CN114065851B - Maneuvering target track clustering method and system - Google Patents
Maneuvering target track clustering method and system Download PDFInfo
- Publication number
- CN114065851B CN114065851B CN202111332340.0A CN202111332340A CN114065851B CN 114065851 B CN114065851 B CN 114065851B CN 202111332340 A CN202111332340 A CN 202111332340A CN 114065851 B CN114065851 B CN 114065851B
- Authority
- CN
- China
- Prior art keywords
- track
- sequence
- acceleration
- track data
- distance
- 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
Links
- 238000000034 method Methods 0.000 title claims abstract description 63
- 230000001133 acceleration Effects 0.000 claims abstract description 79
- 239000011159 matrix material Substances 0.000 claims abstract description 63
- 230000033001 locomotion Effects 0.000 claims abstract description 61
- 238000012353 t test Methods 0.000 claims abstract description 54
- 238000000692 Student's t-test Methods 0.000 claims abstract description 50
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 15
- 238000012360 testing method Methods 0.000 claims abstract description 5
- 230000002776 aggregation Effects 0.000 claims abstract description 4
- 238000004220 aggregation Methods 0.000 claims abstract description 4
- 238000001228 spectrum Methods 0.000 claims abstract description 4
- 238000010606 normalization Methods 0.000 claims description 63
- 238000012545 processing Methods 0.000 claims description 25
- 230000008569 process Effects 0.000 claims description 21
- 238000005070 sampling Methods 0.000 claims description 19
- 238000005452 bending Methods 0.000 claims description 8
- 238000004364 calculation method Methods 0.000 claims description 6
- 238000009826 distribution Methods 0.000 claims description 6
- 239000013598 vector Substances 0.000 claims description 6
- 230000003595 spectral effect Effects 0.000 claims description 5
- 238000003064 k means clustering Methods 0.000 claims description 4
- 230000001131 transforming effect Effects 0.000 claims description 3
- 230000006870 function Effects 0.000 description 11
- 238000004590 computer program Methods 0.000 description 7
- 238000010586 diagram Methods 0.000 description 7
- 238000005259 measurement Methods 0.000 description 7
- 238000003860 storage Methods 0.000 description 5
- 230000008901 benefit Effects 0.000 description 3
- 238000007621 cluster analysis Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
- G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Cheminformatics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Probability & Statistics with Applications (AREA)
- Radar Systems Or Details Thereof (AREA)
Abstract
The invention discloses a maneuvering target track clustering method and system, comprising the following steps: extracting the instantaneous speed and acceleration of each time point in each track data based on a least square method to obtain a speed sequence and an acceleration sequence corresponding to each track data; respectively performing t-test on the speed sequence and the acceleration sequence, classifying the tracks of the motion model according to the labels according to the t-test judging result, and classifying the unknown model track sequences which are not classified according to the labels; calculating the DTW distance between any two track sequences in the unknown model track sequence set, and obtaining a distance matrix according to the DTW distance; according to the distance matrix, a similarity matrix is obtained, a spectrum aggregation algorithm is adopted for the similarity matrix, a clustering label of an unknown model track sequence is obtained, the known motion type is identified through motion feature hypothesis test, and then the unknown type track clustering is achieved based on the inter-track distance and the clustering algorithm, so that the performance of the track clustering is improved.
Description
Technical Field
The invention relates to the technical field of data classification, in particular to a maneuvering target track clustering method and system.
Background
With the development of modern technology, the movement state of the target is more and more complex and changeable, and the maneuvering performance is obviously enhanced. The traditional maneuvering target tracking algorithm is more and more difficult to track the target, so that the research on maneuvering target tracking technology has important theoretical value and practical significance. In certain aspects, achieving target track clustering is a precondition for achieving maneuver target tracking. Firstly, track cluster analysis is realized, and then, proper models are respectively built aiming at different types of targets, so that the accuracy of target tracking is improved.
Track data essentially belongs to time series data. Conventional track clustering generally only implements clustering according to inter-track distance measurement, and the method does not consider a motion model of a maneuvering target. In addition, the movement speed scale of the maneuvering target cannot be considered in most cases only according to the inter-track distance.
In the problem of track clustering, the measurement of inter-track distance is an important aspect affecting the performance of track clustering. Classical time series data distance definitions are generally divided into two categories, lock-step measures (lock-step measures) and elastic measures (elastic measures). The lockstep metric is a "one-to-one" comparison of the time series; the elasticity measure allows the time series to be compared "one-to-many".
The metric method of point-to-point comparison between two time series is called lockstep metric, most commonly Euclidean Distance (ED). The euclidean distance is simple to calculate, but it requires that both tracks have the same length, which limits the use of euclidean distances. Unlike the lockstep metric, the elasticity metric compares the two time series "one-to-many" or "one-to-zero", and the Hausdorff distance does not take into account the time sequence of track points, and the longest common subsequence (LCSS) distance measures the distance of the two tracks by finding the longest common portion of them, the performance of which depends on the setting of additional parameters. In addition, the distance is related to the movement speed scale of the maneuvering target. Therefore, maneuver target track clustering performance based on the distance is poor.
Disclosure of Invention
The invention aims to solve the technical problem of improving the track clustering performance of maneuvering targets, and provides a maneuvering target track clustering method and system.
