further performing target three-dimensional reconstruction based on the obtained track association result, wherein the obtained reconstruction result is shown in fig. 9. FIG. 9(a) is a comparison graph of the reconstruction result of the model of the target point in space and the real structure in three-dimensional space; FIG. 9(b) is a comparison graph of the spatial object point model reconstruction result and the projection of the real structure on the XOY plane; FIG. 9(c) is a comparison graph of the reconstruction result of the spatial object point model and the projection of the real structure thereof on the XOZ plane; FIG. 9(d) is a comparison of the spatial object point model reconstruction and its real structure projected on the YOZ plane.
By observing the graph 9, the three-dimensional scattering center position of the target point model can be accurately reconstructed based on the scattering center track correlation result obtained by the method provided by the invention, and the accuracy of the track correlation result is reflected. The method can accurately process the complex conditions of false alarm, new generation, interruption and the like in the association, and has obvious robustness compared with the prior method.
Although the present invention has been described in detail in this specification with reference to specific embodiments and illustrative embodiments, it will be apparent to those skilled in the art that modifications and improvements can be made thereto based on the present invention. Accordingly, such modifications and improvements are intended to be within the scope of the invention as claimed.

Claims (10)

1. A sequence ISAR image scattering center multi-hypothesis tracking track correlation method is characterized by comprising the following steps:
step 1, obtaining target echoes of ISAR (inverse synthetic aperture radar) long-time continuous observation, and performing sub-aperture division on target echo data to obtain K-segment sub-aperture data; sequentially performing distance compression, envelope alignment and self-focusing operation on each section of sub-aperture data to obtain a target high-resolution two-dimensional ISAR image sequence;
step 2, establishing a target coordinate system and a radar observation imaging model after translation compensation; establishing a state transition equation of the scattering center track between two frames of images and a position state observation equation of the scattering center; extracting the scattering center projection coordinates in the target high-resolution two-dimensional ISAR image sequence by utilizing an ESPRIT algorithm to obtain the observation position of each scattering center track at the corresponding moment of each frame of image;
step 3, performing scattering center track association frame by frame from the positive direction and the negative direction by using a multi-hypothesis tracking algorithm, namely continuously updating scattering center track state transition parameters by using the generated hypothesis trees of the positive direction and the negative direction to generate hypothesis trees of the positive direction and the negative direction;
step 4, performing hypothesis tree management by using an N backtracking pruning method, and determining scattering center tracks generated by association in the positive and negative directions, namely association tracks in the positive and negative directions;
and 5, fusing the correlation tracks in the positive direction and the negative direction to realize the complementation of the correlation information in the positive direction and the negative direction to obtain a fused track, and performing Kalman filtering on the fused track to obtain a final track correlation result.
2. The method for correlating multiple hypothesis tracking tracks of scattering centers of sequential ISAR images according to claim 1, wherein the distance compression, envelope alignment and self-focusing operations are sequentially performed on each segment of sub-aperture data, specifically:
(1.1) constructing a reference signal: taking the distance from the inverse synthetic aperture radar to the scene center as a reference distance, and constructing a linear frequency modulation signal with the same parameter as the transmitted signal as a reference signal;
(1.2) performing pulse compression on each echo data in each sub-aperture based on a reference signal to obtain each sub-aperture pulse pressure data, namely a high-resolution one-dimensional range profile sequence;
(1.3) for each sub-aperture pulse pressure data, firstly, carrying out envelope alignment on each sub-aperture pulse pressure data by adopting an envelope alignment method based on adjacent correlation, and compensating envelope offset caused by target translation; then, compensating echo initial phase errors by adopting a self-focusing method based on a minimum entropy criterion; and finally, carrying out azimuth Fourier transform on the compensated data to obtain a high-resolution two-dimensional image sequence.
