CN108960295A - A kind of full polarimetric SAR for multidate extracts the method for feature and the classification method of vegetation distributed areas - Google Patents
A kind of full polarimetric SAR for multidate extracts the method for feature and the classification method of vegetation distributed areas Download PDFInfo
- Publication number
- CN108960295A CN108960295A CN201810608408.5A CN201810608408A CN108960295A CN 108960295 A CN108960295 A CN 108960295A CN 201810608408 A CN201810608408 A CN 201810608408A CN 108960295 A CN108960295 A CN 108960295A
- Authority
- CN
- China
- Prior art keywords
- target
- tensor
- multidate
- polarimetric sar
- full polarimetric
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 56
- 239000000284 extract Substances 0.000 title claims abstract description 14
- 239000011159 matrix material Substances 0.000 claims abstract description 60
- 230000009467 reduction Effects 0.000 claims abstract description 20
- 238000005352 clarification Methods 0.000 claims abstract description 14
- 238000000354 decomposition reaction Methods 0.000 claims abstract description 8
- 239000000203 mixture Substances 0.000 claims abstract description 8
- 230000011218 segmentation Effects 0.000 claims description 13
- 238000004364 calculation method Methods 0.000 claims description 5
- 238000012549 training Methods 0.000 abstract description 15
- 230000008569 process Effects 0.000 abstract description 5
- 238000000605 extraction Methods 0.000 abstract description 4
- 240000008042 Zea mays Species 0.000 description 44
- 235000002017 Zea mays subsp mays Nutrition 0.000 description 44
- 235000005824 Zea mays ssp. parviglumis Nutrition 0.000 description 41
- 235000005822 corn Nutrition 0.000 description 41
- 230000010287 polarization Effects 0.000 description 13
- 244000068988 Glycine max Species 0.000 description 7
- 235000010469 Glycine max Nutrition 0.000 description 7
- VMXUWOKSQNHOCA-UKTHLTGXSA-N ranitidine Chemical compound [O-][N+](=O)\C=C(/NC)NCCSCC1=CC=C(CN(C)C)O1 VMXUWOKSQNHOCA-UKTHLTGXSA-N 0.000 description 7
- 238000012544 monitoring process Methods 0.000 description 5
- QDGIAPPCJRFVEK-UHFFFAOYSA-N (1-methylpiperidin-4-yl) 2,2-bis(4-chlorophenoxy)acetate Chemical compound C1CN(C)CCC1OC(=O)C(OC=1C=CC(Cl)=CC=1)OC1=CC=C(Cl)C=C1 QDGIAPPCJRFVEK-UHFFFAOYSA-N 0.000 description 4
- 238000004519 manufacturing process Methods 0.000 description 4
- 238000012986 modification Methods 0.000 description 4
- 230000004048 modification Effects 0.000 description 4
- 235000016383 Zea mays subsp huehuetenangensis Nutrition 0.000 description 3
- 230000008901 benefit Effects 0.000 description 3
- 230000008859 change Effects 0.000 description 3
- 238000001914 filtration Methods 0.000 description 3
- 235000009973 maize Nutrition 0.000 description 3
- 238000012545 processing Methods 0.000 description 3
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 235000013305 food Nutrition 0.000 description 2
- 238000003384 imaging method Methods 0.000 description 2
- 238000005259 measurement Methods 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 239000002689 soil Substances 0.000 description 2
- 230000009466 transformation Effects 0.000 description 2
- 240000002853 Nelumbo nucifera Species 0.000 description 1
- 235000006508 Nelumbo nucifera Nutrition 0.000 description 1
- 235000006510 Nelumbo pentapetala Nutrition 0.000 description 1
- 240000007594 Oryza sativa Species 0.000 description 1
- 235000007164 Oryza sativa Nutrition 0.000 description 1
- 244000046052 Phaseolus vulgaris Species 0.000 description 1
- 235000010627 Phaseolus vulgaris Nutrition 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 235000013339 cereals Nutrition 0.000 description 1
- 230000001427 coherent effect Effects 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000012937 correction Methods 0.000 description 1
- 230000001186 cumulative effect Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000003709 image segmentation Methods 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 235000009566 rice Nutrition 0.000 description 1
- 230000000630 rising effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10032—Satellite or aerial image; Remote sensing
- G06T2207/10044—Radar image
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30181—Earth observation
- G06T2207/30188—Vegetation; Agriculture
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Radar Systems Or Details Thereof (AREA)
Abstract
This application provides a kind of full polarimetric SARs for multidate to extract the method for feature and the classification method of vegetation distributed areas.The method that the full polarimetric SAR for multidate extracts feature includes: to obtain the full polarimetric SAR of several different phases of the same area comprising several vegetation;The full polarimetric SAR of the multidate is split to obtain several targets;Multistage tensor is created for each target, the first rank and second-order in the multistage tensor indicate the coherence matrix of the target or the row and column of covariance matrix, and the third rank in the multistage tensor indicates different phases;Dimensionality reduction is carried out using the multistage tensor of the main composition decomposition algorithm of polyteny to each target, to obtain each clarification of objective vector.The application improves the otherness of vegetation by the feature extraction of the full polarimetric SAR to multidate.DT classification results show that the present processes can also obtain higher vegetation accuracy of identification even if in the case where training sample is smaller.
Description
Technical field
This application involves remote sensing application fields, and in particular to and it is the method that the full polarimetric SAR of multidate extracts feature,
And the classification method of the vegetation distributed areas of the full polarimetric SAR of multidate.
Background technique
Corn is one of most important industrial crops in the whole world, is the important component in Chinese agriculture, in agricultural and work
Industry field all occupies important function, and therefore, how to carry out effective monitoring to corn is the problem of being worth people to contemplate.