The invention is realized by the following technical scheme:
a maneuvering target track clustering method comprises the following steps:
s1, acquiring a track data set to be clustered, wherein each track data set comprises position coordinates of a maneuvering target in three directions at each sampling point X, Y, Z; extracting the instantaneous speed and the acceleration of each sampling point in each track data based on a least square method, and carrying out normalization processing to obtain a speed normalization sequence and an acceleration normalization sequence which correspond to each track data and comprise X, Y, Z components;
s2, respectively performing t-test on three components of the speed standardization sequence and the acceleration standardization sequence of each piece of track data, and marking the piece of track data as a label type to which a corresponding target motion model belongs according to a t-test result;
s3, summarizing unknown track data of the label type in the track data set to be clustered to obtain an unknown model track sequence set;
s4, calculating the DTW distance between any two tracks in the unknown model track sequence set, and obtaining a distance matrix D according to the DTW distance;
s5, transforming the distance matrix D to obtain a similarity matrix S, and adopting a spectral clustering algorithm to the similarity matrix S to obtain clustering labels of each track in the unknown model track sequence set.
For target tracking, target track clustering is a precondition of realizing maneuvering target tracking, a proper model is built for targets of different types respectively through track cluster analysis, so that the accuracy of target tracking is improved, lock step measurement and elasticity measurement are often adopted in the prior art to calculate measurement of inter-track distance, but the two methods do not consider the motion characteristics of the targets, such as time sequence of track points, motion speed scale and acceleration of maneuvering targets, so that the maneuvering target track clustering performance based on the distance measurement of the prior art is poor, the known target motion types are classified and identified through target motion characteristic hypothesis test, the final clustering result is obtained by taking the motion characteristics of the targets such as speed and acceleration into consideration in the clustering process, and the motion characteristics of the motion speed and acceleration are considered in the clustering process to improve the measurement accuracy of the inter-track distance, so that the performance of the maneuvering target clustering is improved, and the tracking accuracy of the maneuvering target is improved.
Further, the process of obtaining the velocity normalization sequence and the acceleration normalization sequence, which correspond to each track data and comprise X, Y, Z components, in the S1 is as follows:
s11, respectively representing the position coordinates of the ith track data in 3 directions as follows: x coordinate direction X i ={x ij J=1, 2, …, R }, Y coordinate direction Y i ={y ij J=1, 2, …, R }, height direction Z i ={z ij ,j=1,2,…,R},x ij ,y ij ,z ij Respectively representing the position coordinate components at the j moment, wherein R represents the number of sampling points;
s12, drawing a curve of track data in the height direction, setting a time window timewindow, and setting the height direction Z in the ith track data i Is a continuous timewindow data z ik ,z i(k+1) ,…,z i(k+timewindow-1) The values of (2) are the ordinate values, 1, 2..timewindow is the abscissa value, where k=1, 2, …, r+1-timewindow;
s13, fitting a curve in the height direction by using least square to obtain a function expression of the curve in the height direction, and obtaining a corresponding speed sequence and acceleration sequence according to the function expression, wherein the formula of the function expression is as follows: z=p 2 x 2 +p 1 x+p 0
Solving the function expression to obtain the corresponding velocity vz ik =2p 2 +p 1 Acceleration of az ik =2p 2 ;
From this, a velocity sequence VZ is obtained i ={vz ik K=1, 2, …, r+1-timewindow, acceleration sequence AZ i ={az ik K=1, 2, …, r+1-timewindow, where p 2 、p 1 and p0 Fitting the obtained function coefficients;
s14, carrying out normalization processing on the speed sequence and the acceleration sequence in the S13 to obtain a speed normalization sequence and an acceleration normalization sequence;
s15, solving the X coordinate direction and the Y coordinate direction of the ith track data according to the methods of the steps S12-S14 to obtain a speed normalization sequence VX respectively i and VYi Acceleration sequence AX i and AYi 。
Further, in S2, the tag value of the tag type to which the target motion model corresponding to the track data belongs is: the uniform linear motion is set to be 1, the uniform acceleration linear motion is set to be 2, the unmanned aerial vehicle/armed helicopter motion model is set to be 3, the shell motion model is set to be 4, and the unknown model is set to be 5;
the marking process of the ith track data according to the t-test judging result is as follows:
s21, respectively performing t-test on three components of the speed standardization sequence, judging whether the average value of t-test results of the three components is 0, if so, setting the label type of the track data to be 1, and if not, executing the step S22;
s22, respectively performing t-test on three components of the acceleration standardization sequence, judging whether the average value of t-test results of the three components is 0, if so, setting the label type of the track data to be 2, and if not, executing step S23;
s23, carrying out normalization processing on the position coordinates of each sampling point in the Z direction in the track data to obtain a Z direction coordinate set, carrying out t-test on a plurality of points in the Z direction coordinate set, judging whether the average value of t-test results is 0, if so, setting the label type of the track data to be 3, and if not, executing step S24;
s24, constructing a feature number sequence according to the speed standardization sequence and the acceleration standardization sequence, performing t-test on the feature number sequence, judging whether the average value of the t-test result is smaller than a threshold value, if yes, setting the label type of the track data to be 4, and if not, setting the label type of the track data to be 5.
