CN104992186A - Aurora video classification method based on dynamic texture model representation - Google Patents

Abstract

The invention discloses an aurora video classification method based on a dynamic texture model. The technical scheme is characterized by carrying out universality modeling and characteristic extraction on aurora videos of four kinds of forms and carrying out classification on the aurora videos by utilizing repeatability correlation between the aurora video frames and starting from the dynamic change characteristics between the aurora video frames. The technical scheme is realized through the following steps: 1) obtaining a training aurora video and a test aurora video; 2) extracting dynamic texture characteristics of the test aurora video; 3) extracting a state transition matrix of the test aurora video; 4) calculating the Martin distance between the test aurora video and a training set sample; and 5) carrying out nearest distance classification on the test aurora video according to the Martin distance between the aurora videos. The method can realize automatic computer classification of the four kinds of aurora videos, has the advantage of high classification accuracy, and can be applied to aurora video feature extraction and computer image identification.

Description

Based on the aurora video classification methods that dynamic texture model characterizes

Technical field

The invention belongs to technical field of image processing, relate to the computer classification method of four kinds of form aurora videos, can be used for feature extraction and the computer picture recognition of aurora video.

Background technology

Aurora are solar winds when being injected into earth magnetosphere by a day Cusp/Cleft, side, the gorgeous radiance that fallout particulate interacts along the magnetic line of force and the upper atmosphere and produces.Aurora are observation windows of polar region space weather physical process, reflect solar wind and the coupling process of ground magnetosphere intuitively, contain the electromagnetic activity information of a large amount of solar-terrestrial physics, the Research Significance of own profound.

The all-sky imaging system (All-sky Camera) of Chinese Arctic Yellow River Station carries out Continuous Observation to three typical spectral coverage 427.8nm, 557.7nm and 630.0nm of aurora simultaneously, and produce ten hundreds of aurora images, data volume is huge.Aurora are divided into arcuation, radial, valance shape and focus shape four class by form by the people such as Wang Q in article " Spatial texture based automatic classification of dayside aurora in all-sky images.Journal of Atmospheric and Solar-Terrestrial Physics; 2010; 72 (5): 498-508. ", and have drawn the Statistical Distribution of four kinds of aurora types.Article " the Pedersen T R that the people such as Pedersen deliver, Gerken E A.Creation of visibleartificial optical emission in the aurora by high-power radio waves.Nature, 2005, 433 (7025): 498-500 ", the people such as Hu Zejun publish an article " Hu Z J, Yang H, Huang D, et al.Synoptic distribution ofdayside aurora:Multiple-wavelength all-sky observation at Yellow River Station in Ny-Alesund, Svalbard.J.Atmos.Sol.-Terr.Phys., 2009, 71 (89): 794-804 " and article " LorentzenD A, Moen J, Oksavik K, et al.In situmeasurement of a newly created polar cap patch.J.Geophys.Res., 2010, 115 (A12). ", provide a large amount of research materials, prove that the aurora of different shape correspond to different magnetosphere Boundary Layer Dynamics processes.How accurately and efficiently aurora video to be classified, it is the key disclosing its magnetosphere source region dynamic process, also be the important step of its mechanism research, but aurora form and dynamic change complexity, undoubtedly for polar region researcher brings huge difficulty.