We are by the available accurate image to earth's surface of remote sensing technology, and the analysis to remote sensing images is as a result, not only may be used
It with reallocation soil and ensures physical norm for soils, and provides convenience for monitoring crop growth.
But in fact, the state of cultivation different due to China various regions, corn, which is easy to other vegetation with surrounding, to be obscured.Pass through
Pixel level and the separability of target level compare, it can be found that in the corn growth middle and later periods, backscattering coefficient not can guarantee
Separability between corn and other atural objects.
Summary of the invention
The application provides a kind of method that the full polarimetric SAR for multidate extracts feature;A kind of full pole of multidate
Change the classification method of the vegetation distributed areas of SAR image;To solve the problems, such as the separability between corn and other atural objects.
In order to solve the above-mentioned technical problem, the embodiment of the present application provides the following technical solution:
This application provides a kind of methods that the full polarimetric SAR for multidate extracts feature, and the method includes such as
Lower step:
Obtain the full polarimetric SAR of several different phases of the same area comprising several vegetation;
The full polarimetric SAR of the multidate is split to obtain several targets;
Multistage tensor is created for each target, the first rank and second-order in the multistage tensor indicate the relevant of the target
The row and column of matrix or covariance matrix, the third rank in the multistage tensor indicate different phases;
Dimensionality reduction is carried out using the multistage tensor of the main composition decomposition algorithm of polyteny to each target, to obtain each mesh
Target feature vector.
Optionally, the coherence matrix or covariance matrix include: the full-polarization SAR of several different phases of the same area
Strength information and/or phase information in image.
Optionally, before the full polarimetric SAR to the multidate is split, further includes:
The full polarimetric SAR of the multidate is registrated, and carries out the mean filter of default specification.
Optionally, the step of full polarimetric SAR to the multidate is split to obtain several targets packet
It includes:
Using the average coherence matrix or average covariance matrices of the full polarimetric SAR of the multidate, super picture is calculated
The distance measure of element segmentation;
The pixel of the full polarimetric SAR of the multidate is aggregated into several mesh using the distance measure of super-pixel segmentation
Mark;
Wherein, in the minimum pixel of gradient in the neighborhood for being centrally disposed in default specification of the target.
It is optionally, described to carry out dimensionality reduction using the multistage tensor of the main composition decomposition algorithm of polyteny to each target,
To include: the step of obtaining each clarification of objective vector
Projection matrix is obtained according to the multistage tensor of each targetWherein Ln<In, LnIt is
Dimension after dimensionality reduction, InFor original dimension;
Based on the iterative calculation of the projection matrix, dimensionality reduction is carried out to the multistage tensor of each target, it is each to obtain
A clarification of objective vector.
Further, the projection matrixWherein Ln<In, LnIt is the dimension after dimensionality reduction, InFor
Original dimension, specific formula is as follows:
WhenWhen,
Wherein,
I indicates the i-th rank, and n indicates order;
M indicates that m-th of target tensor, M indicate the quantity of target tensor;
Q value range is [0,1], is user's custom parameter, and letter is saved after projection for controlling original tensor
The ratio of breath.
Further, which is characterized in that the iterative calculation based on the projection matrix, to the described more of each target
Rank tensor carries out dimensionality reduction, to include: the step of obtaining each clarification of objective vector
Obtain several sample objects with tensor representation;
Polyteny projection is carried out to the sample object to calculate, and obtains tensor Y;
Calculating is iterated with the tensor Y based on including the projection matrix;
Meet preset condition when total divergence of the tensor Y converges to or when the number of iterations reaches preset value, then obtain each
A clarification of objective vector.
Optionally, in the multistage tensor, further include fourth order and the 5th rank, indicate the length in the target window neighborhood
And width.
This application provides a kind of classification methods of the vegetation distributed areas of the full polarimetric SAR of multidate, comprising:
Using method described in any one of claim 1-8 extract comprising several vegetation the same area it is several not
With each clarification of objective vector of the full polarimetric SAR of phase;
The feature vector of at least one phase based on each target determines the vegetation classification belonging to it using classifier.
Optionally, the average coherence matrix or average covariance matrices of the target indicate are as follows:
Wherein,
Indicate average coherence matrix or average covariance matrices;
TiIndicate the coherence matrix of ith pixel;
K indicates that the target is made of k pixel.
Disclosure based on the above embodiment can know, the embodiment of the present application have it is following the utility model has the advantages that
The full polarimetric SAR that this application provides a kind of for multidate extract feature method and multidate it is complete
The classification method of the vegetation distributed areas of polarimetric SAR image.The method that the full polarimetric SAR for multidate extracts feature
It include: the full polarimetric SAR for obtaining several different phases of the same area comprising several vegetation;To the complete of the multidate
Polarimetric SAR image is split to obtain several targets;Multistage tensor is created for each target, first in the multistage tensor
Rank and second-order indicate the coherence matrix of the target or the row and column of covariance matrix, and the third rank in the multistage tensor indicates
Different phases;Dimensionality reduction is carried out using the multistage tensor of the main composition decomposition algorithm of polyteny to each target, it is each to obtain
Clarification of objective vector.And the feature vector of at least one phase based on each target, it is determined belonging to it using classifier
Vegetation classification.The application improves the otherness of vegetation by the feature extraction of the full polarimetric SAR to multidate.DT classification
The result shows that even if the present processes can also obtain higher vegetation identification essence in the case where training sample is smaller
Degree.The combination of especially 3 phase data is optimal selection.