3. The sequence ISAR image scattering center multi-hypothesis tracking track association method according to claim 1, wherein a translation-compensated target coordinate system and a radar observation imaging model are established; accordingly, a state transition equation of the scattering center track between two frames of images and a position state observation equation of the scattering center are established, and the method specifically comprises the following steps:
(2.1) establishing a translation compensated target coordinate system O-XYZ, wherein O is a target center; the azimuth angle of the equivalent sight line direction of the radar in the radar observation imaging model is theta, and the pitch angle is
Figure FDA0002920896690000021
The target motion is equivalent to rotation with an angular velocity omega around a fixed axis OZ;
(2.2) correspondingly obtaining K frames of target high-resolution two-dimensional ISAR images by K sections of sub-aperture target data, wherein the K frames of images form an image sequence; let us note that the time interval between adjacent sub-apertures is Δ t, and the coherent processing time of each sub-aperture is tCPI(ii) a Let tkRepresenting the instant of the sub-aperture centre for any nth scattering centre s on the targetn＝(xn，yn，zn)TThe theoretical projection coordinate on the k frame ISAR image is
Figure FDA0002920896690000022
The two satisfy the following projection relation:
Figure FDA0002920896690000023
scattering center snThe flight path is as follows:
Figure FDA0002920896690000024
wherein, the superscript T is the transposition of the matrix; (r)n)2＝(xn)2+(yn)2Compensating the equivalent rotation radius of the back scattering center for translation;
during the actual observation, the parameter rn，znEquivalent pitch angle with radar sight line
Figure FDA0002920896690000031
Obtained by estimation;
(2.3) establishing a state transition equation of the projection position of the scattering center between two adjacent frames of images:
Figure FDA0002920896690000032
wherein the content of the first and second substances,
Figure FDA0002920896690000033
are respectively a parameter zn、
Figure FDA0002920896690000034
An estimate of ω; a. thek|k-1、bnAre respectively asA scattering center track state transition parameter; by estimating parameters
Figure FDA0002920896690000035
Obtaining a scattering center track state transition parameter;
(2.4)
Figure FDA0002920896690000036
the estimation process of (2) is: target three-dimensional equivalent rotation angular velocity omega and imaging effective rotation angular velocity omegaeSatisfy the relation
Figure FDA0002920896690000037
Obtaining an estimate of effective rotational angular velocity by image orientation scaling
Figure FDA0002920896690000038
Thereafter, the estimated values are further combined
Figure FDA0002920896690000039
Calculating an estimated value of equivalent rotational angular velocity
Figure FDA00029208966900000310
Figure FDA00029208966900000311
And
Figure FDA00029208966900000312
the estimation and updating of (2) is carried out during the growth of the hypothetical tree;
(2.5) the one-step state transition equation of the scattering center track between two frames of images is as follows:
Figure FDA00029208966900000313
wherein the content of the first and second substances,
Figure FDA00029208966900000314
Figure FDA00029208966900000315
noise that is the process of state transition;
the position state observation of the scattering center is obtained as follows:
Figure FDA00029208966900000316
wherein the content of the first and second substances,
Figure FDA00029208966900000317
denotes the scattering center snAt the observation position in the image of the k-th frame,
Figure FDA00029208966900000318
a representation of an observation matrix is shown,
Figure FDA00029208966900000319
is a scattering center snObservation noise in the k-th frame image.
4. The method for associating the scattering center multi-hypothesis tracking tracks of the sequence ISAR images according to claim 3, wherein the method for associating the scattering center tracks frame by frame from the forward direction and the backward direction by using the multi-hypothesis tracking algorithm comprises the following steps:
(3.1) assume tree initialization:
for the ISAR image sequence arranged in time sequence, defining that the backward correlation is the forward correlation from the first frame and the forward correlation is the backward correlation from the last frame; for any correlation direction, setting the number of the reserved optimal assumed nodes of each layer in the assumed tree to be M, the assumed tree depth to be N, the maximum Life of the flight path to be k, and the k to be an integer greater than 0;
when the hypothesis tree is initialized, all observations are initialized to new tracks, and the initial innovation covariance of each new track is P0；