Further, the t-test process is as follows:
wherein ,representing a sample mean, S being a sample variance, n=r+1-timewindow, R representing the number of sampling points, timewindow representing a time window, μ being a hypothetical mean set to 0, α representing a confidence level;
if t is greater than or equal to t α (n-1); rejecting the assumption that the distribution mean of the samples is μ;
if t is less than t α (n-1), the original distribution mean of the sample is considered μ with a probability greater than 1- α.
Further, in S24, in order to identify the track data of the projectile motion type, the target feature number of the projectile motion type is deduced, and the process for distinguishing the projectile motion type from other types of motions and constructing the feature number sequence is as follows:
selecting a velocity normalization sequence of an X-direction component from three components of the velocity normalization sequence, and selecting an acceleration normalization sequence of a Z-direction from three components of the acceleration normalization sequence;
for the acceleration in the Z direction and the acceleration in the X direction after adding 9.8 at the kth moment, and then the difference between the acceleration in the Z direction and the velocity in the X direction is obtained, the characteristic sequence at the kth moment is obtained, the processing of the processes is carried out on all the moments in the acceleration normalization sequence in the Z direction and the velocity normalization sequence of the X direction component, the characteristic sequence Ti is obtained,
when t-test is performed on the feature number sequence, the threshold is set to 1.
Further, for an unknown model track sequence set with tag type 5,
s41, selecting any two tracks P= { P 1 ,p 2 ,...,p m And track q= { Q 1 ,q 2 ,..., n q, wherein m and n respectively represent the number of sampling points in the track P, Q; computing the u-th point P in the track P by adopting Euclidean distance u And the c-th point Q in the track Q c The distance between them is denoted as the base distance d base (p u -q c );
S42, calculating the DTW distance between tracks P, Q by adopting a dynamic programming method, wherein the calculation method is as follows:
DTW(p,q)=
d base (head(p),head(q))+min{DTW(rest(p),q),DTW(p,rest(q)),DTW(rest(p),rest(q))};
where rest (p) represents the sample point remaining after the first sample point is removed from the sequence p, and head (p) represents the first sample point of the sequence p.
For an unknown model track sequence set, calculating dynamic time bending distances Dij between a plurality of pairs of tracks according to the process of S42;
obtaining a distance matrix D according to the dynamic time bending distance Dij, and calculating a similarity matrix S according to the distance matrix, wherein the ith row and jth column elements S of the similarity matrix S ij Calculated according to the following formula:
S ij =exp(-D ij /std(D));
where std (D) represents the standard deviation of all elements in matrix D.
Further, in S5, a spectral clustering algorithm is adopted for the similarity matrix S to obtain a cluster label of each track in the unknown model track sequence set, which specifically includes the steps of:
s51, decomposing the similarity matrix S features, s=uΛu -1 Wherein U is a matrix composed of eigenvectors, U -1 Representing the inverse matrix of U, Λ being the diagonal comprised of eigenvaluesA matrix;
s52, w eigenvectors corresponding to the minimum w eigenvalues are taken from the matrix U to form an M-row w-column matrix Um, wherein w is the cluster number;
s53, taking Um as M w-dimensional vectors, and adopting a k-means clustering algorithm for the M w-dimensional vectors;
s54, obtaining the cluster labels of M tracks in the unknown model track sequence set.
In addition, the invention provides a maneuvering target track clustering system, which comprises a track data processing module, a t-test module, a track data classifying module according to labels, a track distance calculating module and an unknown track classifying module; wherein,
the track data processing module is used for extracting the instantaneous speed and the acceleration of each time point in each track data based on a least square method for each track data in a track data set to be clustered, and carrying out normalization processing to obtain a speed normalization sequence and an acceleration normalization sequence which correspond to each track data and comprise X, Y, Z components;
the t-test module is used for normalizing the sequence V according to the speed of each track data i T-test is respectively carried out on the X, Y, Z three components and the acceleration normalization sequence;
performing label assignment on each piece of track data according to a label classification module and a t-test judgment result to obtain a preset target motion model label type to which the piece of track data belongs; wherein uniform linear motion is set to be 1, uniform acceleration linear motion is set to be 2, unmanned aerial vehicle/armed helicopter motion model is set to be 3, shell motion model is set to be 4, and unknown model is set to be 5; obtaining an unknown model track sequence set according to the track data with the label type of 5;
the track distance calculation module is used for calculating the DTW distance between any two track sequences in the unknown model track sequence set based on the dynamic time bending distance to obtain a distance matrix D
The unknown track classification module is used for obtaining a similarity matrix S according to the distance matrix D, and obtaining clustering labels of each track in the unknown model track sequence set by adopting a spectrum aggregation algorithm for the similarity matrix S.
Compared with the prior art, the invention has the following advantages and beneficial effects:
the invention relates to a maneuvering target track clustering method and a maneuvering target track clustering system, which fully consider the movement characteristics of a target track into a clustering process, better realize the identification of known target types based on hypothesis test, deduce the characteristic numbers of the targets based on a shell movement model, be used for distinguishing the movement of other types, realize the clustering of unknown types of tracks by using a k-means clustering algorithm which is not identified and is difficult to describe and is evaluated based on track distance and similarity, consider the movement speed scale of maneuvering targets among tracks, better reflect the movement models of the targets, obtain the clustering results of different tracks, or recluster similar tracks.
