CN111797478B - Strong maneuvering target tracking method based on variable structure multi-model - Google Patents

Strong maneuvering target tracking method based on variable structure multi-model Download PDF

Info

Publication number
CN111797478B
CN111797478B CN202010734714.0A CN202010734714A CN111797478B CN 111797478 B CN111797478 B CN 111797478B CN 202010734714 A CN202010734714 A CN 202010734714A CN 111797478 B CN111797478 B CN 111797478B
Authority
CN
China
Prior art keywords
model
target
tracking
detector
measurement
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.)
Active
Application number
CN202010734714.0A
Other languages
Chinese (zh)
Other versions
CN111797478A (en
Inventor
李君龙
曹颖
高长生
尹童
张超
陈晓波
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.)
Harbin Institute of Technology
Beijing Institute of Electronic System Engineering
Original Assignee
Harbin Institute of Technology
Beijing Institute of Electronic System Engineering
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 Harbin Institute of Technology, Beijing Institute of Electronic System Engineering filed Critical Harbin Institute of Technology
Priority to CN202010734714.0A priority Critical patent/CN111797478B/en
Publication of CN111797478A publication Critical patent/CN111797478A/en
Application granted granted Critical
Publication of CN111797478B publication Critical patent/CN111797478B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/10—Geometric CAD
    • G06F30/15—Vehicle, aircraft or watercraft design
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/20—Analysis of motion
    • G06T7/246—Analysis of motion using feature-based methods, e.g. the tracking of corners or segments
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/20—Analysis of motion
    • G06T7/277—Analysis of motion involving stochastic approaches, e.g. using Kalman filters
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
    • G06F2119/14—Force analysis or force optimisation, e.g. static or dynamic forces
    • 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
    • Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
    • Y02T90/00—Enabling technologies or technologies with a potential or indirect contribution to GHG emissions mitigation

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Geometry (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Multimedia (AREA)
  • Mathematical Optimization (AREA)
  • Mathematical Analysis (AREA)
  • Pure & Applied Mathematics (AREA)
  • Computer Hardware Design (AREA)
  • Evolutionary Computation (AREA)
  • General Engineering & Computer Science (AREA)
  • Computational Mathematics (AREA)
  • Aviation & Aerospace Engineering (AREA)
  • Automation & Control Theory (AREA)
  • Radar Systems Or Details Thereof (AREA)

Abstract

A strong maneuvering target tracking method based on variable structure multi-models relates to the field of target tracking, and aims at the problem that when a high-speed strong maneuvering target in an adjacent space is tracked, the target tracking accuracy is low, and the method comprises the following steps: constructing a dynamic tracking model set by using the dynamic characteristics of the target aircraft, and then acquiring a state equation set of a maneuvering target tracking system; step two: establishing a system measurement model, and obtaining a measurement equation and measurement noise of the system according to the established system measurement model; step three: and carrying out recursive estimation on the motion state and the pneumatic parameters of the target aircraft based on the state equation set of the system, the measurement equation of the system and the measurement noise. According to the method, the dynamic tracking model set is constructed based on the dynamic characteristics of the target aircraft, the description precision of target motion is improved, and the target tracking precision is improved by adopting an improved variable-structure multi-model tracking algorithm.