(3.2) target state transition parameter estimation and state prediction:
let a new hypothesis be generated based on a hypothesis node corresponding to the k-1 th frame of image; order to
Figure FDA0002920896690000041
Representing the set of tracks for which an association has been assumed in the node,
Figure FDA0002920896690000042
representing the nth track, L, in the setk-1Representing the total number of tracks in the set; defining the flight path length as the total number of observation points and prediction points of the flight path;
from Tk-1Zhongren flight path
Figure FDA0002920896690000043
According to
Figure FDA0002920896690000044
The length of the flight path is used for completing the state transition parameter estimation and state prediction of the flight path;
(3.3) hypothesis generation:
(3.3a) calculating the probability that any current observation is old track continuation, new track and false alarm;
(3.3b) making hypothesis assignments to all current observations based on the probabilities calculated in step (3.3a) and calculating the probability that each hypothesis assignment holds;
(3.3c) calculating the current hypothesis assignment based on the probability that the previous hypothesis assignment holds
Figure FDA0002920896690000045
A probability of being established;
(3.4) status update:
if a certain observation hypothesis is a false alarm, removing the observation from the state set; if a certain observation is assumed to be a new flight path, the position of the observation is taken as the initial position of the scattering center flight path, Life is set to be k, and innovation covariance is set to be P0And is combined withBringing the observation into a track set at the current moment; if a certain observation hypothesis is associated with a certain track at the previous moment, the track position is updated to be the current observed position, and the innovation covariance is updated according to the following formula:
Figure FDA0002920896690000051
Figure FDA0002920896690000052
if the flight path at the previous moment is not associated with any observation hypothesis at the current moment, carrying out flight path state classification according to the Life value of the flight path: if Life is greater than 0, considering that the flight path is temporarily interrupted, and updating the position prediction of the target at the current moment to be in the current position state, wherein the innovation covariance is unchanged, and Life is reduced by 1; if Life is equal to 0, the track is considered to be terminated, and the state of the track is not updated;
and (4) repeating the steps (3.1) to (3.4) and continuously generating new hypothesis nodes to form a hypothesis tree.
5. The sequential ISAR image scattering center multi-hypothesis tracking track association method according to claim 4, wherein the slave T is Tk-1Zhongren flight path
Figure FDA0002920896690000053
According to
Figure FDA0002920896690000054
The length of the path is used for completing the state transition parameter estimation and state prediction of the path, and specifically comprises the following steps:
(3.2a) when
Figure FDA0002920896690000055
Length 1:
due to the lack of more track information, the state prediction is performed as follows:
Figure FDA0002920896690000056
wherein the content of the first and second substances,
Figure FDA0002920896690000057
to represent
Figure FDA0002920896690000058
The position state prediction result in the k-th frame image,
Figure FDA0002920896690000059
representing the theoretical observed coordinates of the flight path in the k-th frame image,
Figure FDA00029208966900000510
to represent
Figure FDA00029208966900000511
Observing the actual position in the k-1 frame image;
(3.2b) when
Figure FDA00029208966900000512
When the length is greater than 1 and less than 4, the estimation of the state transition parameters is carried out by approximating the track motion by the circular motion around the image center o, specifically: is provided with
Figure FDA00029208966900000513
The motion of the observation point in the (k-2) th frame to the observation point in the (k-1) th frame image is approximately circular motion with the center o of the image as the center and the radius of the distance r from the observation point in the (k-1) th frame image to the point o, and the rotation angle thetak-1|k-2Taking the anticlockwise direction as positive; the motion of the flight path from the (k-1) th frame to the (k) th frame is approximated to be circular motion with o as the center of a circle and r as the radius, and the motion corners of adjacent frames are the same, namely thetak|k-1＝θk-1|k-2And calculating the position prediction of the flight path in the k frames as follows:
Figure FDA0002920896690000061
if it is
Figure FDA0002920896690000062
If there is no corresponding observation in the k frame image, then order
Figure FDA0002920896690000063
(3.2c) when
Figure FDA0002920896690000064