The magnanimity aurora Data classification research that develops into of computer picture recognition and analytical technology provides possibility.2004, deng people at article " Syrjasuo M, Partamies N.Numeric image features for detection ofaurora [J] .Geoscience and Remote Sensing Letters, IEEE, 2012, 9 (2): 176-179. " method of computer vision is introduced in aurora classification of images system in, the method extracts Fourier operator as feature from the auroral region after segmentation, the automatic classification of aurora image is realized by arest neighbors method, owing to being subject to the impact of partitioning algorithm, the method is only respond well to shape facility obvious arcuation aurora Images Classification, the people such as Wang in 2007 at article " Wang Qian, Liang Jimin, Hu ZeJun, Hu HaiHong, Zhao Heng, Hu HongQiao, Gao Xinbo, Yang Huigen.Spatial texture based automatic classification of dayside aurora in all-sky images.Journal of Atmospheric and Solar-Terrestrial Physics, 2010, 72 (5): 498 – 508. " in use the gray feature of principal component analysis (PCA) PCA to aurora image to extract, propose a kind of aurora sorting technique based on presentation, certain progress is achieved in Coronal aurorae sort research direction, 2008, the people such as Gao publish an article " L.Gao, X.B.Gao, and J.M.Liang.Dayside corona autora detection based on sample selection and adaBoostalgorithm.J.I mage Graph, 2010, 15 (1): 116-121. ", aurora image classification method based on Gabor transformation is proposed, have employed local Gabor filter and extract characteristics of image, reducing feature redundant information when guaranteeing computational accuracy, achieving good classifying quality, 2009, morphology constituent analysis (MCA) combines with aurora image procossing by the people such as Fu in article " Fu Ru, Jie Li and X.B.Gao..Automatic aurora images classification algorithm based on separated texture.Proc.Int.Conf.Robotics and Biomimetics, 2009:1331-1335. ", feature is extracted from the aurora texture subgraph obtained after MCA is separated, for the classification of arc crown two class aurora image, improve the accuracy of arc crown aurora classification.Follow-up correlative study also has: the people such as Han propose again the aurora classification based on the classification of BIFs characteristic sum C average in article " Bing Han; Xiaojing Zhao; Dacheng Tao, et al.Dayside aurora classification via BIFs-basedsparse representation using manifold learning.International Journal of Computer Mathematics.Published online:12Nov 2013. "; The people such as Yang propose multi-level Wavelet Transform conversion and represent aurora characteristics of image in article " Yang Xi; Li Jie; Han Bing; Gao Xinbo.Wavelet hierarchical model foraurora images classification.Journal of Xidian University; 2013; 40 (2): 18-24. ", achieve higher classification accuracy; 2013, the people such as Han introduce implicit Dirichlet distribute model LDA in article " Han B; Yang C; Gao XB.Aurora image classification based on LDA combining withsaliency information.RuanJian Xue Bao/Journal of Software; 2013; 24 (11): 2758-2766. ", and combining image conspicuousness information, further increase again the classification accuracy of aurora image.

But mostly these existing aurora graphical analyses above-mentioned are based on single image and static nature, and the related work about the automatic classification of aurora sequence is still also fewer.Relevant progress mainly contains: the people such as Yang proposed the aurora sequence category theory based on Hidden Markov Model (HMM) in 2013 in article " Yang Qiuju.Auroral Events Detection and Analysis Based on ASI and UVIImages [D] .Xi ' an:Xidian university; 2013. ", but the method essence is still based on single image feature; In addition, when Han constructs sky in article " Han B; Liao Q; GaoXB.Spatial-Temporal poleward volume local binary patterns for aurora sequences eventdetection.Ruan Jian Xue Bao/Journal of Software; 2014; 25 (9): 2172-2179. ", pole characterizes the STP-LBP feature of operator extraction aurora sequence to LBP, for detecting pole in aurora video to motion artifacts.But this algorithm, only for the feature of pole to arcuation aurora video, does not have universality.Due at present not for four class form aurora videos universality modeling and extract the method for video behavioral characteristics, computing machine directly cannot carry out automatic classification to four class aurora videos, single frames operation can only be carried out, or analyze a certain class aurora video to unicity, classification accuracy and classification effectiveness are all lower.

Summary of the invention

The object of the invention is to the deficiency for above-mentioned prior art, a kind of aurora video classification methods based on dynamic texture model is proposed, to utilize the repetition correlativity between aurora frame of video, start with from the dynamic variation characteristic between aurora frame of video, universality modeling is carried out to four class form aurora videos, extract dynamic texture feature, realize the computer automatic sorting of four class aurora videos.