Description

Strong maneuvering target tracking method based on variable structure multi-model
Technical Field
The invention relates to the field of target tracking, in particular to a strong maneuvering target tracking method based on variable structure multi-models.
Background
In the field of target tracking, a kalman filter is generally adopted for tracking, and an algorithm for correcting measured data by combining a target motion model and the measured data is adopted. Mainly solves the problems of three aspects: firstly, the measurement error of the observer comprises constant deviation and random deviation, and the instability and precision limit of the measured data are corrected to a certain extent in a target tracking algorithm; secondly, maneuvering of the target, which is a maneuvering mode of the target which cannot be accurately obtained because the target and a tracking party belong to different mechanisms, has high randomness and uncertainty, and can be used for distinguishing whether the target maneuvers according to residual errors in the tracking process so as to provide a reference basis; and thirdly, the sensor can only detect the position data of the target usually, the speed and related parameter data of the target cannot be acquired, and in a target tracking algorithm, the state vector of the target can be subjected to dimension expansion, so that the estimated value of the related parameter is obtained, the knowledge of the motion mode of the target is increased, and the accuracy of target tracking is also improved.
The near space field is an airspace 20-100 km away from the ground, the flight speed of the high-speed aircraft in the space can reach more than 5 Mach, and the near space contains rarefied atmosphere, so that the aircraft has certain maneuvering capability, and the parameters such as speed, acceleration and the like of the aircraft change violently. The vehicle can flexibly maneuver, quickly respond and super-strengthen the penetration, has strategic deterrence and actual combat application capabilities, and has important strategic significance for deterring strong enemies, controlling crisis and winning wars. In the defense field, the traditional tracking algorithm cannot realize accurate tracking, which has great difficulty in tracking and predicting the ultra-high-speed target in the adjacent space. Therefore, the method has very important military significance and practical significance for developing the near space strong maneuvering target tracking algorithm research.
At present, for tracking a strong maneuvering target, common models comprise a current statistical model, a Jerk model and other kinematic models, and a multi-model combined algorithm is often adopted for tracking a near space to perform fusion output so as to match different maneuvering modes of an aircraft. In the variable structure tracking algorithm, the IMM algorithm has the best tracking precision, but has certain limitation on the set value of the tracking model; the variable structure multi-model algorithm (VSMM algorithm) based on directed graph switching can solve the problem, but the delay problem is serious in model switching; the adaptive variable structure algorithm (AGIMM algorithm) is flexible in reaction and high in tracking precision, but overfitting occurs in the tracking and fitting process, so that the fitting result is changed quickly and unstably, and the final tracking error is increased.
Disclosure of Invention
The purpose of the invention is: aiming at the problem of low target tracking accuracy in the process of tracking a high-speed strong maneuvering target in an adjacent space, the strong maneuvering target tracking method based on the variable structure multi-model is provided.
The technical scheme adopted by the invention for solving the technical problems is as follows:
a strong maneuvering target tracking method based on variable structure multiple models comprises the following steps:
the method comprises the following steps: constructing a dynamic tracking model set by using the dynamic characteristics of the target aircraft, and then acquiring a state equation set of a maneuvering target tracking system;
step two: establishing a system measurement model, and obtaining a measurement equation and measurement noise of the system according to the established system measurement model;
step three: and carrying out recursive estimation on the motion state and the pneumatic parameters of the target aircraft based on the state equation set of the system, the measurement equation of the system and the measurement noise.
Further, the specific steps of the first step are as follows:
the method comprises the following steps: selecting the pneumatic parameters of the tracking model according to the maneuvering characteristics of the target,
the tracking model pneumatic parameters are as follows:
Figure BDA0002604484020000021
wherein, C L (. Alpha.) and C D (α) is the pneumatic parameter, C L (α) is the coefficient of lift, C D (α) is the drag coefficient, S is the characteristic area of the target aircraft, m is the mass of the target aircraft, α D ，α L Resistance parameters and lift parameters;
the first step is: modeling the change characteristic of the pneumatic parameters by using Gaussian white noise to obtain a state equation of the maneuvering target tracking system, wherein the state equation of the maneuvering target tracking system is as follows:
Figure BDA0002604484020000022
wherein, ω is e Is the angular velocity vector of the earth rotation, g is the gravity acceleration of the earth, r is the position vector of the target under the detection system, v is the velocity of the target under the detection system,
Figure BDA0002604484020000023
is a transfer matrix from a ballistic coordinate system to a detection system, theta is a velocity inclination angle, sigma is an azimuth angle, and gamma is v Is the angle of inclination of the speed, omega γ White gaussian noise, omega, at the roll angle D ，ω L White Gaussian noise as an aerodynamic parameter, R is aerodynamic force, wherein
Figure BDA0002604484020000024
S is the characteristic area of the target aircraft:
Figure BDA0002604484020000025
step one is three: the roll angle gamma of the speed in the first step and the second step v As model variables, different speed roll angles γ are selected v And constructing a state equation set of the maneuvering target tracking system.
Further, the step of obtaining the measurement equation and the measurement noise of the system in the step two specifically includes:
step two, firstly: establishing a detection coordinate system according to the requirement of the tracking task, and determining position vectors of the detector and the target under the detection coordinate system;
step two: acquiring a three-dimensional position coordinate of a target aircraft in a detector body system according to an infrared detection principle;