When the length is greater than or equal to 4, fitting the ellipse parameters by adopting a least square method, and calculating a transfer equation of the track position state;
traverse T as abovekL in (1)kAnd (4) tracking the flight path to complete the state prediction of all the flight paths.
6. The sequence ISAR image scattering center multi-hypothesis tracking track correlation method according to claim 5, wherein the method for fitting the ellipse parameters by using the least square method and calculating the transfer equation of the track position state comprises the following specific steps: the scattering center projection trajectory satisfies the elliptical motion equation,
Figure FDA0002920896690000065
the corresponding equation of motion is:
Figure FDA0002920896690000066
converting the above equation to a general form:
Figure FDA0002920896690000067
Figure FDA0002920896690000068
the following relationship is further obtained:
Figure FDA0002920896690000071
Figure FDA0002920896690000072
Figure FDA0002920896690000073
e2＝e4＝0
in the formula (I), the compound is shown in the specification,
Figure FDA0002920896690000074
to know
Figure FDA0002920896690000075
Parameters to be estimated in the state transition equation;
let e be [ e1，e3，e5，e6]TFormation matrix W [ u ], [ v ], [ 1 ]]Wherein u and v respectively represent an azimuth coordinate vector and a distance coordinate vector formed by all scattering center coordinates on the associated track, < > represents a Hadamard product, and 1 represents a column vector of all elements 1; if the flight path length is xi, then
Figure FDA0002920896690000076
Figure FDA0002920896690000077
Defining an error function:
F(e)＝(We)TWe＝eTWTWe
when the e parameter is the best fitting parameter of the ellipse, the value of F (e) is minimized, thus converting the solution of the ellipse parameter into the solution of the following constrained optimization problem:
arg min F(e)＝eTWTWe
s.t.eTFe＝1
wherein the content of the first and second substances,
Figure FDA0002920896690000078
the constraint optimization problem is solved by adopting a Lagrange multiplier method, and a Lagrange multiplier lambda is introduced and subjected to derivation to obtain the constraint optimization problem
2Fe-2λWTWe＝0
eTFe＝1
Let S be WTW, further obtaining
Figure FDA00029208966900000819
Solving by adopting a generalized eigenvalue solving method to obtain an eigenvalue and eigenvector pair (lambda)i，ei) It is clear that for any non-zero real number μ, (μ λ)i，μei) For a set of solutions that still satisfy the constraints, there are
Figure FDA0002920896690000081
Further obtaining an ellipse parameter set table
Figure FDA0002920896690000082
Then the elliptical optimal parameter estimate eoptComprises the following steps:
Figure FDA0002920896690000083
by means of eoptEach element of (A) and
Figure FDA0002920896690000084
and
Figure FDA0002920896690000085
by obtaining the relationship of
Figure FDA0002920896690000086
And
Figure FDA0002920896690000087
a value of (d); further obtaining a parameter A in the state transition equationk|k-1、bnThereby obtaining the state prediction of the target at the next moment
Figure FDA0002920896690000088
7. The method according to claim 4, wherein the calculating the probability that any current observation is an old track continuation, a new track and a false alarm specifically comprises: recording the probability that the track point is detected in the image as PdThe nth observation in the kth frame image
Figure FDA0002920896690000089
The probability associated with the mth track is:
Figure FDA00029208966900000810
wherein the content of the first and second substances,
Figure FDA00029208966900000811
is Gaussian distributed and has
Figure FDA00029208966900000812
Wherein the content of the first and second substances,
Figure FDA00029208966900000813
respectively representing the innovation and innovation covariance corresponding to the association:
Figure FDA00029208966900000814
Figure FDA00029208966900000815
Figure FDA00029208966900000816
Figure FDA00029208966900000817
and
Figure FDA00029208966900000818
respectively representing the noise covariance of the distance and the azimuth position in the observation process; furthermore, let PnewRepresenting the probability that a track point is a new track, PfThe probability that the track point is a false alarm is represented.
Suppose an observation set D of k frame imageskHas a total of NsAn observation, i group association hypothesis at previous time
Figure FDA0002920896690000091
The number of tracks in is NTIn a
Figure FDA0002920896690000092
On the basis of a priori assumptions on NsAssigning each observation, and constructing an assignment matrix omega:
Ω1(m，n)＝-ln(Pmn)
Ω2(m，n)＝-ln(Pnew)δ(m，n)
Ω3(m，n)＝-ln(Pf)δ(m，n)
Ω＝[Ω1，Ω2，Ω3]
and adopting a Murty algorithm to assign M groups of hypotheses with the maximum possibility to all current observations, wherein the j group of possibility hypotheses
Figure FDA0002920896690000093