Drawings
In order to more clearly illustrate the technical solutions of the exemplary embodiments of the present invention, the drawings that are needed in the examples will be briefly described below, it being understood that the following drawings only illustrate some examples of the present invention and therefore should not be considered as limiting the scope, and that other related drawings may be obtained from these drawings without inventive effort for a person skilled in the art. In the drawings:
FIG. 1 is a flow chart of a method for clustering target tracks according to an embodiment of the present invention;
FIG. 2 is a process of clustering target tracks for a known model in one embodiment of the invention;
FIG. 3 is a track library of track clusters in one embodiment of the invention;
FIG. 4 is a graph illustrating the uniform acceleration of tracks in a track cluster in accordance with one embodiment of the present invention;
FIG. 5 is a diagram of a projectile track in a track cluster in accordance with one embodiment of the invention;
FIG. 6 is a view of a guided munition track in a track cluster in accordance with one embodiment of the invention.
Detailed Description
For the purpose of making apparent the objects, technical solutions and advantages of the present invention, the present invention will be further described in detail with reference to the following examples and the accompanying drawings, wherein the exemplary embodiments of the present invention and the descriptions thereof are for illustrating the present invention only and are not to be construed as limiting the present invention.
Example 1
As shown in fig. 1, the embodiment provides a maneuvering target track clustering method, which includes the following steps:
s1, acquiring a track data set to be clustered, wherein each track data set comprises position coordinates of a maneuvering target in three directions at each sampling point X, Y, Z; extracting the instantaneous speed and the acceleration of each sampling point in each track data based on a least square method, and carrying out normalization processing to obtain a speed normalization sequence and an acceleration normalization sequence which correspond to each track data and comprise X, Y, Z components;
specifically, the corresponding velocity normalization sequence and acceleration normalization sequence containing X, Y, Z components can be obtained for the ith track:
s11, respectively representing the position coordinates of the ith track data in 3 directions as follows: x coordinate direction X i ={x ij J=1, 2, …, R }, Y coordinate direction Y i ={y ij J=1, 2, …, R }, height direction Z i ={z ij ,j=1,2,…,R},x ij ,y ij ,z ij Respectively representing the position coordinate components at the j moment, wherein R represents the number of sampling points;
s12, drawing a curve of track data in the height direction, setting a time window timewindow, and setting the height direction Z in the ith track data i Is a continuous timewindow data z ik ,z i(k+1) ,…,z i(k+timewindow-1) The values of (2) are the ordinate values, 1, 2..timewindow is the abscissa value, where k=1, 2, …, r+1-timewindow;
s13, curve in height directionFitting by least square to obtain a functional expression of the height direction curve, and obtaining a corresponding speed sequence and acceleration sequence according to the functional expression, wherein the formula of the functional expression is as follows: z=p 2 x 2 +p 1 x+p 0
Solving the function expression to obtain the corresponding velocity vz ik =2p 2 +p 1 Acceleration of az ik =2p 2 ;
From this, a velocity sequence VZ is obtained i ={vz ik K=1, 2, …, r+1-timewindow, acceleration sequence AZ i ={az ik K=1, 2, …, r+1-timewindow, where p 2 、p 1 and p0 Fitting the obtained function coefficients;
s14, carrying out normalization processing on the speed sequence and the acceleration sequence in the S13 to obtain a speed normalization sequence and an acceleration normalization sequence;
s15, solving the X coordinate direction and the Y coordinate direction of the ith track data according to the methods of the steps S12-S14 to obtain a speed normalization sequence VX respectively i and VYi Acceleration sequence AX i and AYi 。
S2, respectively performing t-test on three components of the speed standardization sequence and the acceleration standardization sequence of each piece of track data, and marking the piece of track data as a label type to which a corresponding target motion model belongs according to a t-test result;
as shown in fig. 2, the tag values of the tag types to which the target motion model corresponding to the track data belongs are: the uniform linear motion is set to be 1, the uniform acceleration linear motion is set to be 2, the unmanned aerial vehicle/armed helicopter motion model is set to be 3, the shell motion model is set to be 4, and the unknown model is set to be 5;
the specific process of track clustering for the known motion model is: for the ith track data, the marking process according to the t-test judging result is as follows:
s21, respectively performing t-test on three components of the speed standardization sequence, judging whether the average value of t-test results of the three components is 0, if so, setting the label type of the track data to be 1, and if not, executing the step S22;
s22, respectively performing t-test on three components of the acceleration standardization sequence, judging whether the average value of t-test results of the three components is 0, if so, setting the label type of the track data to be 2, and if not, executing step S23;
s23, carrying out normalization processing on the position coordinates of each sampling point in the Z direction in the track data to obtain a Z direction coordinate set, carrying out t-test on a plurality of points in the Z direction coordinate set, judging whether the average value of t-test results is 0, if so, setting the label type of the track data to be 3, and if not, executing step S24;
s24, constructing a feature number sequence according to the speed standardization sequence and the acceleration standardization sequence, performing t-test on the feature number sequence, judging whether the average value of the t-test result is smaller than a threshold value, if yes, setting the label type of the track data to be 4, and if not, setting the label type of the track data to be 5.
In order to identify the flight path data of the projectile motion type, deducing the target feature number of the projectile motion type, which is used for distinguishing the projectile motion type from other types of motions, and constructing a feature number sequence comprises the following steps:
selecting a velocity normalization sequence of an x-direction component from three components of the velocity normalization sequence, and selecting an acceleration normalization sequence of a z-direction from three components of the acceleration normalization sequence;
selecting a velocity normalization sequence of an X-direction component from three components of the velocity normalization sequence, and selecting an acceleration normalization sequence of a Z-direction from three components of the acceleration normalization sequence;
for the acceleration in the Z direction and the acceleration in the X direction after adding 9.8 at the kth moment, and then the difference between the acceleration in the Z direction and the velocity in the X direction is obtained, the characteristic sequence at the kth moment is obtained, the processing of the processes is carried out on all the moments in the acceleration normalization sequence in the Z direction and the velocity normalization sequence of the X direction component, the characteristic sequence Ti is obtained,
when t-test is performed on the feature number sequence, the threshold is set to 1.
Specifically, the t-test method comprises the following steps:
wherein ,representing a sample mean, S being a sample variance, n=r+1-timewindow, R representing the number of sampling points, timewindow representing a time window, μ being a hypothetical mean set to 0, α representing a confidence level;
if t is greater than or equal to t α (n-1); rejecting the assumption that the distribution mean of the samples is μ;
if t is less than t α (n-1), the original distribution mean of the sample is considered μ with a probability greater than 1- α.
For example, in step S21, for normalized velocity sequence V i ={v ik ,k=1,2,...,R+1-timewindow}
n=R-timewindow+ 1
the tag type is set to 1, otherwise the next step S22 is performed.
S3, summarizing the track data with the label type of 5 in the track data set to be clustered to obtain an unknown model track sequence set;
s4, calculating the DTW distance between any two track sequences in the unknown model track sequence set, and obtaining a distance matrix D according to the DTW distance;
specifically, for an unknown model track sequence set with a tag type of 5,
s41, selecting any two tracks P= { P 1 ,p 2 ,...,p m And track q= { Q 1 ,q 2 ,..., n q, wherein m and n respectively represent the number of sampling points in the track P, Q; computing the u-th point P in the track P by adopting Euclidean distance u And the c-th point Q in the track Q c The distance between them is denoted as the base distance d base (p u -q c );
S42, calculating the DTW distance between tracks P, Q by adopting a dynamic programming method, wherein the calculation method is as follows:
DTW(p,q)=
d base (head(p),head(q))+min{DTW(rest(p),q),DTW(p,rest(q)),DTW(rest(p),rest(q))};
where rest (p) represents the sample point remaining after the first sample point is removed from the sequence p, and head (p) represents the first sample point of the sequence p.
The DTW distance for the calculated sequences p and q is defined as follows:
DTW(<>,<>)=0
DTW(p,<>)=DTW(<>,q)=∞
DTW(p,q)=
d base (head(p),head(q))+min{DTW(rest(p),q),DTW(p,rest(q)),DTW(rest(p),rest(q))}
wherein rest (·) represents the point left after the first sampling point is removed from the sequence, head (·) represents the first point of the sequence x, and the DTW distance between any two track sequences P and Q is obtained by adopting a dynamic programming method.
For an unknown model track sequence set, calculating dynamic time between a plurality of pairs of tracks according to the process of S42The inter-bending distance Dij; obtaining a distance matrix D according to the dynamic time bending distance Dij, and calculating a similarity matrix S according to the distance matrix, wherein the ith row and jth column elements S of the similarity matrix S ij Calculated according to the following formula:
S ij =exp(-D ij /std(D));
where std (D) represents the standard deviation of all elements in matrix D.
S5, transforming the distance matrix D to obtain a similarity matrix S, and adopting a spectral clustering algorithm to the similarity matrix S to obtain clustering labels of each track in the unknown model track sequence set, wherein the method specifically comprises the following steps of:
s51, decomposing the similarity matrix S features, s=uΛu -1 Wherein U is a matrix composed of eigenvectors, U -1 Representing the inverse matrix of U, Λ being a diagonal matrix consisting of eigenvalues;
s52, w eigenvectors corresponding to the minimum w eigenvalues are taken from the matrix U to form an M-row w-column matrix Um, wherein w is the cluster number;
s53, taking Um as M w-dimensional vectors, and adopting a k-means clustering algorithm for the M w-dimensional vectors;
s54, obtaining cluster labels of M tracks in the unknown model track sequence set, for example, 5-1,5-2.
Example 2
The embodiment provides a maneuvering target track clustering system, which comprises a track data processing module, a t-test module, a track data classifying module according to labels, a track distance calculating module and an unknown track classifying module; wherein,
the track data processing module is used for extracting the instantaneous speed and the acceleration of each time point in each track data based on a least square method for each track data in a track data set to be clustered, and carrying out normalization processing to obtain a speed normalization sequence and an acceleration normalization sequence which correspond to each track data and comprise X, Y, Z components;
the t-test module is used for normalizing the sequence V according to the speed of each track data i X, Y, Z components and acceleration normalizationRespectively performing t-test on the sequences;
performing label assignment on each piece of track data according to a label classification module and a t-test judgment result to obtain a preset target motion model label type to which the piece of track data belongs; wherein uniform linear motion is set to be 1, uniform acceleration linear motion is set to be 2, unmanned aerial vehicle/armed helicopter motion model is set to be 3, shell motion model is set to be 4, and unknown model is set to be 5; obtaining an unknown model track sequence set according to the track data with the label type of 5;
the track distance calculation module is used for calculating the DTW distance between any two track sequences in the unknown model track sequence set based on the dynamic time bending distance to obtain a distance matrix D;
the unknown track classification module is used for obtaining a similarity matrix S according to the distance matrix D, and obtaining clustering labels of each track in the unknown model track sequence set by adopting a spectrum aggregation algorithm for the similarity matrix S.
To better illustrate the benefits of the present invention, each curve corresponds to a track, as shown in FIG. 3, with track data comprising 3 types of objects, namely, a uniform acceleration track, a projectile track, and a guided munition track, respectively. The experimental objective is to group track data into three categories by the method of the invention. The final clustering results are shown in fig. 4, 5 and 6, and it can be seen that the experimental data are classified into three categories, which shows that the method of the invention performs well.
It will be appreciated by those skilled in the art that embodiments of the present application may be provided as a method, system, or computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
The present application is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the application. It will be understood that each flow and/or block of the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
Those of ordinary skill in the art will appreciate that implementing all or part of the above facts and methods may be accomplished by a program to instruct related hardware, the program involved or the program may be stored in a computer readable storage medium, the program when executed comprising the steps of: the corresponding method steps are introduced at this time, and the storage medium may be a ROM/RAM, a magnetic disk, an optical disk, or the like.
The foregoing description of the embodiments has been provided for the purpose of illustrating the general principles of the invention, and is not meant to limit the scope of the invention, but to limit the invention to the particular embodiments, and any modifications, equivalents, improvements, etc. that fall within the spirit and principles of the invention are intended to be included within the scope of the invention.
Claims (9)
1. The maneuvering target track clustering method is characterized by comprising the following steps of:
s1, acquiring a track data set to be clustered, wherein each track data set comprises position coordinates of a maneuvering target in three directions at each sampling point X, Y, Z; extracting the instantaneous speed and the acceleration of each sampling point in each track data based on a least square method, and carrying out normalization processing to obtain a speed normalization sequence and an acceleration normalization sequence which correspond to each track data and comprise X, Y, Z components;
s2, respectively performing t-test on three components of the speed standardization sequence and the acceleration standardization sequence of each piece of track data, and marking the piece of track data as a label type to which a corresponding target motion model belongs according to a t-test result;
s3, summarizing unknown track data of the label type in the track data set to be clustered to obtain an unknown model track sequence set;
s4, calculating the DTW distance between any two tracks in the unknown model track sequence set, and obtaining a distance matrix D according to the DTW distance;
s5, transforming the distance matrix D to obtain a similarity matrix S, and adopting a spectral clustering algorithm to the similarity matrix S to obtain clustering labels of each track in the unknown model track sequence set.
2. The maneuvering target track clustering method according to claim 1, wherein the process of obtaining a speed normalization sequence and an acceleration normalization sequence corresponding to each track data and containing X, Y, Z components in S1 is as follows:
s11, respectively representing the position coordinates of the ith track data in three directions as follows: x coordinate direction X i ={x ij J=1, 2, …, R }, Y coordinate direction Y i ={y ij J=1, 2, …, R }, height direction Z i ={z ij ,j=1,2,…,R},x ij ,y ij ,z ij Respectively representing the position coordinate components at the j moment, wherein R represents the number of sampling points;
s12, drawing a curve of track data in the height direction, setting a time window timewindow, and setting the height direction Z in the ith track data i Is a continuous timewindow data z ik ,z i(k+1) ,…,z i(k+timewindow-1) The values of (2) are the ordinate values, 1, 2..timewindow is the abscissa value, where k=1, 2, …, r+1-timewindow;
s13, fitting a curve in the height direction by using least square to obtain a function expression of the curve in the height direction, and obtaining a corresponding speed sequence and acceleration sequence according to the function expression, wherein the formula of the function expression is as follows: z=p 2 x 2 +p 1 x+p 0
Solving the function expression to obtain the corresponding velocity vz ik =2p 2 +p 1 Acceleration of az ik =2p 2 ;
From this, a velocity sequence VZ is obtained i ={vz ik K=1, 2, …, r+1-timewindow, acceleration sequence AZ i ={az ik K=1, 2, …, r+1-timewindow, where p 2 、p 1 and p0 Fitting the obtained function coefficients;
s14, carrying out normalization processing on the speed sequence and the acceleration sequence in the S13 to obtain a speed normalization sequence and an acceleration normalization sequence;
s15, solving the X coordinate direction and the Y coordinate direction of the ith track data according to the methods of the steps S12-S14 to obtain a speed normalization sequence VX respectively i and VYi Acceleration sequence AX i and AYi 。
3. The maneuvering target track clustering method according to claim 1, wherein in S2, a tag value of a tag type to which a target motion model corresponding to track data belongs is: the uniform linear motion is set to be 1, the uniform acceleration linear motion is set to be 2, the unmanned aerial vehicle/armed helicopter motion model is set to be 3, the shell motion model is set to be 4, and the unknown model is set to be 5;
the marking process according to the t test result is as follows for the ith track data:
s21, respectively performing t-test on three components of the speed standardization sequence, judging whether the average value of t-test results of the three components is 0, if so, setting the label type of the track data to be 1, and if not, executing the step S22;
s22, respectively performing t-test on three components of the acceleration standardization sequence, judging whether the average value of t-test results of the three components is 0, if so, setting the label type of the track data to be 2, and if not, executing step S23;
s23, carrying out normalization processing on the position coordinates of each sampling point in the Z direction in the track data to obtain a Z direction coordinate set, carrying out t-test on a plurality of points in the Z direction coordinate set, judging whether the average value of t-test results is 0, if so, setting the label type of the track data to be 3, and if not, executing step S24;
s24, constructing a feature number sequence according to the speed standardization sequence and the acceleration standardization sequence, performing t-test on the feature number sequence, judging whether the average value of the t-test result is smaller than a threshold value, if yes, setting the label type of the track data to be 4, and if not, setting the label type of the track data to be 5.
4. A maneuvering target track clustering method according to claim 3, wherein the t-test process is as follows:
wherein ,representation sampleThe mean value is S is a sample variance, n=R+1-timewindow, R represents the number of sampling points, timewindow represents a time window, mu is assumed that the mean value is set to 0, and alpha represents confidence;
if t is greater than or equal to t α (n-1); rejecting the assumption that the distribution mean of the samples is μ;
if t is less than t α (n-1), the original distribution mean of the sample is considered μ with a probability greater than 1- α.
5. A method of clustering maneuvering targets according to claim 3, wherein in S24, the process of constructing the feature number sequence is:
selecting a velocity normalization sequence of an X-direction component from three components of the velocity normalization sequence, and selecting an acceleration normalization sequence of a Z-direction from three components of the acceleration normalization sequence;
for the acceleration in the Z direction and the acceleration in the X direction after adding 9.8 at the kth moment, and then the difference between the acceleration in the Z direction and the velocity in the X direction is obtained, the characteristic sequence at the kth moment is obtained, the processing of the processes is carried out on all the moments in the acceleration normalization sequence in the Z direction and the velocity normalization sequence of the X direction component, the characteristic sequence Ti is obtained,
when t-test is performed on the feature number sequence, the threshold is set to 1.
6. A maneuver target track clustering method as defined in claim 1 wherein, for an unknown model track sequence set with a tag type of 5,
s41, selecting any two tracks P= { P 1 ,p 2 ,..., m p and track q= { Q 1 ,q 2 ,...,q n Wherein m and n respectively represent the number of sampling points in the track P, Q;computing the u-th point P in the track P by adopting Euclidean distance u And the c-th point Q in the track Q c The distance between them is denoted as the base distance d base (p u -q c );
S42, calculating the DTW distance between tracks P, Q by adopting a dynamic programming method, wherein the calculation method is as follows:
DTW(p,q)=
d base (head(p),head(q))+min{DTW(rest(p),q),DTW(p,rest(q)),DTW(rest(p),rest(q))};
where rest (p) represents the sample point remaining after the first sample point is removed from the sequence p, and head (p) represents the first sample point of the sequence p.
7. The method for clustering maneuvering targets according to claim 6, wherein for a set of unknown model track sequences, calculating dynamic time warping distances Dij between pairs of tracks according to the process of S42;
obtaining a distance matrix D according to the dynamic time bending distance Dij, and calculating a similarity matrix S according to the distance matrix, wherein the ith row and jth column elements S of the similarity matrix S ij Calculated according to the following formula:
S ij =exp(-D ij /std(D));
where std (D) represents the standard deviation of all elements in matrix D.
8. The method for clustering maneuvering target tracks according to claim 7, wherein in S5, a spectral clustering algorithm is adopted for the similarity matrix S to obtain a cluster label of each track in the unknown model track sequence set, and the specific steps include:
s51, decomposing the similarity matrix S features, s=uΛu -1 Wherein U is a matrix composed of eigenvectors, U -1 Representing the inverse matrix of U, Λ being a diagonal matrix consisting of eigenvalues;
s52, w eigenvectors corresponding to the minimum w eigenvalues are taken from the matrix U to form an M-row w-column matrix Um, wherein w is the cluster number;
s53, taking Um as M w-dimensional vectors, and adopting a k-means clustering algorithm for the M w-dimensional vectors;
s54, obtaining the cluster labels of M tracks in the unknown model track sequence set.