Technical scheme of the present invention comprises the steps: for achieving the above object

(1) appoint from other aurora video database of marking class and get N number of video composition training set { y ₁, y ₂..., y _k..., y _n, y _ka kth training set sample, k=1,2 ..., N, will remain aurora video composition test set, get a sample as test aurora video y from test set _test;

(2) the dynamic texture feature of aurora video is extracted;

(2a) aurora video y will be tested _testbe expressed as y (t), y (t) ∈ R ^m, m=I ₁× I ₂, I ₁for the line number of current video frame picture element matrix, I ₂for the columns of current video frame picture element matrix, t=1 ..., τ, τ are video totalframes;

(2b) average of all frames is deducted with aurora frame of video y (t) observed obtain video matrix Y:

$Y=\[y\left(1\right)-\stackrel{\‾}{y},y\left(2\right)-\stackrel{\‾}{y},\mathrm{...},y\left(t\right)-\stackrel{\‾}{y},\mathrm{...},y\left(\mathrm{\τ}\right)-\stackrel{\‾}{y}\];$

(2c) SVD decomposition is carried out to video matrix Y, i.e. Y=USV ^t, wherein, U ∈ R ^{m × n}for left basis matrix; S ∈ R ^{n × n}for singular value matrix; V ∈ R ^{n × τ}for right basis matrix;

(2d) represent the observing matrix C of aurora video with left basis matrix U, i.e. C=U, and ask the dynamic texture eigenmatrix X=SV of aurora video ^t, X ∈ R here ^{n × τ}, extract each row of dynamic texture eigenmatrix X as dynamic texture characteristic frame x (t), namely X=[x (1), x (2) ..., x (t) ..., x (τ)];

(3) with the dynamic texture eigenmatrix X obtained, the state-transition matrix A of t dynamic texture characteristic frame x (t) to t+1 dynamic texture characteristic frame x (t+1) is calculated,

$\begin{array}{c}A=\arg \min \left|\right|X_{2,\mathrm{...},\mathrm{\τ}}-{\mathrm{AX}}_{1,\mathrm{...},\mathrm{\τ}-1}|{|}_F^2\\ =\left(X_{2,\mathrm{...},\mathrm{\τ}}{X}_{1,\mathrm{...},\mathrm{\τ}-1}\right){\left(X_{1,\mathrm{...},\mathrm{\τ}-1}{X}_{1,\mathrm{...},\mathrm{\τ}-1}^T\right)}^{-1}\end{array},$

X in formula _{1 ..., τ-1}=[x (1), x (2) ..., x (t) ..., x (τ-1)]; X _{2 ..., τ}=[x (2), x (3) ..., x (t) ..., x (τ)]; represent the F norm asking matrix;

(4) test aurora video y is asked _testto training set sample y _kmartin's distance:

(4a) with calculating test pole light video y _testobserving matrix C and the method for state-transition matrix A, calculation training collection { y ₁, y ₂..., y _k..., y _nmiddle training set sample y _kobserving matrix C _kwith state-transition matrix A _k, obtain observing matrix collection { A ₁, A ₂..., A _k..., A _nand state-transition matrix collection { C ₁, C ₂..., C _k..., C _n, wherein k=1,2 ..., N;

(4b) according to observing matrix and state-transition matrix, test aurora video y is calculated _testto training set sample y _kmartin distance d ²(y _test, y _k), k=1,2 ..., N;

(5) according to the Martin's distance between aurora video, to test aurora video y _testcarry out minimum distance classification:

By the N number of Martin distance d obtained in step 3 ²(y _test, y _k) by order arrangement from small to large, take out minimum Martin's distance corresponding aurora sequences y _min, and will with this aurora sequences y _minplesiomorphic test aurora video y _testbe divided into and aurora sequences y _minsame class, completes test aurora video y _testautomatic classification.

Tool of the present invention has the following advantages:

The present invention utilizes SVD to decompose the dynamic texture feature extracting aurora video, take full advantage of the repetition correlativity between aurora frame of video, have found the dynamic perfromance of four class aurora videos, overcome prior art to start with from single frames feature, cause model generalization scarce capacity, the shortcoming that counting yield is not high, makes the method can carry out universality modeling to four class form aurora videos, realizes the computer automatic sorting of four class aurora videos.

Accompanying drawing explanation

Fig. 1 is embodiments of the invention process flow diagrams;

Fig. 2 is with the visualization result figure of the present invention to arcuation aurora video observing matrix;

Fig. 3 is with the visualization result figure of the present invention to radial aurora video observing matrix;

Fig. 4 is with the visualization result figure of the present invention to focus shape aurora video observing matrix;

Fig. 5 is with the visualization result figure of the present invention to valance shape aurora video observing matrix.

Embodiment

Below in conjunction with accompanying drawing, content of the present invention and effect are described further.

One. know-why

The present invention utilizes the repetition correlativity between aurora frame of video, starts with from the dynamic variation characteristic between aurora frame of video, four class form aurora videos is carried out to universality modeling and extracts feature, realizes the classification to aurora video.

Its modeling process is: tie up dynamic texture characteristic sequence frame x (t) with n and characterize m dimension aurora frame of video y (t): wherein, y (t) ∈ R ^m, m=I ₁× I ₂, I ₁× I ₂for the dimension of current video frame picture element matrix, t=1 ..., τ, τ are video totalframes; X (t) ∈ R ⁿand n < < m; C ∈ R ^{m × n}represent observing matrix; W (t) ∈ R ^mfor mean value is 0, variance matrix is zI _mwhite Gaussian noise; represent average on time orientation of aurora sequences y (t) that observe.And dynamic texture characteristic sequence x (t) is expressed as second-order stationary stochastic process: x (t+1)=Ax (t)+v (t).Wherein, parameter A ∈ R ^{n × n}represent state-transition matrix; V (t) ~ N (0, Q) represents that mean value is 0, and variance matrix is the white Gaussian noise of Q.

The average of all frames is deducted with aurora frame of video y (t) observed obtain video matrix Y:

$Y=\&lsqb;y\left(1\right)-\stackrel{\&OverBar;}{y},y\left(2\right)-\stackrel{\&OverBar;}{y},\mathrm{...},y\left(t\right)-\stackrel{\&OverBar;}{y},\mathrm{...},y\left(\mathrm{\&tau;}\right)-\stackrel{\&OverBar;}{y}\&rsqb;;$

SVD decomposition is carried out to video matrix Y, is decomposed into left basis matrix U ∈ R ^{m × n}; Singular value matrix S ∈ R ^{n × n}; Right basis matrix V ∈ R ^{n × τ}, i.e. Y=USV ^t.

Represent the observing matrix C of aurora video with left basis matrix U, i.e. C=U, and ask the dynamic texture eigenmatrix X=SV of aurora video ^t, X ∈ R here ^{n × τ}, extract each row of dynamic texture eigenmatrix X as dynamic texture characteristic frame x (t), namely X=[x (1), x (2) ..., x (t) ..., x (τ)];

With the dynamic texture eigenmatrix X obtained, calculate the state-transition matrix A of t dynamic texture characteristic frame x (t) to t+1 dynamic texture characteristic frame x (t+1),

$\begin{array}{c}A=\arg \min \left|\right|X_{2,\mathrm{...},\mathrm{\&tau;}}-{\mathrm{AX}}_{1,\mathrm{...},\mathrm{\&tau;}-1}|{|}_F^2\\ =\left(X_{2,\mathrm{...},\mathrm{\&tau;}}{X}_{1,\mathrm{...},\mathrm{\&tau;}-1}\right){\left(X_{1,\mathrm{...},\mathrm{\&tau;}-1}{X}_{1,\mathrm{...},\mathrm{\&tau;}-1}^T\right)}^{-1}\end{array},$

X in formula _{1 ..., τ-1}=[x (1), x (2) ..., x (t) ..., x (τ-1)]; X _{2 ..., τ}=[x (2), x (3) ..., x (t) ..., x (τ)]; represent the F norm asking matrix.

For test aurora video y _testwith any one training set sample y _k, ask test aurora video y as stated above _testwith training set sample y _kobserving matrix C and C _kand state-transition matrix A and A _k, and according to observing matrix C _aand C _band state-transition matrix A _aand A _b, build expansion observing matrix:

Φ _test＝[C ^T,A ^TC ^T,(A ^T) ²C ^T,...,(A ^T) ⁿC ^T,...]

${\mathrm{\&Phi;}}_k=\&lsqb;C_k^T,A_k^T{C}_k^T,{\left(A_k^T\right)}^2{C}_k^T,\mathrm{...},{\left(A_k^T\right)}^n{C}_k^T,\mathrm{...}\&rsqb;$

According to expansion observing matrix Φ _testand Φ _kbetween i-th characteristic angle θ _i, calculate Martin's distance $d^2(y_{test},y_k)=-\log \underset{i=1}{\overset{n}{\&Pi;}}{\cos }^2{\mathrm{\&theta;}}_i.$

According to the Martin's distance between aurora video, to test aurora video y _testcarry out minimum distance classification.

Two. performing step

Step 1: obtain training aurora video and test aurora video.

Appoint from other aurora video database of marking class and get N number of video composition training set { y ₁, y ₂..., y _k..., y _n, wherein y _ka kth training aurora video, k=1,2 ..., N;

By remaining aurora video composition test set, from test set, get a sample as test aurora video y _test.

Step 2: extract test aurora video y _testdynamic texture feature.

(2a) aurora video y will be tested _testbe expressed as y (t), y (t) ∈ R ^m, m=I ₁× I ₂, I ₁for the line number of current video frame picture element matrix, I ₂for the columns of current video frame picture element matrix, t=1 ..., τ, τ are video totalframes;

(2b) average of all frames is deducted with aurora frame of video y (t) observed obtain video matrix Y:

$Y=\&lsqb;y\left(1\right)-\stackrel{\&OverBar;}{y},y\left(2\right)-\stackrel{\&OverBar;}{y},\mathrm{...},y\left(t\right)-\stackrel{\&OverBar;}{y},\mathrm{...},y\left(\mathrm{\&tau;}\right)-\stackrel{\&OverBar;}{y}\&rsqb;;$

(2c) SVD decomposition is carried out to video matrix Y, i.e. Y=USV ^t, wherein, U ∈ R ^{m × n}for left basis matrix; S ∈ R ^{n × n}for singular value matrix; V ∈ R ^{n × τ}for right basis matrix;

(2d) represent the observing matrix C of aurora video with left basis matrix U, i.e. C=U, and ask the dynamic texture eigenmatrix X=SV of aurora video ^t, X ∈ R here ^{n × τ}, extract each row of dynamic texture eigenmatrix X as dynamic texture characteristic frame x (t), namely X=[x (1), x (2) ..., x (t) ..., x (τ)];

Step 3: with the dynamic texture eigenmatrix X obtained, calculate the state-transition matrix A of t dynamic texture characteristic frame x (t) to t+1 dynamic texture characteristic frame x (t+1),

$\begin{array}{c}A=\arg \min \left|\right|X_{2,\mathrm{...},\mathrm{\&tau;}}-{\mathrm{AX}}_{1,\mathrm{...},\mathrm{\&tau;}-1}|{|}_F^2\\ =\left(X_{2,\mathrm{...},\mathrm{\&tau;}}{X}_{1,\mathrm{...},\mathrm{\&tau;}-1}\right){\left(X_{1,\mathrm{...},\mathrm{\&tau;}-1}{X}_{1,\mathrm{...},\mathrm{\&tau;}-1}^T\right)}^{-1}\end{array},$

X in formula _{1 ..., τ-1}=[x (1), x (2) ..., x (t) ..., x (τ-1)]; X _{2 ..., τ}=[x (2), x (3) ..., x (t) ..., x (τ)]; represent the F norm asking matrix;

Step 4: ask test aurora video y _testto training set sample y _kmartin's distance.

(4a) with calculating test pole light video y _testobserving matrix C and the method for state-transition matrix A, calculation training collection { y ₁, y ₂..., y _k..., y _nmiddle training set sample y _kobserving matrix C _kwith state-transition matrix A _k, obtain observing matrix collection { A ₁, A ₂..., A _k..., A _nand state-transition matrix collection { C ₁, C ₂..., C _k..., C _n, wherein k=1,2 ..., N;

(4b) according to observing matrix and state-transition matrix, test aurora video y is calculated _testto training set sample y _kmartin distance d ²(y _test, y _k), k=1,2 ..., N;

(3b1) with observing matrix C and C _kand state-transition matrix A and A _k, build expansion observing matrix:

Φ _test＝[C ^T,A ^TC ^T,(A ^T) ²C ^T,...,(A ^T) ⁿC ^T,...]

${\mathrm{\&Phi;}}_k=\&lsqb;C_k^T,A_k^T{C}_k^T,{\left(A_k^T\right)}^2{C}_k^T,\mathrm{...},{\left(A_k^T\right)}^n{C}_k^T,\mathrm{...}\&rsqb;$

(3b2) according to expansion observing matrix Φ _testand Φ _kbetween i-th characteristic angle θ _i, calculate Martin's distance $d^2(y_{test},y_k)=-log\underset{i=1}{\overset{n}{\&Pi;}}{\cos }^2{\mathrm{\&theta;}}_i.$

Step 5: according to the Martin's distance between aurora video, to test aurora video y _testcarry out minimum distance classification.

By the N number of Martin distance d obtained in step 3 ²(y _test, y _k) by order arrangement from small to large, take out minimum Martin's distance corresponding aurora sequences y _min, and will with this aurora sequences y _minplesiomorphic test aurora video y _testbe divided into and aurora sequences y _minsame class, completes test aurora video y _testclassification.

Effect of the present invention is further illustrated by following emulation experiment:

1. simulated conditions and method:

Hardware platform is: Intel Core i5,2.93GHz, 3.45GB RAM;

Software platform is: the MATLAB R2012b under Windows7 operating system;

Experiment data: in Dec, 2003 in the all-sky aurora data of China's Arctic Yellow River Station by 2004 totally 115557 width G-band images carry out manual markings, remove the invalid data that the factors such as weather cause, mark 93 radial aurora sequences, 102 arcuation aurora sequences, 73 focus shape aurora sequences, 95 valance shape aurora sequences, sequence length is between 15 frames to 35 frames, forms typical aurora video database and carries out four classification experiments.

2. emulate content and result:

Experiment simulation 1, from typical aurora video database, random choose goes out each one of four quasi-representative aurora video, i.e. arcuation aurora video, radial aurora video, focus shape aurora video and valance shape aurora video.Carry out down-sampling process to every width frame of video, the frame of video down-sampling by 512 × 512 sizes is 128 × 128.Solve observing matrix C by method of the present invention, the C obtained is the matrix of 16384 × 10 sizes, each column weight of observing matrix C is newly converted into 128 × 128 large minor matrixs, then shows in the form of images.

To the display result of arcuation aurora video as shown in Figure 2, wherein Fig. 2 (a) is former frame of video, and Fig. 2 (b) is the visualization result of the observing matrix of arcuation aurora video;

To the display result of radial aurora video as shown in Figure 3, wherein Fig. 3 (a) is former frame of video, and Fig. 3 (b) is the visualization result of the observing matrix of radial aurora video;

To the display result of focus shape aurora video as shown in Figure 4, wherein Fig. 4 (a) is former frame of video, and Fig. 4 (b) is the visualization result of the observing matrix of focus shape aurora video;

To the display result of valance shape aurora video as shown in Figure 5, wherein Fig. 5 (a) is former frame of video, and Fig. 5 (b) is the visualization result of the observing matrix of valance shape aurora video;

Fig. 2-Fig. 5 shows, observing matrix C only has former row to contain the texture information of aurora image sequence usually, proves the repetition correlativity in the certain life period direction of aurora image sequence, and dynamic texture model effectively can extract its spatial texture feature.

Experiment simulation 2, at training set video number N=30, when 50,80,100,120, respectively with the present invention to typical aurora video database classification experiments, obtain classification accuracy as shown in the table.

Table 1 aurora visual classification accuracy rate

As can be seen from Table 1, when number of training is more than 100, classification accuracy is 77.09%, and category of model accuracy rate is higher, can realize the automatic classification of computing machine to aurora video.

Claims (2)

1., based on the aurora video classification methods that dynamic texture model characterizes, comprise the steps:

(1) appoint from other aurora video database of marking class and get N number of video composition training set { y ₁, y ₂..., y _k..., y _n, y _ka kth training set sample, k=1,2 ..., N, will remain aurora video composition test set, get a sample as test aurora video y from test set _test;

(2) the dynamic texture feature of aurora video is extracted;

(2a) aurora video y will be tested _testbe expressed as y (t), y (t) ∈ R ^m, m=I ₁× I ₂, I ₁for the line number of current video frame picture element matrix, I ₂for the columns of current video frame picture element matrix, t=1 ..., τ, τ are video totalframes;

(2b) average of all frames is deducted with aurora frame of video y (t) observed obtain video matrix Y:

$Y=\&lsqb;y\left(1\right)-\stackrel{\&OverBar;}{y},y\left(2\right)-\stackrel{\&OverBar;}{y},...,y\left(t\right)-\stackrel{\&OverBar;}{y},...,y\left(\mathrm{\&tau;}\right)-\stackrel{\&OverBar;}{y}\&rsqb;;$

(2c) SVD decomposition is carried out to video matrix Y, i.e. Y=USV ^t, wherein, U ∈ R ^{m × n}for left basis matrix; S ∈ R ^{n × n}for singular value matrix; V ∈ R ^{n × τ}for right basis matrix;

(2d) represent the observing matrix C of aurora video with left basis matrix U, i.e. C=U, and ask the dynamic texture eigenmatrix X=SV of aurora video ^t, X ∈ R here ^{n × τ}, extract each row of dynamic texture eigenmatrix X as dynamic texture characteristic frame x (t), namely X=[x (1), x (2) ..., x (t) ..., x (τ)];

(3) with the dynamic texture eigenmatrix X obtained, the state-transition matrix A of t dynamic texture characteristic frame x (t) to t+1 dynamic texture characteristic frame x (t+1) is calculated,

$\begin{array}{c}A=\arg \min \left|\right|X_{2,\mathrm{...},\mathrm{\&tau;}}-{\mathrm{AX}}_{1,\mathrm{...},\mathrm{\&tau;}-1}|{|}_F^2\\ =\left(X_{2,\mathrm{...},\mathrm{\&tau;}}{X}_{1,\mathrm{...},\mathrm{\&tau;}-1}\right){\left(X_{1,\mathrm{...},\mathrm{\&tau;}-1}{X}_{1,\mathrm{...},\mathrm{\&tau;}-1}^T\right)}^{-1}\end{array},$

X in formula _{1 ..., τ-1}=[x (1), x (2) ..., x (t) ..., x (τ-1)]; X _{2 ..., τ}=[x (2), x (3) ..., x (t) ..., x (τ)]; represent the F norm asking matrix;

(4) test aurora video y is asked _testto training set sample y _kmartin's distance:

(4a) with calculating test pole light video y _testobserving matrix C and the method for state-transition matrix A, calculation training collection { y ₁, y ₂..., y _k..., y _nmiddle training set sample y _kobserving matrix C _kwith state-transition matrix A _k, obtain observing matrix collection { A ₁, A ₂..., A _k..., A _nand state-transition matrix collection { C ₁, C ₂..., C _k..., C _n, wherein k=1,2 ..., N;

(4b) according to observing matrix and state-transition matrix, test aurora video y is calculated _testto training set sample y _kmartin distance d ²(y _test, y _k), k=1,2 ..., N;

(5) according to the Martin's distance between aurora video, to test aurora video y _testcarry out minimum distance classification:

By the N number of Martin distance d obtained in step 3 ²(y _test, y _k) by order arrangement from small to large, take out minimum Martin's distance corresponding aurora sequences y _min, and will with this aurora sequences y _minplesiomorphic test aurora video y _testbe divided into and aurora sequences y _minsame class, completes test aurora video y _testclassification.

2. method according to claim 1, solves test aurora video y in wherein said step (4b) _testto training set sample y _kmartin distance d ²(y _test, y _k), carry out as follows:

(4b1) with observing matrix C and C _kand state-transition matrix A and A _k, build expansion observing matrix:

$\begin{array}{c}{\mathrm{\&Phi;}}_{test}=\&lsqb;C^T,A^T{C}^T,{\left(A^T\right)}^2{C}^T,\mathrm{...},{\left(A^T\right)}^n{C}^T,\mathrm{...}\&rsqb;\\ {\mathrm{\&Phi;}}_k=\&lsqb;C_k^T,A_k^T{C}_k^T,{\left(A_k^T\right)}^2{C}_k^T,\mathrm{...},{\left(A_k^T\right)}^n{C}_k^T,\mathrm{...}\&rsqb;\end{array};$

(4b2) according to expansion observing matrix Φ _testand Φ _kbetween i-th characteristic angle θ _i, calculate Martin's distance:

$d^2(y_{test},y_k)=-log\underset{i=1}{\overset{n}{\&Pi;}}{\cos }^2{\mathrm{\&theta;}}_i.$