Middle to NsOne observation is assigned: old track continuation NtNew born target NnewFalse alarm NfA plurality of; based on
Figure FDA0002920896690000094
Computing
Figure FDA0002920896690000095
Probability of being true:
Figure FDA0002920896690000096
calculating the probability of the assignment of the rest M-1 groups in the same way;
device set
Figure FDA0002920896690000097
To comprise
Figure FDA0002920896690000098
And the related assumptions made at all previous timesk-1Then the probability that its assumption is true is:
Figure FDA0002920896690000099
wherein c is a normalization coefficient, D1：kRepresents a set of all observations up to the k-th frame;
the probability that the remaining M-1 groups are assigned to be established is calculated in the same way, and a new layer of hypothetical tree nodes is formed on the hypothetical tree.
8. The sequence ISAR image scattering center multi-hypothesis tracking track association method according to claim 1, wherein the hypothesis tree management is performed by using an N backtracking pruning method, specifically: setting a Murty algorithm to select M optimal hypotheses each time, assuming that the depth of a tree is N, when the depth of the assumed tree is greater than N, selecting a node with the highest probability of establishment assigned by the hypotheses in all current leaf nodes, tracing back an N-1 layer upwards, taking the traced node as a new root hypothesis, and cutting off all side branches irrelevant to the new root hypothesis; the path correlation hypothesis in the newly generated root node is determined as the scattering center path.
9. The sequence ISAR image scattering center multi-hypothesis tracking track correlation method according to claim 1, wherein the correlation tracks in the forward direction and the backward direction are fused to realize correlation information complementation in the forward direction and the backward direction, and the method is implemented by the following steps:
(5.1) track matching
Setting the forward correlation track result as
Figure FDA0002920896690000101
Figure FDA0002920896690000102
The number is the ith' track, and I is the total number of tracks; the inversely correlated track results are
Figure FDA0002920896690000103
Definition of
Figure FDA0002920896690000104
Representing flight paths
Figure FDA0002920896690000105
And
Figure FDA0002920896690000106
counting the number of the overlapped tracks of the two tracks; definition ai′j′Representing flight paths
Figure FDA0002920896690000107
And
Figure FDA0002920896690000108
a matching relationship ofi′j′1 represents
Figure FDA0002920896690000109
Matching
Figure FDA00029208966900001010
ai′j′Is 0 represents
Figure FDA00029208966900001011
And
Figure FDA00029208966900001012
mismatch is not achieved; the track matching problem is converted into the following optimization problem:
Figure FDA00029208966900001013
Figure FDA00029208966900001014
solving the following optimization problem through a Hungarian algorithm to obtain a track matching relation set A ═ ai′j′＝1}；
(5.2) taking any element a in the set Ai′j′1, first rejecting
Figure FDA00029208966900001015
And
Figure FDA00029208966900001016
the first four frame results of (1); then go through
Figure FDA00029208966900001017
And
Figure FDA00029208966900001018
in the mutual exclusion frame time, the track point in the mutual exclusion frame time is taken as the fusion track ti′j′Track points of; then go through
Figure FDA00029208966900001019
To know
Figure FDA00029208966900001020
The track points with the same coordinates in the corresponding frame times are taken as the fusion track ti′j′Taking the coordinate mean value of the track points with different coordinates in the corresponding frame number as the fusion track ti′j′Predicting track points;
wherein the mutually exclusive frame is next to
Figure FDA00029208966900001021
And
Figure FDA00029208966900001022
of the frames appearing only once, the total number of frames being
Figure FDA00029208966900001023
To know
Figure FDA0002920896690000111
The number of frames present in each;
(5.3) repeating the step (5.2) until all elements of the set A are traversed to obtain a fused track Tfuse。
10. The method of claim 9, wherein the sequential ISAR images scatter center multi-hypothesis tracking track correlation method is characterized in thatThe Kalman filtering is performed on the fusion track by using a mode of TfusePerforming Kalman filtering along the positive direction and then performing Kalman filtering along the negative direction to obtain a filtering result T for each flight path in the systemfilterAnd obtaining the final track correlation result.
CN202110117517.9A 2021-01-28 2021-01-28 Sequence ISAR image scattering center multi-hypothesis tracking track correlation method Active CN112782696B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110117517.9A CN112782696B (en) 2021-01-28 2021-01-28 Sequence ISAR image scattering center multi-hypothesis tracking track correlation method