9. The maneuvering target track clustering system is characterized by comprising a track data processing module, a t-test module, a track data classifying module according to labels, a track distance calculating module and an unknown track classifying module; wherein,
the track data processing module is used for extracting the instantaneous speed and the acceleration of each time point in each track data based on a least square method for each track data in a track data set to be clustered, and carrying out normalization processing to obtain a speed normalization sequence and an acceleration normalization sequence which correspond to each track data and comprise X, Y, Z components;
the t-test module is used for normalizing the sequence V according to the speed of each track data i T-test is respectively carried out on the X, Y, Z three components and the acceleration normalization sequence;
performing label assignment on each piece of track data according to a label classification module and a t-test judgment result to obtain a preset target motion model label type to which the piece of track data belongs; wherein uniform linear motion is set to be 1, uniform acceleration linear motion is set to be 2, unmanned aerial vehicle/armed helicopter motion model is set to be 3, shell motion model is set to be 4, and unknown model is set to be 5; obtaining an unknown model track sequence set according to the track data with the label type of 5;
the track distance calculation module is used for calculating the DTW distance between any two track sequences in the unknown model track sequence set based on the dynamic time bending distance to obtain a distance matrix D;
the unknown track classification module is used for obtaining a similarity matrix S according to the distance matrix D, and obtaining clustering labels of each track in the unknown model track sequence set by adopting a spectrum aggregation algorithm for the similarity matrix S.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111332340.0A CN114065851B (en) | 2021-11-11 | 2021-11-11 | Maneuvering target track clustering method and system |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111332340.0A CN114065851B (en) | 2021-11-11 | 2021-11-11 | Maneuvering target track clustering method and system |
Publications (2)
Publication Number | Publication Date |
---|---|
CN114065851A CN114065851A (en) | 2022-02-18 |
CN114065851B true CN114065851B (en) | 2023-05-02 |
Family
ID=80275201
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202111332340.0A Active CN114065851B (en) | 2021-11-11 | 2021-11-11 | Maneuvering target track clustering method and system |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN114065851B (en) |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2003130948A (en) * | 2001-10-25 | 2003-05-08 | Mitsubishi Electric Corp | Target tracking apparatus |
CN110033017A (en) * | 2019-02-27 | 2019-07-19 | 中国人民解放军空军工程大学 | A kind of more radar track substep Interconnected Fuzzy clustering algorithms |
CN110135451A (en) * | 2019-03-27 | 2019-08-16 | 中电莱斯信息系统有限公司 | A kind of track clustering method arriving line-segment sets distance based on point |
CN110399686A (en) * | 2019-07-30 | 2019-11-01 | 中国民航大学 | A kind of unrelated aircraft flight profiles clustering method of parameter based on silhouette coefficient |
-
2021
- 2021-11-11 CN CN202111332340.0A patent/CN114065851B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2003130948A (en) * | 2001-10-25 | 2003-05-08 | Mitsubishi Electric Corp | Target tracking apparatus |
CN110033017A (en) * | 2019-02-27 | 2019-07-19 | 中国人民解放军空军工程大学 | A kind of more radar track substep Interconnected Fuzzy clustering algorithms |
CN110135451A (en) * | 2019-03-27 | 2019-08-16 | 中电莱斯信息系统有限公司 | A kind of track clustering method arriving line-segment sets distance based on point |
CN110399686A (en) * | 2019-07-30 | 2019-11-01 | 中国民航大学 | A kind of unrelated aircraft flight profiles clustering method of parameter based on silhouette coefficient |
Non-Patent Citations (1)
Title |
---|
基于近邻传播算法的航迹聚类分析;张榆薪等;《软件导刊》;20180415(第04期);全文 * |
Also Published As
Publication number | Publication date |
---|---|
CN114065851A (en) | 2022-02-18 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN112308881B (en) | Ship multi-target tracking method based on remote sensing image | |
US5636291A (en) | Continuous parameter hidden Markov model approach to automatic handwriting recognition | |
CN110197502B (en) | Multi-target tracking method and system based on identity re-identification | |
CN106778501B (en) | Video face online identification method based on compression tracking and IHDR incremental learning | |
EP0539749A2 (en) | A statistical mixture approach to automatic handwriting recognition | |
CN103415825A (en) | System and method for gesture recognition | |
CN105261367B (en) | A kind of method for distinguishing speek person | |
CN112149705A (en) | Method and system for training classification model, computer equipment and storage medium | |
CN107609590B (en) | Multi-scale mouse track feature extraction method, device and system | |
Horak et al. | Classification of SURF image features by selected machine learning algorithms | |
CN114065851B (en) | Maneuvering target track clustering method and system | |
Zhang et al. | Detecting abnormal events via hierarchical Dirichlet processes | |
CN111738194B (en) | Method and device for evaluating similarity of face images | |
Orrite et al. | One-sequence learning of human actions | |
JP4928193B2 (en) | Face image recognition apparatus and face image recognition program | |
Li et al. | Fault Diagnosis Based on Improved Kernel Fisher Discriminant Analysis. | |
Yin et al. | Abnormal behavior recognition using self-adaptive hidden markov models | |
Su et al. | Learning low-dimensional temporal representations with latent alignments | |
Dong et al. | On-line signature verification based on template matching approach and support vector data description | |
CN114090967A (en) | APT (android package) organization tracing and tracing method and system based on PSO-MSVM (Power System-Mobile virtual machine) | |
Shon et al. | Large-scale speaker retrieval on random speaker variability subspace | |
CN112016417A (en) | Pedestrian re-identification method and system based on free energy fractional space | |
Gao et al. | Feature tracking for target identification in acoustic image sequences | |
Yang et al. | Flexible trajectory indexing for 3d motion recognition | |
Shah et al. | Image pattern classification using MFCC and HMM |
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 |