Detailed description of the invention
Fig. 1 is the flow chart that the full polarimetric SAR that the embodiment of the present application is multidate extracts the method for feature;
Fig. 2 is the stream of the classification method of the vegetation distributed areas of the full polarimetric SAR of the multidate of the embodiment of the present application
Cheng Tu.
Specific embodiment
In the following, being described in detail in conjunction with specific embodiment of the attached drawing to the application, but not as the restriction of the application.
It should be understood that various modifications can be made to disclosed embodiments.Therefore, description above should not regard
To limit, and only as the example of embodiment.Those skilled in the art can in the scope and spirit of the present application into
Other modifications of row.
The attached drawing being included in the description and forms part of the description shows embodiments herein, and with it is upper
What face provided is used to explain the application together to substantially description and the detailed description given below to embodiment of the application
Principle.
By the description of the preferred form with reference to the accompanying drawings to the embodiment for being given as non-limiting example, the application's
These and other characteristic will become apparent.
It is also understood that although the application is described referring to some specific examples, those skilled in the art
Member realizes many other equivalents of the application in which can determine, they have feature as claimed in claim and therefore all
In the protection scope defined by whereby.
When read in conjunction with the accompanying drawings, in view of following detailed description, above and other aspect, the feature and advantage of the application will become
It is more readily apparent.
The specific embodiment of the application is described hereinafter with reference to attached drawing;It will be appreciated, however, that the disclosed embodiments are only
Various ways implementation can be used in the example of the application.Known and/or duplicate function and structure and be not described in detail to avoid
Unnecessary or extra details makes the application smudgy.Therefore, specific structural and functionality disclosed herein is thin
Section is not intended to restrictions, but as just the basis of claim and representative basis be used to instructing those skilled in the art with
Substantially any appropriate detailed construction diversely uses the application.
This specification can be used phrase " in one embodiment ", " in another embodiment ", " in another embodiment
In " or " in other embodiments ", it can be referred to one or more of the identical or different embodiment according to the application.
Needed for food in order to meet nearly 1,400,000,000 population, China is faced with stern challenge, people for food production and
The concern of safety is also gradually increasing.As one of cereal most important in world wide, corn is all accounted in agricultural and industrial circle
According to important function, therefore, how to carry out effective monitoring to corn is the problem of being worth people to contemplate.
But due to the different state of cultivation in China various regions, corn, which is easy to other vegetation with surrounding, to be obscured.Pass through pixel water
The separability of gentle target level compares, it can be found that in the corn growth middle and later periods, existing monitoring method not can guarantee corn with
Separability between other atural objects.
The application provides a kind of method that the full polarimetric SAR for multidate extracts feature;The application also provides one kind
The classification method of the vegetation distributed areas of the full polarimetric SAR of multidate.It carries out one by one in the following embodiments specifically
It is bright.
To first embodiment provided by the present application, i.e., the method for a kind of full polarimetric SAR extraction feature for multidate
Embodiment.
The present embodiment is described in detail below with reference to Fig. 1, wherein Fig. 1 is that the embodiment of the present application is multidate
Full polarimetric SAR extracts the flow chart of the method for feature.
Step S101 obtains the full polarimetric SAR of several different phases of the same area comprising several vegetation.
Synthetic aperture radar (Synthetic Aperture Radar, abbreviation SAR), is a kind of high-resolution imaging thunder
It reaches, similar photo-optical high resolution radar image can be obtained under the extremely low meteorological condition of visibility., SAR is due to can be with
The influence for overcoming weather and time has more monitoring advantage compared with optical remote sensing.
The SAR image is exactly the image obtained by SAR.
Polarization, is one of essential attribute of electromagnetic wave, is the another dimension important information in addition to frequency, amplitude, phase, complete
The whole vector property feature for embodying electromagnetic wave.As polarization SAR (PolSAR) technology becomes more mature, what SAR can be provided
Information content is also richer.Four kinds of polarization body formula patterns --- HH, VV, HV, VH is commonly used in SAR.The above two are polarization in the same direction, after
The two is incorgruous (intersection) polarization.Full-polarization SAR (FP SAR) technical difficulty highest, can provide the intensity and phase in 4 channels
Dry phase information, it is desirable that while emitting H and V, that is, HH, HV, VV, tetra- kinds of polarization body formula patterns of VH.
The full polarimetric SAR is exactly that SAR passes through the intensity in 4 channels and the image letter of coherent phase information acquisition
Breath.
Multidate is often referred to the feature that one group of remote sensing image of reflection has on time series.In a broad aspect, all not
With a group image, map or the geodata of the same region that the time obtains, it all can be considered the data of " multidate ".
The full polarimetric SAR of several different phases of described the same area, is exactly several different times of same region
Full polarimetric SAR.For example, Tianjin Wuqing District (centre coordinate is about 117 ° of 2 ' 2 " E, 39 ° of 28 ' 29 " N), the region
It is one of most important corn producing region of China positioned at the North China Plain.Three scape shadows of the region corn are obtained by full-polarization SAR
Picture, such as DOY207, DOY231, DOY255 image.
Year day of year (abbreviation DOY) is the method on the only Continuous plus date used in 1 year, is from January 1 current year
Start the number of days calculated.Such as: DOY207 be exactly since the January 1 of current year the 207th day.
Registration refers in the same area with of the geographical coordinate of different imaging means different images figure obtained
Match.Processing including three aspect of geometric correction, projective transformation and common scale ruler.