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110117517.9A CN112782696B (en) 2021-01-28 2021-01-28 Sequence ISAR image scattering center multi-hypothesis tracking track correlation method

Publications (2)

Publication Number Publication Date
CN112782696A true CN112782696A (en) 2021-05-11
CN112782696B CN112782696B (en) 2023-04-11

Family

ID=75759364

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110117517.9A Active CN112782696B (en) 2021-01-28 2021-01-28 Sequence ISAR image scattering center multi-hypothesis tracking track correlation method

Country Status (1)

Country Link
CN (1) CN112782696B (en)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114415180A (en) * 2022-03-30 2022-04-29 中国人民解放军火箭军工程大学 Stable tracking method fusing SAR high-resolution image and one-dimensional range profile
CN114779205A (en) * 2022-06-23 2022-07-22 北京海兰信数据科技股份有限公司 Radar-target-based multi-state parameter track matching processing method and system
CN117724064A (en) * 2024-02-07 2024-03-19 中国电子科技集团公司第十五研究所 Radar detection real-time evaluation method and device, electronic equipment and storage medium

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2004012351A (en) * 2002-06-07 2004-01-15 Mitsubishi Electric Corp Equipment, method, and program for tracking target
US8314394B1 (en) * 2009-11-04 2012-11-20 Science Applications International Corporation System and method for three-dimensional imaging using scattering from annihilation coincidence photons
CN105259553A (en) * 2015-11-11 2016-01-20 西安电子科技大学 Micro-motion target scattering point track association method based on distance-instant Doppler image
KR20160053520A (en) * 2014-11-05 2016-05-13 국방과학연구소 Efficient Method and Apparatus of Generating Inverse Synthetic Aperture Radar Image of Multiple Targets Using Flight Trajectory and Morphological Processing
CN108226928A (en) * 2017-12-18 2018-06-29 西安电子科技大学 Based on the inverse synthetic aperture radar imaging method for it is expected propagation algorithm
CN111208513A (en) * 2020-01-15 2020-05-29 西安电子科技大学 Space target ISAR image sequence energy back projection and three-dimensional reconstruction method

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2004012351A (en) * 2002-06-07 2004-01-15 Mitsubishi Electric Corp Equipment, method, and program for tracking target
US8314394B1 (en) * 2009-11-04 2012-11-20 Science Applications International Corporation System and method for three-dimensional imaging using scattering from annihilation coincidence photons
KR20160053520A (en) * 2014-11-05 2016-05-13 국방과학연구소 Efficient Method and Apparatus of Generating Inverse Synthetic Aperture Radar Image of Multiple Targets Using Flight Trajectory and Morphological Processing
CN105259553A (en) * 2015-11-11 2016-01-20 西安电子科技大学 Micro-motion target scattering point track association method based on distance-instant Doppler image
CN108226928A (en) * 2017-12-18 2018-06-29 西安电子科技大学 Based on the inverse synthetic aperture radar imaging method for it is expected propagation algorithm
CN111208513A (en) * 2020-01-15 2020-05-29 西安电子科技大学 Space target ISAR image sequence energy back projection and three-dimensional reconstruction method