step two and step three: and (3) performing expansion analysis on the mean square error of positioning by using the three-dimensional position coordinates of the target aircraft in the system of the detector, and determining a measurement equation and measurement noise of the tracking system.
Further, in the first step, the position vectors of the detector and the target in the detection coordinate system are:
position vector of the target under the detection coordinate system: r = (x, y, z);
position vector of the detector under the detection coordinate system: s l ＝(x l ,y l ,z l ) And l represents the l-th detector.
Further, the second step specifically comprises the following steps:
first, the vector of the probe pointing to the target is:
Figure BDA0002604484020000031
detector detection angle alpha l And beta l Expressed as:
Figure BDA0002604484020000032
and (3) conversion obtaining:
Figure BDA0002604484020000033
Figure BDA0002604484020000034
Figure BDA0002604484020000035
Figure BDA0002604484020000036
obtaining three-dimensional position coordinates X = (X, y, z) of the target aircraft under the detector by using a least square method
The least squares are expressed as:
Figure BDA0002604484020000037
where M is the measurement matrix, X is the state quantity, and Y is the quantity measurement.
Further, the specific steps of determining the measurement equation and the measurement noise of the tracking system in the second step and the third step are as follows:
firstly, determining the relative position and angle relationship between the target and the detector according to the geometric principle:
Figure BDA0002604484020000041
in the formula,
Figure BDA0002604484020000042
x 1 ，y 1 ，z 1 the position components of the first detector in the x direction, the y direction and the z direction under a detection system are obtained; x is the number of 2 ，y 2 ，z 2 The position components of the second detector in the x direction, the y direction and the z direction under a detection system are obtained; Δ κ = κ 2 -κ 1 ，
Figure BDA0002604484020000043
Then, obtaining noise R through a measurement noise formula, wherein the measurement noise formula is as follows:
Figure BDA0002604484020000044
wherein,
Figure BDA0002604484020000045
Figure BDA0002604484020000046
Figure BDA0002604484020000047
in the above formula, c 1 ＝κ 2 (x 2 -x 1 )-(y 2 -y 1 )，c 2 ＝-κ 1 (x 2 -x 1 )+(y 2 -y 1 )；
Figure BDA0002604484020000048
Figure BDA0002604484020000051
And
Figure BDA0002604484020000052
respectively the mean square error of the position coordinates of the detector itself,
Figure BDA0002604484020000053
the mean square error is located for the target,
Figure BDA0002604484020000054
for the mean square error of the detector detection angle alpha 1,
Figure BDA0002604484020000055
the mean square error of the detector detection angle alpha 2,
Figure BDA0002604484020000056
the mean square error of the angle beta 1 is detected for the detector,
Figure BDA0002604484020000057
the mean square error of the angle β 2 is detected for the detector.
Further, in the third step, an improved variable structure multi-model algorithm is utilized to carry out recursive estimation on the motion state and the pneumatic parameters of the target aircraft; the improved variable structure multi-model algorithm comprises the following specific steps:
first, input interaction:
Figure BDA0002604484020000058
m is the set of the model totality, i, j represents the ith, j model,
Figure BDA0002604484020000059
wherein p is ij Is the transition probability, μ, of model i transitioning to j i,j (k-1/k-1) is a mixed probability
Figure BDA00026044840200000510
Figure BDA00026044840200000511
Second step, filtering
For model M j (k) To be provided with
Figure BDA00026044840200000512
Performing Kalman filtering, phi j For the state transition matrix, the prediction is:
Figure BDA00026044840200000513
the prediction error covariance is:
Figure BDA00026044840200000514
the filtering is as follows:
Figure BDA00026044840200000515
Figure BDA00026044840200000516
the filtering covariance is:
Figure BDA00026044840200000517
and (4) estimating an observation equation, wherein the Kalman gain is as follows:
Figure BDA0002604484020000061
Figure BDA0002604484020000062
third, model probability updating
Figure BDA0002604484020000063
Wherein, Λ j (k)＝N(r j (k),0,S j (k) K) likelihood function of the pattern j at time k, defining a variable r j (k) Mean 0 and variance S j (k) Gaussian distribution of Λ j (k) Comprises the following steps:
Figure BDA0002604484020000064
step four, outputting interaction:
Figure BDA0002604484020000065
Figure BDA0002604484020000066
fifthly, updating the model set:
calculating the maximum model probability:
u max ＝max{u 1 ,u 2 ,u 3 }
and the corresponding model is set as a model j, the distance function is as follows:
Figure BDA0002604484020000067
tracking model adjustment, which is performed by analogy with the principle of directed graph switching,
when u1 max When, if D 1 (k) Not less than M, then
Figure BDA0002604484020000068
Wherein G is 0 Is the minimum grid spacing, k 1 ，k 2 In order to be able to adjust the parameters,
if D is 1 (k) If < M, updating the model according to a method of switching a directed graph, namely
Figure BDA0002604484020000069
When u2 max When, if D 2 (k) Not less than M, then
Figure BDA0002604484020000071
If D is 2 (k) If < M, updating the model according to a method of directed graph switching, namely
Figure BDA0002604484020000072
When u3 max When, if D 3 (k) Not less than M, then
Figure BDA0002604484020000073
If D is 3 (k) If < M, updating the model according to a method of switching a directed graph, namely
Figure BDA0002604484020000074
For the initialization of the state vector and covariance of the newly activated model, the probability weighted combination of each model at the previous moment is adopted for initialization, that is:
Figure BDA0002604484020000075
Figure BDA0002604484020000076
the invention has the beneficial effects that:
the method constructs a dynamic tracking model set based on the dynamic characteristics of the target aircraft, improves the description precision of target motion, further improves the target tracking precision by adopting an improved variable-structure multi-model tracking algorithm, and improves the robustness of target maneuvering by updating the model and the corresponding probability by detecting the target maneuvering.
Drawings
FIG. 1 is a schematic block diagram of a variable structure multi-model algorithm tracking;
FIG. 2 is a diagram of the actual motion trajectory of a target in a simulation example;
FIG. 3 is a diagram illustrating a variation of an actual movement velocity and a roll angle of a target in a simulation example;
FIG. 4 is a tracking error diagram of a variable structure tracking algorithm in a weak maneuvering state;
FIG. 5 is a diagram of the error of the fit of each algorithm to the target roll angle under weak maneuvering conditions;
FIG. 6 is a tracking error diagram of a variable structure tracking algorithm under a strong maneuvering condition;
FIG. 7 is a diagram of the error of the fit of each algorithm to the target roll angle under a strong maneuver condition.
Detailed Description
The first embodiment is as follows:
the embodiment is specifically described by referring to fig. 1, and the invention aims to improve the tracking and positioning accuracy of a high-speed maneuvering target in an adjacent space and predict the pneumatic characteristic parameters of the target and detect the maneuvering target by combining the tracking principle. Expanding the pneumatic characteristic parameters of the target into a target tracking state quantity by establishing an aircraft dynamics model, and carrying out filtering tracking; and simultaneously, tracking data of the target is obtained by combining a variable structure multi-model algorithm based on maneuvering detection. The above object is achieved by the following technical scheme:
in step 1, a dynamic tracking model set is constructed according to the dynamic characteristics of a target aircraft, and a state equation set of a maneuvering target tracking system is obtained.
In step 2 of the invention, a system measurement model is established according to the principle and distribution of the detection devices; a measurement equation and measurement noise for the system are obtained. The infrared observed quantity is the azimuth angle of the target, belongs to passive direction finding positioning, and the target is positioned through two or more infrared detectors.
In step 3, based on a state equation of the system, a measurement equation of the system and measurement noise, the motion state and the control parameters of the target aircraft are recursively estimated by using an improved variable structure multi-model algorithm based on target maneuver detection.
In the invention, the specific method for constructing the dynamic tracking model set according to the dynamic characteristics of the target aircraft and acquiring the state equation set of the maneuvering target tracking system in the step 1 is as follows:
step 1-1, analyzing target maneuvering characteristics, selecting and tracking pneumatic parameters of a model:
the maneuvering of the aircraft in the free glide phase mainly comes from aerodynamic force in such a way that aerodynamic acceleration can be divided into resistive acceleration, turning acceleration and climbing acceleration along three directions of a ballistic system through pairs, namely:
Figure BDA0002604484020000081
from this equation, the aerodynamic parameter C L (alpha) Presence with pneumatic parameters
Figure BDA0002604484020000082
The relationship (c) in (c). The frequent and great adjustment of range of angle of attack alpha of flight can lead to that pneumatic parameter changes are complicated, the aircraft is violent to vibrate, is difficult to control, and the rate of change of alpha is very little in the process of usually gliding, and pneumatic parameter is comparatively mild along with the angle of attack change, and the parameter variation is less in the short time promptly, can regard as the state quantity to estimate, and the pneumatic parameter in the tracking model of so here sets up to:
Figure BDA0002604484020000091
wherein S is the characteristic area of the target aircraft, and m is the mass of the target aircraft. Alpha is alpha D ，α L Are a drag parameter and a lift parameter. Modeling the change characteristic of the pneumatic parameters by using Gaussian white noise to obtain a state equation of the maneuvering target tracking system:
Figure BDA0002604484020000092
wherein gamma is v Is the angle of inclination of the speed, omega γ Is white gaussian noise at the roll angle,ω D ，ω L the Gaussian white noise is an aerodynamic parameter, R is aerodynamic force, and the expression is as follows:
Figure BDA0002604484020000093
the invention inclines the speed by the angle gamma v As model variables, different γ's were selected v To construct a set of mobile target tracking models. Further, in the present invention, the specific method for obtaining the measurement equation and the measurement noise of the system in step 2 is:
step 2-1, establishing a detection system according to the tracking task requirement, and determining position vectors of a detector and a target under a detection coordinate system;
2-2, acquiring a three-dimensional position coordinate of the target aircraft under a detector according to an infrared detection principle, and realizing positioning of the target aircraft;
and 2-3, carrying out expansion analysis on the positioning mean square error according to the three-dimensional position coordinates of the target aircraft under the detector, and determining a measurement equation and measurement noise of the tracking system.
Further, in the invention, a detection coordinate system is established according to the position of the detector in the step 2-1, and the position vectors of the base point of the detector and the target in the detection coordinate system are determined;
position vector of the target under the detection system: r = (x, y, z);
position vector of the base point of the detector under the detection system: s l ＝(x l ,y l ,z l ) And l represents the l-th detector;
the vector pointed to the target by the detector is: r l ＝r-S l ＝(x-x l ,y-y l ,z-z l )。
Further, in the present invention, the step 2-2 of obtaining the three-dimensional position coordinates of the target aircraft under the detector according to the infrared detection principle includes the specific steps of:
let the target aircraft be at a distance from the probe:
Figure BDA0002604484020000094
due to the detection angle alpha of the detector l And beta l Comprises the following steps:
Figure BDA0002604484020000101
and (3) conversion obtaining:
Figure BDA0002604484020000102
Figure BDA0002604484020000103
Figure BDA0002604484020000104
Figure BDA0002604484020000105
the least square method is used as follows:
Figure BDA00026044840200001011
and obtaining three-dimensional position coordinates X = (X, y, z) of the target aircraft under the detector.
Further, in the present invention, the specific method for determining the noise measured by the tracking system in step 2-3 is as follows:
determining according to geometric principles:
Figure BDA0002604484020000106
in the formula,
Figure BDA0002604484020000107
wherein x is 1 ，y 1 ，z 1 The position components of the first detector in the x direction, the y direction and the z direction under a detection system are obtained; x is the number of 2 ，y 2 ，z 2 The position components of the second detector in the x direction, the y direction and the z direction under the detection system are obtained; Δ κ = κ 2 -κ 1 ，
Figure BDA0002604484020000108
By measuring the noise formula:
Figure BDA0002604484020000109
the noise R is obtained, wherein,
Figure BDA00026044840200001010
Figure BDA0002604484020000111
Figure BDA0002604484020000112
in the above formula, c 1 ＝κ 2 (x 2 -x 1 )-(y 2 -y 1 )，c 2 ＝-κ 1 (x 2 -x 1 )+(y 2 -y 1 )；
Figure BDA0002604484020000113
Figure BDA0002604484020000114
And
Figure BDA0002604484020000115
respectively the mean square error of the position coordinates of the detector itself,
Figure BDA0002604484020000116
the mean square error is located for the target,
Figure BDA0002604484020000117
the mean square error of the detector detection angle alpha 1,
Figure BDA0002604484020000118
the mean square error of the detector detection angle alpha 2,
Figure BDA0002604484020000119
the mean square error of the angle beta 1 is detected for the detector,
Figure BDA00026044840200001110
the mean square error of the angle β 2 is detected for the detector.
Further, in step 3 of the invention, the motion state and the pneumatic parameters of the target aircraft are recursively estimated by using an improved variable structure multi-model algorithm based on target maneuver detection:
determining initial state quantity and initial covariance of a filter:
Figure BDA00026044840200001111
wherein,
Figure BDA00026044840200001112
as initial state quantity of the filter, E (x) 0 ) The mean value of the initial state quantities of the target aircraft is obtained; taking the mean value
Figure BDA00026044840200001113
Is the initial covariance, x 0 Is the initial state quantity of the target aircraft;
the mathematical expectation of a matrix of random variables is defined as the matrix of the mathematical expectation of their respective elements, E (X) = μ. The covariance matrix of the multidimensional random variables is a symmetric matrix and plays a role of one-dimensional random variable variance in a certain sense. The covariance matrix C of X can be expressed as:
C＝E((X-μ) T (X-μ))
is provided with
Figure BDA0002604484020000121
Then:
Figure BDA0002604484020000122
Cov(X 1 ,X 2 )＝ρσ 1 σ 1
then (X) 1 ,X 2 ) Has a covariance matrix of
Figure BDA0002604484020000123
The probability density is:
Figure BDA0002604484020000124
c has a determinant of
Figure BDA0002604484020000125
Inverse matrix is
Figure BDA0002604484020000126
Let x = (x) 1 ,x 2 )，
Then there are:
Figure BDA0002604484020000127
thus (X) 1 ,X 2 ) The probability density of (d) can be expressed as:
Figure BDA0002604484020000128
if n-dimensional random variable X = (X) 1 ,X 2 ,…X n ) The probability density of (a) is:
Figure BDA0002604484020000129
let n j For events where model j is correct, with a prior probability
Figure BDA00026044840200001210
j =1,2 \ 8230r, then under the premise of the model j, the likelihood function of the measured data at the time k is:
Figure BDA00026044840200001211
wherein
Figure BDA00026044840200001212
Innovation residual calculated for filter j
Figure BDA00026044840200001213
The probability density of (a) is as follows:
Figure BDA0002604484020000131
under the guidance of Bayes theory, the posterior probability density of the model j at the moment k is as follows:
Figure BDA0002604484020000132
1. variable structure multi-model algorithm
The first step, input interaction:
Figure BDA0002604484020000133
Figure BDA0002604484020000134
wherein p is ij Is the transition probability of model i transitioning to j. Mu.s i,j (k-1/k-1) Mixed probability
Figure BDA0002604484020000135
Figure BDA0002604484020000136
Second step, filtering
Corresponding model M j (k) To be provided with
Figure BDA0002604484020000137
Performing Kalman filtering
And (3) prediction:
Figure BDA0002604484020000138
prediction error covariance:
Figure BDA0002604484020000139
filtering:
Figure BDA00026044840200001310
Figure BDA00026044840200001311
filtering covariance:
Figure BDA00026044840200001312
and (3) estimating an observation equation:
kalman gain:
Figure BDA0002604484020000141
Figure BDA0002604484020000142
third, model probability updating
Figure BDA0002604484020000143
Λ j (k)＝N(r j (k),0,S j (k) K) likelihood function of the time pattern j, a variable r is defined j (k) Mean 0 and variance S j (k) Gaussian distribution (also called normal distribution).
Figure BDA0002604484020000144
Fourth, output interaction
Figure BDA0002604484020000145
Figure BDA0002604484020000146
Fifthly, updating the model set
And further obtaining a corresponding model decision according to the model probability updating, wherein the model decision is a model self-adaptive process depending on a set model transformation algorithm and is used for filtering at the next moment. For the initialization of the state vector and covariance of the newly activated model, the probability weighted combination of each model at the previous moment is adopted for initialization, that is:
Figure BDA0002604484020000147
Figure BDA0002604484020000148
2. improved variable structure multi-model method
The variable structure tracking algorithm has a good tracking effect on the problem of strong maneuvering, and mainly comprises the steps of working a plurality of tracking filters simultaneously, calculating the output probability of each filter according to the residual error and the covariance of each filter, and finally combining the interactive output result of each filter to serve as tracking data. The specific flow is shown in fig. 1.
Aiming at the tracking of the strong maneuvering flying target, whether the target has strong maneuvering or not is detected in the tracking process, and the filtering innovation and the covariance thereof in the target tracking process are utilized to carry out discrimination detection by combining a variable structure multi-model algorithm. Past information should be forgotten much at the maneuvering time; the past information is used for a large amount of non-maneuvering time, and the model is adjusted in a self-adaptive mode by using the past information, so that the model parameters are closer to a real model, and the tracking precision of the target non-maneuvering time is improved. The tracking result of the improved multi-model algorithm (GIMM algorithm) is between the IMM algorithm and the AGIMM algorithm, but compared with the IMM algorithm, the improved multi-model algorithm is not limited by a preset model, and compared with the AGIMM algorithm, the improved multi-model algorithm is relatively stable without the phenomenon of overfitting.
The specific method comprises the following steps:
during the filtering process, the innovation r k+1|k Sum innovation covariance S k+1|k And the error of the one-step prediction is shown, when the one-step prediction has a large error, the target is most likely to have strong maneuvering, and the tracking system does not detect and make corresponding adjustment at the moment. Therefore, after the filtering is finished at each moment, the following are calculated: u _ max = max (u) L ,u C ,u R ) Model generation with maximum posterior probabilityRepresenting the model most consistent at the moment, if the innovation of the corresponding model is still large, the aircraft is considered to have strong maneuver, at the moment, whether the aircraft has the strong maneuver needs to be judged, and D (k) = r (k) is calculated T S(k) -1 r (k) representing the innovation of the maximum probability model at the time k; s (k) represents the innovation covariance of the maximum probability model at time k. And reasonably setting a threshold value M, and if D (k) > M, considering that the target has strong maneuvering.
The specific content of the method is as follows:
and after k filtering is finished at a certain moment, obtaining innovation corresponding to the maximum probability model and covariance of the innovation.
Calculating the maximum model probability:
u max ＝max{u 1 ,u 2 ,u 3 }
the corresponding model is set as model j, then the distance function
Figure BDA0002604484020000151
Adjusting a tracking model:
the model adjustment is performed by analogy to the principle of directed graph switching.
When u1 max When, if D 1 (k) Not less than M, then
Figure BDA0002604484020000152
Wherein G is 0 Is the minimum grid spacing, k 1 ，k 2 Is an adjustable parameter.
If D is 1 (k) If < M, updating the model according to a method of directed graph switching, namely
Figure BDA0002604484020000153
When u2 max When, if D 2 (k) Not less than M, then
Figure BDA0002604484020000161
If D is 2 (k) If < M, updating the model according to a method of directed graph switching, namely
Figure BDA0002604484020000162
When u3 max When, if D 3 (k) Not less than M, then
Figure BDA0002604484020000163
If D is 3 (k) If < M, updating the model according to a method of switching a directed graph, namely
Figure BDA0002604484020000164
Example (b):
the computer adopted by the simulation of the embodiment of the invention is configured as follows: the CPU is i7-8550U, the main frequency is 1.80GHz, and the memory is 8GB.
Simulation scene: the target does maneuvering motion in the adjacent space, the motion trail is shown in fig. 2, the attack angle of the target changes slowly in the process, the maneuvering condition of the target is mainly realized according to the change of the roll angle of the moving speed of the target, and the change of the roll angle is shown in fig. 3.
The method is characterized in that multiple multi-model algorithms are combined with a target tracking dynamic model to realize the tracking of a near space strong maneuvering target, multiple Monte Carlo simulations are carried out, and the precision, the convergence speed and the stability of each algorithm are mainly analyzed and compared.
The target tracking state quantity is X = [ X y z v = [ ] x v y v z α D α L ] T Angle of inclination gamma v The angular velocity in the circular turning model is used as a multi-model distinguishing value, a ckf filtering method is adopted for filtering and tracking, andthe velocity roll angle used by each model is output alternately as an estimated value of the roll angle, which is one of the determination tracking accuracies.
The measurement adopts double infrared detection, the height of the original point of the infrared detection system is 5000m from the ground, the infrared detection system is coincident with the local geographic coordinate system, the north-sky-east coordinate system is adopted, the distance between the two infrared detectors is 900km, the detection distance of each infrared detector is 1100km, and the detection angle error is 6 multiplied by 10 -4 rad。
The algorithm tracking performance index is expressed by Normalized Position Error (NPE), which is the ratio of Root Mean Square Error (RMSE) of position filtering to RMSE of position measurement, and is specifically expressed as follows:
Figure BDA0002604484020000171
wherein M is the simulation times of the Monte Carlo.
Simulation one: and tracking the target for the time period of 590-640s, wherein the target is in a weak maneuvering state, the tracking result is shown in fig. 4, and the tilt angle fitting error is shown in fig. 5. The following are NEP error cases:
TABLE 1 comparison of simulation results
Figure BDA0002604484020000172
Simulation II: and tracking the target for 540-590s, wherein the target is in a strong maneuvering state, the tracking result is shown in fig. 6, and the tilt angle fitting error is shown in fig. 7. The following are simulation results of the NEP error case:
TABLE 1 tracking result comparison
Figure BDA0002604484020000173
From the above simulation images and data, it can be seen that:
1) Tracking precision: when the target maneuverability is not strong, the tracking errors of the four variable structure multi-model tracking methods reach within one hundred meters, and the tracking precision is kept about one hundred meters when strong maneuverability occurs; and comprehensively comparing, the GIMM algorithm has the best comprehensive tracking precision in the aspects of weak maneuver and strong maneuver.
2) In the convergence rate aspect: the AGIMM algorithm reacts fastest under the condition of strong maneuvering and can keep up with maneuvering change of a target, and the GIMM algorithm is adopted secondly, but the methods have no great difference in data processing time.
3) Stability aspect: the four methods can be adjusted at any time according to the target maneuvering condition in the filtering process, can ensure that the tracking error is stable in a certain range, and has no dispersion condition.
The method is used for discussing the tracking problem of the near space strong maneuvering aircraft, and a target dynamic model and a variable structure multi-model method are combined to position and track the target. The tracking accuracy is improved, the pneumatic characteristic and the maneuvering state of the target are obtained, and the method is greatly helpful for recognizing the target and forecasting the track.
From the above, the embodiment of the invention realizes the problem of strong maneuvering aircraft tracking in the near space. However, the above description is only a preferred embodiment of the present invention, and is not intended to limit the present invention, and the present invention is also applicable to other related tracking problems. Any modification, equivalent replacement, and improvement made within the spirit and scope of the present invention shall be included in the protection scope of the present invention.
It should be noted that the detailed description is only for explaining and explaining the technical solution of the present invention, and the scope of protection of the claims is not limited thereby. It is intended that all such modifications and variations be included within the scope of the invention as defined in the following claims and the description.

Claims (5)

1. A strong maneuvering target tracking method based on variable structure multiple models is characterized by comprising the following steps:
the method comprises the following steps: constructing a dynamic tracking model set by using the dynamic characteristics of the target aircraft, and then acquiring a state equation set of a maneuvering target tracking system;
step two: establishing a system measurement model, and obtaining a measurement equation and measurement noise of the system according to the established system measurement model;
step three: carrying out recursive estimation on the motion state and the pneumatic parameters of the target aircraft based on the state equation set of the system, the measurement equation of the system and the measurement noise;
the step two of obtaining the measurement equation and the measurement noise of the system specifically comprises the following steps:
step two, firstly: establishing a detection coordinate system according to the requirement of the tracking task, and determining position vectors of the detector and the target under the detection coordinate system;
step two: acquiring a three-dimensional position coordinate of a target aircraft in a detector body system according to an infrared detection principle;
step two and step three: utilizing the three-dimensional position coordinates of the target aircraft in the system of the detector to perform expansion analysis on the positioning mean square error, and determining a measurement equation and measurement noise of the tracking system;
in the third step, the motion state and the pneumatic parameters of the target aircraft are recursively estimated by using an improved variable structure multi-model algorithm; the improved variable structure multi-model algorithm comprises the following specific steps:
first, input interaction:
Figure FDA0003835090150000011
m is the set of the model totality, i, j represents the ith, j model,
Figure FDA0003835090150000012
wherein p is ij Is the transition probability, μ, of model i transitioning to j i,j (k-1/k-1) is a mixing probability
Figure FDA0003835090150000013
Figure FDA0003835090150000014
And step two, filtering:
for model M j (k) To be provided with
Figure FDA0003835090150000015
Performing Kalman filtering, phi j For the state transition matrix, the prediction is:
Figure FDA0003835090150000016
the prediction error covariance is:
Figure FDA0003835090150000021
the filtering is:
Figure FDA0003835090150000022
Figure FDA0003835090150000023
the filtering covariance is:
Figure FDA0003835090150000024
and (4) estimating an observation equation, wherein the Kalman gain is as follows:
Figure FDA0003835090150000025
Figure FDA0003835090150000026
thirdly, updating model probability:
Figure FDA0003835090150000027
Λ j (k)＝N(r j (k),0,S j (k) K) the likelihood function of the pattern j at time k, defines a variable r j (k) Mean 0 and variance S j (k) (ii) a gaussian distribution of;
Figure FDA0003835090150000028
step four, outputting interaction:
Figure FDA0003835090150000029
Figure FDA00038350901500000210
fifthly, updating the model set:
calculating the maximum model probability:
u max ＝max{u 1 ,u 2 ,u 3 }
and the corresponding model is set as a model j, the distance function is as follows:
Figure FDA00038350901500000211
tracking model adjustment, which is performed by analogy with the principle of directed graph switching,
when u1 max When, if D 1 (k) Not less than M, then
Figure FDA00038350901500000212
Wherein G 0 Is the minimum grid spacing, k 1 ，k 2 In order to be an adjustable parameter, the device is provided with a power supply,
if D is 1 (k) If < M, updating the model according to a method of directed graph switching, namely
Figure FDA0003835090150000031
When u2 max When, if D 2 (k) Not less than M, then
Figure FDA0003835090150000032
If D is 2 (k) If < M, updating the model according to a method of switching a directed graph, namely
Figure FDA0003835090150000033
When u3 max When, if D 3 (k) Not less than M, then
Figure FDA0003835090150000034
If D is 3 (k) If < M, updating the model according to a method of switching a directed graph, namely
Figure FDA0003835090150000035
For the initialization of the state vector and covariance of the newly activated model, the probability weighted combination of each model at the previous moment is adopted for initialization, that is:
Figure FDA0003835090150000036
Figure FDA0003835090150000037
2. the method for tracking the strong maneuvering target based on the variable structure multiple models as claimed in claim 1, characterized in that the specific steps of the first step are as follows:
the method comprises the following steps: selecting the pneumatic parameters of the tracking model according to the maneuvering characteristics of the target,
the pneumatic parameters of the tracking model are as follows:
Figure FDA0003835090150000041
wherein, C L (. Alpha.) and C D (α) is the pneumatic parameter, C L (α) is the coefficient of lift, C D (α) is the drag coefficient, S is the characteristic area of the target aircraft, m is the mass of the target aircraft, α D ，α L Resistance parameters and lift parameters;
the first step is: modeling the change characteristic of the pneumatic parameters by using Gaussian white noise to obtain a state equation of the maneuvering target tracking system, wherein the state equation of the maneuvering target tracking system is as follows:
Figure FDA0003835090150000042
wherein, ω is e Is the angular velocity vector of the earth rotation, g is the gravity acceleration of the earth, r is the position vector of the target under the detection system, v is the velocity of the target under the detection system,
Figure FDA0003835090150000043
is a transfer matrix from a ballistic coordinate system to a detection system, theta is a velocity inclination angle, sigma is an azimuth angle, and gamma is v Is the angle of inclination of the speed, omega γ White gaussian noise, omega, at the roll angle D ，ω L White Gaussian noise as an aerodynamic parameter, R is aerodynamic force, wherein
Figure FDA0003835090150000044
S is the characteristic area of the target aircraft:
Figure FDA0003835090150000045
step one is three: the roll angle gamma of the speed in the first step and the second step v As model variables, different speed roll angles γ are selected v And constructing a state equation set of the maneuvering target tracking system.
3. The method for tracking the strong maneuvering target based on the variable structure multiple models according to claim 1, characterized in that in the first step, the position vectors of the detector and the target under the detection coordinate system are:
position vector of the target under the detection coordinate system: r = (x, y, z);
position vector of the detector under the detection coordinate system: s. the l ＝(x l ,y l ,z l ) And l represents the l-th detector.
4. The variable structure multi-model-based strong maneuvering target tracking method according to claim 3, characterized in that the second step comprises the specific steps of:
first, the vector of the probe pointing to the target is:
Figure FDA0003835090150000046
detector probeAngle measurement alpha l And beta l Expressed as:
Figure FDA0003835090150000051
the conversion obtains:
Figure FDA0003835090150000052
Figure FDA0003835090150000053
Figure FDA0003835090150000054
Figure FDA0003835090150000055
obtaining three-dimensional position coordinates X = (X, y, z) of the target aircraft under the detector by using a least square method
The least squares are expressed as:
Figure FDA0003835090150000056
where M is the measurement matrix, X is the state quantity, and Y is the quantity measurement.
5. The variable structure multi-model-based strong maneuvering target tracking method according to claim 4, characterized in that the specific steps of determining the measurement equation and the measurement noise of the tracking system in the second and third steps are as follows:
firstly, determining the relative position and angle relationship between the target and the detector according to the geometric principle:
Figure FDA0003835090150000057
in the formula,
Figure FDA0003835090150000058
x 1 ，y 1 ，z 1 the position components of the first detector in the x direction, the y direction and the z direction under a detection system are obtained; x is the number of 2 ，y 2 ，z 2 The position components of the second detector in the x direction, the y direction and the z direction under the detection system are obtained; Δ κ = κ 2 -κ 1 ，
Figure FDA0003835090150000059
Then, noise R is obtained through a measurement noise formula, wherein the measurement noise formula is as follows:
Figure FDA00038350901500000510
wherein,
Figure FDA0003835090150000061
Figure FDA0003835090150000062
Figure FDA0003835090150000063
in the above formula, c 1 ＝κ 2 (x 2 -x 1 )-(y 2 -y 1 )，c 2 ＝-κ 1 (x 2 -x 1 )+(y 2 -y 1 )；
Figure FDA0003835090150000064
Figure FDA0003835090150000065
Figure FDA0003835090150000066
And
Figure FDA0003835090150000067
respectively the mean square error of the position coordinates of the detector itself,
Figure FDA0003835090150000068
the mean square error is located for the target,
Figure FDA0003835090150000069
the mean square error of the detector detection angle alpha 1,
Figure FDA00038350901500000610
the mean square error of the detector detection angle alpha 2,
Figure FDA00038350901500000611
the mean square error of the angle beta 1 is detected for the detector,
Figure FDA00038350901500000612
the mean square error of the angle β 2 is detected for the detector.
CN202010734714.0A 2020-07-27 2020-07-27 Strong maneuvering target tracking method based on variable structure multi-model Active CN111797478B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010734714.0A CN111797478B (en) 2020-07-27 2020-07-27 Strong maneuvering target tracking method based on variable structure multi-model

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010734714.0A CN111797478B (en) 2020-07-27 2020-07-27 Strong maneuvering target tracking method based on variable structure multi-model

Publications (2)

Publication Number Publication Date
CN111797478A CN111797478A (en) 2020-10-20
CN111797478B true CN111797478B (en) 2022-11-11

Family

ID=72827390

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010734714.0A Active CN111797478B (en) 2020-07-27 2020-07-27 Strong maneuvering target tracking method based on variable structure multi-model

Country Status (1)

Country Link
CN (1) CN111797478B (en)

Families Citing this family (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111798491B (en) * 2020-07-13 2022-09-06 哈尔滨工业大学 Maneuvering target tracking method based on Elman neural network
CN112257259B (en) * 2020-10-21 2023-08-22 中国人民解放军战略支援部队信息工程大学 Method and system for estimating whole-course trajectory of ballistic missile based on improved autonomous multiple models
CN112577489B (en) * 2020-12-08 2024-05-07 北京电子工程总体研究所 Seeker sight rotation rate extraction method based on interactive multi-model filtering
CN112784506B (en) * 2021-01-29 2023-04-07 中国人民解放军空军工程大学 Reentry maneuvering trajectory target tracking algorithm based on variable structure multi-model
CN114019792B (en) * 2021-10-08 2023-08-01 北京控制工程研究所 Mars atmosphere entry process lift force guidance error analysis method and system
CN115268487B (en) * 2022-07-13 2024-05-28 北京电子工程总体研究所 Aircraft altitude control method and system based on disturbance estimation and compensation LOS guidance law
CN115390560B (en) * 2022-08-18 2023-09-15 哈尔滨工业大学 Ground target track tracking method based on mixed grid multi-model
CN118194609B (en) * 2024-05-17 2024-09-03 北京理工大学前沿技术研究院 Group target tracking method, device and storage medium based on generalization multivariable model

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104020466A (en) * 2014-06-17 2014-09-03 西安电子科技大学 Maneuvering target tracking method based on variable structure multiple models
CN104793201A (en) * 2015-05-04 2015-07-22 哈尔滨工业大学 Modified variable-structure grid interaction multi-model filtering method for tracking hypersonic-speed target of near space
CN107544067A (en) * 2017-07-06 2018-01-05 西北工业大学 One kind is based on the approximate Hypersonic Reentry Vehicles tracking of Gaussian Mixture

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104020466A (en) * 2014-06-17 2014-09-03 西安电子科技大学 Maneuvering target tracking method based on variable structure multiple models
CN104793201A (en) * 2015-05-04 2015-07-22 哈尔滨工业大学 Modified variable-structure grid interaction multi-model filtering method for tracking hypersonic-speed target of near space
CN107544067A (en) * 2017-07-06 2018-01-05 西北工业大学 One kind is based on the approximate Hypersonic Reentry Vehicles tracking of Gaussian Mixture

Non-Patent Citations (5)

* Cited by examiner, † Cited by third party
Title
Review of the algorithms for radar single target tracking;Wei H等;《IOP Conference Series: Earth and Environmental Science》;20171231;全文 *
变结构多模型机动目标跟踪算法研究;娄益茂;《中国优秀硕士学位论文全文数据库信息科技辑》;20160415(第04期);全文 *
基于变结构多模型的机动目标跟踪算法;于红娟等;《火力与指挥控制》;20121231;全文 *
基于变结构多模型算法的高机动目标跟踪研究;刘德虎;《中国优秀硕士学位论文全文数据库信息科技辑》;20131215(第12期);全文 *
强机动目标自适应跟踪算法研究;陈志锋;《中国优秀硕士学位论文全文数据库工程科技Ⅱ辑》;20150615(第06期);全文 *

Also Published As

Publication number Publication date
CN111797478A (en) 2020-10-20

Similar Documents

Publication Publication Date Title
CN111797478B (en) Strong maneuvering target tracking method based on variable structure multi-model
Ryan et al. Particle filter based information-theoretic active sensing
CN111798491B (en) Maneuvering target tracking method based on Elman neural network
CN105785359B (en) A kind of multiple constraint maneuvering target tracking method
Lin et al. A gated recurrent unit-based particle filter for unmanned underwater vehicle state estimation
CN110514567B (en) Gas source searching method based on information entropy
CN108152812B (en) Improved AGIMM tracking method for adjusting grid spacing
CN113095504B (en) Target track prediction system and method
CN103247057A (en) Road target multi-hypothesis tracking algorithm under target-echo-road network data association
CN114462293B (en) Hypersonic speed target medium-long term track prediction method
Hossain et al. A real-time dynamic optimal guidance scheme using a general regression neural network
Xu et al. 3D pseudolinear Kalman filter with own-ship path optimization for AOA target tracking
CN114608585A (en) Method and device for synchronous positioning and mapping of mobile robot
CN117249835A (en) Space-based passive cooperative multi-target observation data association positioning method
CN113761662B (en) Generation method of trajectory prediction pipeline of gliding target
CN114660587A (en) Jump and glide trajectory target tracking method and system based on Jerk model
Brunet et al. Stereo Vision for Unmanned Aerial VehicleDetection, Tracking, and Motion Control
Kumar et al. Target tracking using adaptive Kalman Filter
CN112965530A (en) Multi-unmanned aerial vehicle self-adaptive variable-scale dynamic target searching method
Amor et al. Performance comparison of particle swarm optimization and extended Kalman filter methods for tracking in non-linear dynamic systems
Aguiar et al. Kalman filtering for differential drive robots tracking
Zhu et al. An Accurate Line-of-Sight Rate Estimation Method Based on LSTM Recurrent Neural Network for Strapdown Imaging Seeker
Lee Optimization of moving objects trajectory using particle filter
CN105277928B (en) Without thrust high-speed flight target classification identification system and method in a kind of dense atmosphere
Zeng et al. A trajectory prediction method for boost phase BM based on adaptive tracking and GPR

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