The full polarimetric SAR to the multidate is registrated, and exactly obtains the same area by full-polarization SAR
The full polarimetric SAR of interior different periods chooses the image of one of those period as standard, carries out geometry to other images
It corrects, the processing of three aspect of projective transformation and common scale ruler.For example, in Tianjin Wuqing District, (centre coordinate is about 117 ° 2 '
2 " E, 39 ° of 28 ' 29 " N) obtain corn three scape image DOY207, DOY231, DOY255 image, then can be by DOY207's
Image is registrated the image of DOY231 and DOY255 as standard.
Mean filter is typical linear filtering algorithm, it refers on the image to object pixel to a template, the mould
Plate include surrounding adjacent pixels (such as the Filtering Template of 3 × 3 sizes be centered on object pixel around 8 pictures
Element, in addition object pixel itself, constitutes the Filtering Template comprising 9 pixels), then being averaged with the entire pixels in template
Value replaces original pixel value.In the present embodiment using mean filter purpose be reduce spot make an uproar, meanwhile, avoid lose difference obtain
The heterogeneity of polarization characteristic in the time is taken, the heterogeneity of the polarization characteristic plays a significant role accurately identifying for vegetation.
The mean filter for carrying out default specification is exactly to choose the window of default specification as sliding window, to the window
The image of mouth carries out mean filter, for example, default specification is 5 × 5.
Step S102 is split the full polarimetric SAR of the multidate to obtain several targets.Specific steps packet
It includes:
Step S102-1 utilizes the average coherence matrix or average covariance square of the full polarimetric SAR of the multidate
Battle array, calculates the distance measure of super-pixel segmentation.
Super-pixel, the cell that a series of and color adjacent by positions, brightness, the similar pixel of Texture eigenvalue form
Domain.These zonules remain the effective information of further progress image segmentation mostly, and will not generally destroy objects in images
Boundary information.
Super-pixel segmentation refers to for digital picture being subdivided into multiple images subregion (pixel in computer vision field
Set) (also referred to as super-pixel) process.
Distance scale between distance measure, that is, measurement different target.In the present embodiment, dimensional measurement is borrowed
The concept of middle distance measure, indicate the distance between pixel and pixel, the distance between two pixels estimate it is smaller, belong to together
A possibility that a kind of, is bigger.
The distance measure for calculating super-pixel segmentation, is exactly to calculate pixel and picture in normal image or optical image
The difference of color data between element can effectively divide the region of different pixels in image by the difference.The present embodiment is used
The distance measure of super-pixel segmentation replaces the Euclidean distance in primitive color space, and SLIC algorithm is applied in FP SAR image.
In order to simplify operation, the present embodiment, using multidate N scape image full polarimetric SAR average coherence matrix or
Average covariance matrices calculate the distance measure of super-pixel segmentation.
The distance measure of the super-pixel segmentation, comprising: Wishart distance measure, cartesian coordinate distance measure;
The average coherence matrix or covariance matrix using multidate N scape image full polarimetric SAR calculates
Wishart distance measure D:
Wherein,
dw(i, j) indicates the Wishart distance between ith pixel and j-th of target's center;Further, the mesh
Target is centrally disposed in the pixel in the neighborhood of default specification having gradient minimum, so that target's center be avoided to be located at boundary;
TiIndicate the coherence matrix or covariance matrix of ith pixel;Further, the coherence matrix or covariance square
Battle array includes: strength information and/or phase information in the full polarimetric SAR of several different phases of the same area;
ΣjIndicate the average coherence matrix or average covariance matrices of j-th of target;
Indicate the average Wishart distance measure of multidate N scape image full polarimetric SAR;
dsIndicate the space length between ith pixel and j-th of target's center, i.e. cartesian coordinate distance measure,
Coordinate (ri, ci) and (rj, cj) respectively indicate column locations of the pixel in SAR image;
Weight parameter m and area parameters S are customized by the user, for example, to N=3, weight parameter is used in the classification of corn
M=1, area parameters S=15.
Step S102-2, using the distance measure of super-pixel segmentation by the pixel of the full polarimetric SAR of the multidate
Aggregate into several targets.
Step S103 creates multistage tensor for each target, and the first rank and second-order in the multistage tensor indicate should
The coherence matrix of target or the row and column of covariance matrix, the third rank in the multistage tensor indicate different phases.
TensorIt can be considered as a n dimension group, n indicates order.Tensor representation method avoids more
The vectorization procedure usually carried out in phase N scape image full polarimetric SAR treatment process, maintains the data knot of coherence matrix
Structure.
For polarization SAR data, coherence matrix and covariance matrix are all 3 × 3Hermitian positive semidefinite matrixs, they
Belong to Riemann manifold, therefore the form of second-order tensor can be considered as.Further, the coherence matrix or covariance matrix packet
Contain: strength information and phase information in the full polarimetric SAR of several different phases of the same area.For example, for corn
The full polarimetric SAR of several difference phases, each pixel can be expressed as a three rank tensors.Coherence matrix and covariance square
Battle array is all 3 × 3Hermitian positive semidefinite matrix, and therefore, the first rank and second-order are 3, in the present embodiment, due to having selected 3
The full polarimetric SAR of scape difference phase, therefore, third rank are 3.
In the full polarimetric SAR of more different phases, it can simply increase very much order of a tensor number.For example, not
In the case where carrying out super-pixel segmentation, neighborhood and texture information in order to protect data, the multistage tensor can also include the
Quadravalence and the 5th rank indicate length and width in the target window neighborhood.
Step S104 carries out dimensionality reduction using the multistage tensor of the main composition decomposition algorithm of polyteny to each target, with
Obtain each clarification of objective vector.
The main composition decomposition algorithm of polyteny (multilinear principle component analysis, referred to as
MPCA) it is tensor-tensor projecting method, by extracting the raw information of core tensor, reduction feature dimensions can be played
The effect of degree.
Step S104-1 obtains projection matrix according to the multistage tensor of each targetIts
Middle Ln<In, LnIt is the dimension after dimensionality reduction, InFor original dimension.Specific formula is as follows:
WhenWhen,
Wherein,
1)
2)
3) i indicates the i-th rank, and n indicates order;
4) m indicates that m-th of target tensor, M indicate the quantity of target tensor;
5) Q value range is [0,1], is user's custom parameter, is saved after projection for controlling original tensor
The ratio of information.
6) symbol | | | |FIndicate Frobenius norm.For example, for tensorFrobenius
The form that norm can be expressed as,<>indicate two inners product of tensors:
7) matrix U(i)By matrix Φ(i)Maximum LiFeature vector corresponding to a characteristic value is constituted:
8) dimension LiBy matrixCumulative probability determine.By square
Battle arrayEigenvalue λt,1≤t≤IiIt is arranged according to sequence from big to small, and calculator cumulative probability pt,1≤t≤Ii, then LiIt is corresponding
In meeting ptThe maximum t of < Q.
Step S104-2 drops the multistage tensor of each target based on the iterative calculation of the projection matrix
Dimension, to obtain each clarification of objective vector.
Specific algorithm are as follows:
Step S104-2-1 obtains several sample objects with tensor representation.
In a particular application, when extracting feature for the full polarimetric SAR of multidate, often never with the full pole of phase
Change in the target that SAR image generates and select some samples, feature is extracted from sample.Sample object tensor set can be expressed as X
={ X1,X2,…,XM, wherein the quantity of M expression sample object.For example, in the sample object tensor set of corn, Xi∈R3 ×3×N;Represent i-th of sample object.
Step S104-2-2 carries out polyteny projection to the sample object and calculates, obtains tensor Y.
Y=X ×1U(1)×2U(2)…×nU(n); (3)
Wherein, the tensorFor the target tensor sample set X={ X1, X2..., XMIt is multi-thread
Property projection result.
Step S104-2-3 is iterated calculating with the tensor Y based on including the projection matrix.
Matrix U(i)WithMaximum LiA characteristic vector is calculated as initial value by alternately least-squares algorithm
The sequential iteration of formula (2) (3) (1) calculates.
Step S104-2-4 meets preset condition or the number of iterations reaches default when total divergence of the tensor Y converges to
When value, then each clarification of objective vector is obtained.
For example, preset condition isPreset value is that 4, k value is current iteration number.
Based on the concrete application of first embodiment provided by the present application, present invention also provides second embodiments, i.e., a kind of
The classification method of the vegetation distributed areas of the full polarimetric SAR of multidate.Since second embodiment is based on first embodiment, institute
To describe fairly simple, relevant part refers to the corresponding explanation of first embodiment.Described below second is implemented
Example is only schematical.
Fig. 2 shows a kind of classification sides of the vegetation distributed areas of the full polarimetric SAR of multidate provided by the present application
The embodiment of method.Fig. 2 is the classification method of the vegetation distributed areas of the full polarimetric SAR of the multidate of the embodiment of the present application
Flow chart.
Referring to FIG. 2, the application provides a kind of classification side of the vegetation distributed areas of the full polarimetric SAR of multidate
Method, comprising:
The method extracts the complete of several different phases of the same area comprising several vegetation according to first embodiment
Each clarification of objective vector of polarimetric SAR image.
The step 101 of first embodiment the method, 102,103 and 104 and its sub-step, respectively correspond the present embodiment
Step 201,202,203 and 204 and its sub-step.
Described according to first embodiment, output order of a tensor number is determined by parameter Q, represents projection matrix characteristic value
The ratio of original information is kept, thus the different sample object collection with tensor representation must not having same Q value yet
To the output result of identical dimensional.For example, in survey region, with the order of the sample object collection of tensor representation rapidly from 27 (3 ×
3 × 3) rank is reduced to 12 (2 × 2 × 3) ranks and 8 ranks (2 × 2 × 2).As Q=0.82, output tensor order be all 4 (2 × 2 ×
1) rank.
The present embodiment mainly studies 7 class atural object distributed areas classification methods.7 class atural objects are as follows: 1 corn;2 is big
Beans;3 rice;4 lotuses;5 meadows;6 water bodys;7 buildings.The purpose of research is: the full-polarization SAR of the multidate based on tensor representation
The classification of the vegetation distributed areas of image.
For accurately identifying for corn-growing regions, the texture information in plot needs as uniform as possible.Therefore, the target
Average coherence matrix or average covariance matrices indicate are as follows:
Wherein,
Indicate average coherence matrix or average covariance matrices;
TiIndicate the coherence matrix of ith pixel;
K indicates that the target is made of k pixel.
Studies have shown that corn-rice, corn-lotus and corn-meadow separability all highly significants.Corn and soybean
All there is relatively specific cluster, but the two all has certain overlapping in all subspaces, this shows corn and soybean
Polarization characteristic the middle and late growth stage have certain similitude.Corn and the JM of soybean distance are only utilizing all three scape images
When can reach 1.7, and be herein even more by dimensionality reduction bring information loss, therefore its separability is by a degree of shadow
It rings.But if tensor set is also added in the sample of water body and building, the projection model that MPCA algorithm obtains also can accordingly change
Become, the separability between corn and soybean has obtained a degree of promotion.
Further, the difference of corn and soybean on single image is not significant, and there are portions in the variation of different phases
Divide subtle difference, can chart for the corn middle and later periods and useful information is provided.
The step of connecting method described in first embodiment:
Step S205, the feature vector of at least one phase based on each target are determined belonging to it using classifier
Vegetation classification.
After Feature Dimension Reduction, classification can be completed with polyteny classifier, the present embodiment selects the DT of fast speed
Classifier.
Training sample includes whole 7 class atural objects, and the parameter Q of MPCA algorithm is set as 0.95, obtains 4 (2 × 2 × 1) therefrom
Feature after dimension dimensionality reduction.The training dataset of the present embodiment includes 7 class atural object, 1758 samples.
It is concentrated from training data and randomly chooses a certain number of samples, and calculate its mean accuracy, including overall accuracy,
The user's precision and producer's precision of kappa coefficient, corn.The producer's precision and user's precision of corn, training sample ratio
It is improved when rising to 20% by 10%, is further added by training sample, then the producer's precision and user's precision of corn are enclosed respectively
It is fluctuated around mean value 0.9388 and 0.9533.
When training sample ratio is 10%, overall accuracy, kappa coefficient and the maize production person's precision of 231 data of DOY
It is apparently higher than DOY 207 and DOY 255, it was demonstrated that the data at heading stage can preferably portray the characteristic of corn.But 231 He of DOY
The data combination of DOY 255 improves corn recognition result.Producer's precision of corn is compared with single SAR image and DOY 207
It combines and is obviously improved with 231 data, while user's precision of corn has been more than 0.95.The combination of 3 phase data has most
High overall accuracy, kappa coefficient and maize production person's precision.
On the other hand, when training sample ratio is 50%, the nicety of grading and DOY 231 of DOY 255 is quite, all obvious excellent
In 207 data of DOY.The data group conjunction of DOY 231 and 255 achieves highest overall accuracy and kappa coefficient, but number of phases when 3
Comparable level therewith is also reached according to combination, and there is slightly advanced maize production person's precision.Compared to training sample ratio
When example is 10%, the overall accuracy and kappa coefficient of 255 data of DOY 207 and DOY are all significantly improved, but with
The case where 255 DOY, is contrasted, and the corn classification results of DOY 207, which have no, to be significantly improved.The corn of 255 data of DOY
Producer's precision of producer's precision and user's precision and 231 data of DOY is all improved with the increase of training sample.
It is beneficial that this shows that sufficient training sample charts to corn.But when the validity of input feature vector itself is slightly poor, such as DOY
207 data, the increase of training sample will not then bring and be obviously improved.From the same available same conclusion of multi-temporal data,
Data combination for DOY 207 and 231, the increase of training sample have only brought about the raising of corn user's precision, and for 3
When be combined and combined with the data of DOY 231 and 255, corn discrimination and the overall performance of classification all increase.
In order to study vegetation growth middle and later periods, the validity that polarization SAR charts to corn, the present embodiment proposes a kind of base
Classification method in the vegetation distributed areas of the full polarimetric SAR of the multidate of tensor representation, passes through the MPCA based on tensor
Algorithm realizes feature extraction.This gives the general tensor representations of the coherence matrix of the full polarimetric SAR of multidate
Form.Corn and soybean polarization response are compared herein, and the multidate variation for further highlighting polarization information helps to be promoted
The otherness of corn and soybean.DT classification results show the processing proposed in this paper even if in the case where training sample is smaller
Process can also obtain higher corn accuracy of identification.In addition, heading stage data know corn compared to jointing stage and maturity period
Not advantageously, nevertheless, the combination of 3 phase data is still optimal selection.
It is apparent to those skilled in the art that for convenience and simplicity of description, the data of foregoing description
The electronic equipment that processing method is applied to, can be with reference to the corresponding description in before-mentioned products embodiment, and details are not described herein.
Above embodiments are only the exemplary embodiment of the application, are not used in limitation the application, the protection scope of the application
It is defined by the claims.Those skilled in the art can make respectively the application in the essence and protection scope of the application
Kind modification or equivalent replacement, this modification or equivalent replacement also should be regarded as falling within the scope of protection of this application.
Claims (10)
1. a kind of method that the full polarimetric SAR for multidate extracts feature, which is characterized in that the method includes walking as follows
It is rapid:
Obtain the full polarimetric SAR of several different phases of the same area comprising several vegetation;
The full polarimetric SAR of the multidate is split to obtain several targets;
Multistage tensor is created for each target, the first rank and second-order in the multistage tensor indicate the coherence matrix of the target
Or the row and column of covariance matrix, the third rank in the multistage tensor indicate different phases;
Dimensionality reduction is carried out using the multistage tensor of the main composition decomposition algorithm of polyteny to each target, to obtain each target
Feature vector.
2. the method according to claim 1, wherein the coherence matrix or covariance matrix include: same area
Strength information and/or phase information in the full polarimetric SAR of several different phases in domain.
3. the method according to claim 1, wherein the full polarimetric SAR to the multidate into
Before row segmentation, further includes:
The full polarimetric SAR of the multidate is registrated, and carries out the mean filter of default specification.
4. the method according to claim 1, wherein the full polarimetric SAR to the multidate carries out
Segmentation to obtain several targets the step of include:
Using the average coherence matrix or average covariance matrices of the full polarimetric SAR of the multidate, super-pixel point is calculated
The distance measure cut;
The pixel of the full polarimetric SAR of the multidate is aggregated into several targets using the distance measure of super-pixel segmentation;
Wherein, in the minimum pixel of gradient in the neighborhood for being centrally disposed in default specification of the target.
5. the method according to claim 1, wherein described utilize the main composition decomposition algorithm of polyteny to each mesh
Multistage tensor described in target carries out dimensionality reduction, to include: the step of obtaining each clarification of objective vector
Projection matrix is obtained according to the multistage tensor of each targetWherein Ln<In, LnIt is dimensionality reduction
Dimension afterwards, InFor original dimension;
Based on the iterative calculation of the projection matrix, dimensionality reduction is carried out to the multistage tensor of each target, to obtain each mesh
Target feature vector.
6. according to the method described in claim 5, it is characterized in that, the projection matrixWherein Ln<
In, LnIt is the dimension after dimensionality reduction, InFor original dimension, specific formula is as follows:
WhenWhen,
Wherein,
I indicates the i-th rank, and n indicates order;
M indicates that m-th of target tensor, M indicate the quantity of target tensor;
Q value range is [0,1], is user's custom parameter, is protected for controlling original tensor after projection stored
Ratio.
7. according to the method described in claim 6, it is characterized in that, the iterative calculation based on the projection matrix, to each
The multistage tensor of a target carries out dimensionality reduction, to include: the step of obtaining each clarification of objective vector
Obtain several sample objects with tensor representation;
Polyteny projection is carried out to the sample object to calculate, and obtains tensor Y;
Calculating is iterated with the tensor Y based on including the projection matrix;
Meet preset condition when total divergence of the tensor Y converges to or when the number of iterations reaches preset value, then obtain each mesh
Target feature vector.
It further include fourth order and the 5th rank 8. the method according to claim 1, wherein in the multistage tensor,
Indicate the length and width in the target window neighborhood.
9. a kind of classification method of the vegetation distributed areas of the full polarimetric SAR of multidate characterized by comprising
When extracting several differences of the same area comprising several vegetation using method described in any one of claim 1-8
Each clarification of objective vector of the full polarimetric SAR of phase;
The feature vector of at least one phase based on each target determines the vegetation classification belonging to it using classifier.
10. classification method according to claim 9, which is characterized in that the average coherence matrix of the target or average association
Variance matrix indicates are as follows:
Wherein,
Indicate average coherence matrix or average covariance matrices;
TiIndicate the coherence matrix of ith pixel;
K indicates that the target is made of k pixel.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810608408.5A CN108960295B (en) | 2018-06-13 | 2018-06-13 | Multi-temporal fully-polarized SAR image feature extraction method and classification method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810608408.5A CN108960295B (en) | 2018-06-13 | 2018-06-13 | Multi-temporal fully-polarized SAR image feature extraction method and classification method |
Publications (2)
Publication Number | Publication Date |
---|---|
CN108960295A true CN108960295A (en) | 2018-12-07 |
CN108960295B CN108960295B (en) | 2022-08-26 |
Family
ID=64488441
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810608408.5A Active CN108960295B (en) | 2018-06-13 | 2018-06-13 | Multi-temporal fully-polarized SAR image feature extraction method and classification method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108960295B (en) |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110399832A (en) * | 2019-07-25 | 2019-11-01 | 内蒙古工业大学 | TomoSAR vegetation pest and disease monitoring method and device based on coherence |
CN111523567A (en) * | 2020-03-31 | 2020-08-11 | 彩虹无人机科技有限公司 | Polarization feature extraction and classification method based on tensor local retention projection |
CN111967537A (en) * | 2020-04-13 | 2020-11-20 | 江西理工大学 | SAR target classification method based on two-way capsule network |
CN112257515A (en) * | 2020-09-29 | 2021-01-22 | 河南大学 | SAR image terrain classification method and device based on complex terrain |
CN112749705A (en) * | 2019-10-31 | 2021-05-04 | 深圳云天励飞技术有限公司 | Training model updating method and related equipment |
CN113642399A (en) * | 2021-07-12 | 2021-11-12 | 广东省国土资源测绘院 | Method for identifying abandoned land in paddy field based on SAR data |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104700110A (en) * | 2015-04-03 | 2015-06-10 | 电子科技大学 | Plant covering information extracting method based on perfect polarization SAR images |
CN104751183A (en) * | 2015-04-08 | 2015-07-01 | 西安电子科技大学 | Polarimetric SAR image classification method based on tensor MPCA |
US9239384B1 (en) * | 2014-10-21 | 2016-01-19 | Sandia Corporation | Terrain detection and classification using single polarization SAR |
CN105930846A (en) * | 2016-04-05 | 2016-09-07 | 西安电子科技大学 | Neighborhood information and SVGDL (support vector guide dictionary learning)-based polarimetric SAR image classification method |
CN107239757A (en) * | 2017-05-23 | 2017-10-10 | 西安电子科技大学 | A kind of polarization SAR silhouette target detection method based on depth ladder net |
CN107292920A (en) * | 2017-06-15 | 2017-10-24 | 武汉大学 | A kind of multidate full-polarization SAR remote sensing imagery change detection method of joint classification |
-
2018
- 2018-06-13 CN CN201810608408.5A patent/CN108960295B/en active Active
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US9239384B1 (en) * | 2014-10-21 | 2016-01-19 | Sandia Corporation | Terrain detection and classification using single polarization SAR |
CN104700110A (en) * | 2015-04-03 | 2015-06-10 | 电子科技大学 | Plant covering information extracting method based on perfect polarization SAR images |
CN104751183A (en) * | 2015-04-08 | 2015-07-01 | 西安电子科技大学 | Polarimetric SAR image classification method based on tensor MPCA |
CN105930846A (en) * | 2016-04-05 | 2016-09-07 | 西安电子科技大学 | Neighborhood information and SVGDL (support vector guide dictionary learning)-based polarimetric SAR image classification method |
CN107239757A (en) * | 2017-05-23 | 2017-10-10 | 西安电子科技大学 | A kind of polarization SAR silhouette target detection method based on depth ladder net |
CN107292920A (en) * | 2017-06-15 | 2017-10-24 | 武汉大学 | A kind of multidate full-polarization SAR remote sensing imagery change detection method of joint classification |
Non-Patent Citations (4)
Title |
---|
LU XU: "Comparative analysis of classification results between compact and fully polarimetric SAR images in random forest classifier", 《2017 IEEE INTERNATIONAL GEOSCIENCE AND REMOTE SENSING SYMPOSIUM (IGARSS)》 * |
LU XU等: "Corn mapping uisng multi-temporal fully and compact SAR data", 《 2017 SAR IN BIG DATA ERA: MODELS, METHODS AND APPLICATIONS (BIGSARDATA)》 * |
SERTAÇ ARISOY: "Superpixel-based classification of SAR images", 《 2015 23ND SIGNAL PROCESSING AND COMMUNICATIONS APPLICATIONS CONFERENCE (SIU)》 * |
蒋婷等: "极化SAR图像分割方法研究", 《无线互联科技》 * |
Cited By (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110399832A (en) * | 2019-07-25 | 2019-11-01 | 内蒙古工业大学 | TomoSAR vegetation pest and disease monitoring method and device based on coherence |
CN112749705A (en) * | 2019-10-31 | 2021-05-04 | 深圳云天励飞技术有限公司 | Training model updating method and related equipment |
CN112749705B (en) * | 2019-10-31 | 2024-06-11 | 深圳云天励飞技术有限公司 | Training model updating method and related equipment |
CN111523567A (en) * | 2020-03-31 | 2020-08-11 | 彩虹无人机科技有限公司 | Polarization feature extraction and classification method based on tensor local retention projection |
CN111523567B (en) * | 2020-03-31 | 2023-09-01 | 彩虹无人机科技有限公司 | Polarization feature extraction and classification method based on tensor local preservation projection |
CN111967537A (en) * | 2020-04-13 | 2020-11-20 | 江西理工大学 | SAR target classification method based on two-way capsule network |
CN111967537B (en) * | 2020-04-13 | 2022-04-08 | 江西理工大学 | SAR target classification method based on two-way capsule network |
CN112257515A (en) * | 2020-09-29 | 2021-01-22 | 河南大学 | SAR image terrain classification method and device based on complex terrain |
CN112257515B (en) * | 2020-09-29 | 2022-09-23 | 河南大学 | SAR image terrain classification method and device based on complex terrain |
CN113642399A (en) * | 2021-07-12 | 2021-11-12 | 广东省国土资源测绘院 | Method for identifying abandoned land in paddy field based on SAR data |
Also Published As
Publication number | Publication date |
---|---|
CN108960295B (en) | 2022-08-26 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108960295A (en) | A kind of full polarimetric SAR for multidate extracts the method for feature and the classification method of vegetation distributed areas | |
CN105678281B (en) | Remote sensing monitoring method for mulching film farmland based on spectrum and texture characteristics | |
Tian et al. | Aboveground mangrove biomass estimation in Beibu Gulf using machine learning and UAV remote sensing | |
Ok et al. | Evaluation of random forest method for agricultural crop classification | |
Yang et al. | Evaluating high resolution SPOT 5 satellite imagery for crop identification | |
CN112800973B (en) | Spartina alterniflora extraction method based on vegetation phenological feature decision | |
CN107609526A (en) | Rule-based fine dimension city impervious surface rapid extracting method | |
CN111242224B (en) | Multi-source remote sensing data classification method based on unmanned aerial vehicle extraction classification sample points | |
CN106023133B (en) | A kind of high-resolution remote sensing image Clean water withdraw method based on multiple features combining processing | |
CN103914678B (en) | Abandoned land remote sensing recognition method based on texture and vegetation indexes | |
CN110020635A (en) | Growing area crops sophisticated category method and system based on unmanned plane image and satellite image | |
CN111598045B (en) | Remote sensing farmland change detection method based on object spectrum and mixed spectrum | |
Zhang et al. | Crop classification and acreage estimation in North Korea using phenology features | |
CN109146948A (en) | The quantization of crop growing state phenotypic parameter and the correlation with yield analysis method of view-based access control model | |
CN105758806B (en) | Remote sensing monitoring method for mulching film farmland based on spectral characteristics | |
CN109977991A (en) | Forest resourceies acquisition method based on high definition satellite remote sensing | |
CN112686086A (en) | Crop classification method based on optical-SAR (synthetic aperture radar) cooperative response | |
CN107688776A (en) | A kind of urban water-body extracting method | |
Stroppiana et al. | Estimating crop density from multi-spectral uav imagery in maize crop | |
CN105678280B (en) | Mulching film mulching farmland remote sensing monitoring method based on textural features | |
Singh et al. | Advanced image processing approach for color-texture analysis of UAV imagery for weed detection in sugarcane crop | |
Xu et al. | Spatiotemporal distribution of cage and raft aquaculture in China's offshore waters using object-oriented random forest classifier | |
CN109636882A (en) | Black soil region digital soil drafting method based on digital elevation model and remote sensing image | |
CN113420780B (en) | Culture pond extraction method based on remote sensing spatial-temporal spectral feature fusion | |
CN113129320B (en) | Method and system for extracting culture pond boundary from aerial orthographic image |
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 | ||
TA01 | Transfer of patent application right | ||
TA01 | Transfer of patent application right |
Effective date of registration: 20220815 Address after: 100190 No. 19 West North Fourth Ring Road, Haidian District, Beijing Applicant after: Aerospace Information Research Institute,Chinese Academy of Sciences Address before: No. 9 Dengzhuang South Road, Haidian District, Beijing 100094 Applicant before: Institute of Remote Sensing and Digital Earth, Chinese Academy of Sciences |