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
T. SATHYAN ET AL.: "Multiple hypothesis tracking with multiframe assignment using range and range-rate measurements", 《14TH INTERNATIONAL CONFERENCE ON INFORMATION FUSION》 *
赵博 等: "基于电磁散射模型的ISAR空中目标欺骗干扰方法", 《电子与信息学报》 *

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114415180A (en) * 2022-03-30 2022-04-29 中国人民解放军火箭军工程大学 Stable tracking method fusing SAR high-resolution image and one-dimensional range profile
CN114415180B (en) * 2022-03-30 2022-07-01 中国人民解放军火箭军工程大学 Stable tracking method fusing SAR high-resolution image and one-dimensional range profile
CN114779205A (en) * 2022-06-23 2022-07-22 北京海兰信数据科技股份有限公司 Radar-target-based multi-state parameter track matching processing method and system
CN114779205B (en) * 2022-06-23 2022-09-13 北京海兰信数据科技股份有限公司 Radar-target-based multi-state parameter track matching processing method and system
CN117724064A (en) * 2024-02-07 2024-03-19 中国电子科技集团公司第十五研究所 Radar detection real-time evaluation method and device, electronic equipment and storage medium
CN117724064B (en) * 2024-02-07 2024-04-12 中国电子科技集团公司第十五研究所 Radar detection real-time evaluation method and device, electronic equipment and storage medium

Also Published As

Publication number Publication date
CN112782696B (en) 2023-04-11

Similar Documents

Publication Publication Date Title
CN112782696B (en) Sequence ISAR image scattering center multi-hypothesis tracking track correlation method
Loo et al. CNN-SVO: Improving the mapping in semi-direct visual odometry using single-image depth prediction
Almalioglu et al. SelfVIO: Self-supervised deep monocular Visual–Inertial Odometry and depth estimation
US7657059B2 (en) Method and apparatus for tracking an object
WO2012009947A1 (en) Device and method of signature-driven multi-target tracking
CN110967690B (en) Multi-target tracking method based on multiple Bernoulli distributed multiple sensors
Jaegle et al. Fast, robust, continuous monocular egomotion computation
US20130080111A1 (en) Systems and methods for evaluating plane similarity
Du et al. A new scatterer trajectory association method for ISAR image sequence utilizing multiple hypothesis tracking algorithm
CN104133983A (en) Multiple-target Bernoulli distribution type fusion method based on weighting negative first-order RD (Renyi Divergence)
CN115880364A (en) Robot pose estimation method based on laser point cloud and visual SLAM
CN115457086A (en) Multi-target tracking algorithm based on binocular vision and Kalman filtering
Park et al. Robust photogeometric localization over time for map-centric loop closure
White et al. An iterative pose estimation algorithm based on epipolar geometry with application to multi-target tracking
Liang et al. Neural enhanced belief propagation for data association in multiobject tracking
Ke et al. Deep multi-view depth estimation with predicted uncertainty
CN116630370A (en) Multi-model PBP-TPMB maneuver expansion target tracking method
Fleck et al. Robust tracking of reference trajectories for autonomous driving in intelligent roadside infrastructure
Iyer et al. CalibNet: Self-supervised extrinsic calibration using 3D spatial transformer networks
CN114529615A (en) Radar calibration method, device and storage medium
Kodaira et al. SST-Calib: simultaneous spatial-temporal parameter calibration between lidar and camera
Kim Active visual SLAM with exploration for autonomous underwater navigation
CN117433538A (en) Multi-source heterogeneous sensor track fusion method
CN117092638A (en) Multi-sensor information online data fusion method
Majdik et al. Micro air vehicle localization and position tracking from textured 3d cadastral models

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant