CN109784413A - A kind of Classification of Polarimetric SAR Image method based on long short-term memory circulation nerve net - Google Patents
A kind of Classification of Polarimetric SAR Image method based on long short-term memory circulation nerve net Download PDFInfo
- Publication number
- CN109784413A CN109784413A CN201910065078.4A CN201910065078A CN109784413A CN 109784413 A CN109784413 A CN 109784413A CN 201910065078 A CN201910065078 A CN 201910065078A CN 109784413 A CN109784413 A CN 109784413A
- Authority
- CN
- China
- Prior art keywords
- sar image
- polarimetric sar
- term memory
- classification
- long short
- 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.)
- Pending
Links
Landscapes
- Radar Systems Or Details Thereof (AREA)
Abstract
The invention discloses a kind of Classification of Polarimetric SAR Image methods based on long short-term memory circulation nerve net, carry out exquisiteness Lee filtering processing to polarization SAR data；Super-pixel segmentation is carried out to filtered polarization coherence matrix；Multidimensional characteristic polarimetric SAR image is obtained using the spatial information of polarimetric SAR image；Obtain the sample data and test data of polarimetric SAR image to be sorted；Deep learning is carried out to long memory network in short-term using sample data；Using trained length, memory network classifies to test data in short-term；Tag along sort is obtained, and obtains color classification result figure.Present invention incorporates the spatial informations of polarimetric SAR image, obtain the polarimetric SAR image of multidimensional characteristic, and use the polarimetric SAR image data of multidimensional characteristic as the input of long memory network in short-term, the accuracy rate for effectively increasing Classification of Polarimetric SAR Image can be used for terrain classification and the identification of polarimetric SAR image.
Description
Technical field
The invention belongs to technical field of image processing, and in particular to a kind of polarization based on long short-term memory circulation nerve net
SAR image classification method can be applied to the terrain classification and target identification of polarimetric SAR image.
Background technique
SAR is synthetic aperture radar, is a kind of advanced remote-sensing imaging system, has round-the-clock, round-the-clock, high resolution
The advantages that.Polarization SAR is the synthetic aperture radar with multichannel, it is an important component of SAR, can compared to SAR
To obtain the richer information of target, the remote sensing images more common than other of the interpretation to polarimetric SAR image have more researching value,
In many fields, (such as earth resource generaI investigation, flood and waterlog monitoring, vegetation type is distinguished, sea ship detection and atural object characteristic divide
Analysis etc.) in, polarization SAR sorting technique, which suffers from, to be widely applied.Research for Classification of Polarimetric SAR Image is by beauty earliest
State scholar, which realizes atural object with polarization SAR, to classify, from this after, Many researchers have carried out relevant work, various classification calculations
Method also emerges.Mainly include following three classes method according to the difference of the feature for classification: one is be based on Polarization scattering
The decomposition method of mechanism, one is the classification method based on statistical nature, the method usually combines polarization decomposing theory to implement, and one
Kind is feature to be extracted based on machine Learning Theory, and then the method for classification is unfolded.
Classification method based on Polarization scattering mechanism.Such methods divide target according to the physical scatterers mechanism of polarization SAR
Solution is odd times scattering, rescattering, volume scattering and unclassified, and is classified using each scattering mechanism as characteristic of division, this
Significant for the algorithm classified according to target physical scattering mechanism, wherein classical way has Pauli to decompose, Cloude divides
Solution, the H/ α polarization SAR decomposition method that Freeman-Durden three-component decomposes and Cloude and Pottier is proposed.
Classification method based on statistical distribution.Such methods using polarization SAR classical approximation distribution multiple Gauss distribution and
Multiple Wishart distribution proposes the speckle noise for solving data for more maximum likelihood classifiers optionally and imitates to classification
The influence problem of fruit.Wishart classifier is added in Lee et al. on the basis of H/ α is decomposed, and proposes H/ α-Wishart polarization
SAR classification method improves the accuracy rate individually classified according to Several Kinds of Target Polar.
Classification method based on machine Learning Theory is relatively new method, with the development of neural network, Hen Duoshen
The algorithm of degree study is applied to Classification of Polarimetric SAR Image.The classification method based on vector did not accounted for polarization SAR data in the past
Spatial information, therefore complete image information can not be extracted, cause the accuracy rate of image classification low.The pixel sheet of polarization SAR
There is the data structure based on sequence in matter, and the Recognition with Recurrent Neural Network in deep learning is mainly for the treatment of sequence data, sheet
Invention proposes a kind of classification method based on long short-term memory Recognition with Recurrent Neural Network, and this method, which uses, utilizes long short-term memory net
Network characterizes the sequence characteristic of polarimetric SAR image pixel vectors, determines information category by network reasoning, reaches classification
Purpose.
Summary of the invention
In view of the above-mentioned deficiencies in the prior art, the technical problem to be solved by the present invention is that providing a kind of based on length
When memory circulation nerve net Classification of Polarimetric SAR Image method, for solves the classification method based on vector expression polarization SAR
The low technical problem of image classification accuracy rate caused by spatial information is lost when pixel.
The invention adopts the following technical scheme:
Polarization SAR data are carried out essence by a kind of Classification of Polarimetric SAR Image method based on long short-term memory circulation nerve net
Cause Lee filtering processing；Super-pixel segmentation is carried out to filtered polarization coherence matrix；Utilize the spatial information of polarimetric SAR image
Obtain multidimensional characteristic polarimetric SAR image；Obtain the sample data and test data of polarimetric SAR image to be sorted；Use sample number
Deep learning is carried out according to long memory network in short-term；Using trained length, memory network classifies to test data in short-term；
Tag along sort is obtained, and obtains color classification result figure.
Specifically, carrying out super-pixel segmentation to filtered polarization coherence matrix, the spatial information of polarimetric SAR image is calculated
Specific step is as follows:
S201, linear transformation is carried out to filtered polarization coherence matrix T, obtains polarization covariance matrix C, and according to pole
Change covariance matrix C, calculates polarimetric SAR image general power；
S202, Eigenvalues Decomposition is carried out to filtered polarization coherence matrix T, and calculate H/A/ α points using decomposition result
It solves parameter and polarimetric SAR image H/A/ α decomposes general power；
S203, the H/A/ α resolution parameter for calculating polarimetric SAR image and H/A/ α decompose general power.
Further, in step S201, polarimetric SAR image general power；
SPAN1=C11+C22+C33
Wherein, SPAN1 is polarimetric SAR image general power, and C11, C22, C33 respectively indicate the master of polarization covariance matrix C
Diagonal entry, C=A-1TA, A are transformation matrix, A-1Indicate the inverse matrix of transformation matrix A, specifically:
Further, in step S202, it is as follows that Eigenvalues Decomposition is carried out to filtered polarization coherence matrix T:
T=U Λ U*
Wherein, U is unitary matrice, and U* indicates the associate matrix of unitary matrice U, and Λ is diagonal matrix, λ1, λ2And λ3Respectively
Indicate three characteristic values of filtered polarization coherence matrix T.
Further, in step S203, scattering entropy H calculates as follows:
Negative entropy A calculates as follows:
Average scattering angleIt calculates as follows:
It is as follows that polarimetric SAR image H/A/ α decomposes general power SPAN2 calculating:
SPAN2=λ1+λ2+λ3
Wherein, H is scattering entropy, and A is negative entropy, and Σ indicates sum operation,λi、λjFor filtered polarization phase
The characteristic value of dry matrix T, i, j respectively correspond which characteristic value, cos-1Indicate anticosine operation, | | it is modulo operation, U1i
Indicate the i-th column element of unitary matrice U the first row.
Specifically, the spatial information using polarimetric SAR image is as follows the step of obtaining multidimensional characteristic polarimetric SAR image:
S301, vectorization is carried out to filtered polarization coherence matrix T, the vector for obtaining polarimetric SAR image indicates, refers to
Filtered polarization coherence matrix T is transformed to the vector I form of 9 dimensions；
S302, the vector of polarimetric SAR image is indicated and polarimetric SAR image general power, H/A/ α resolution parameter, polarization SAR
Image H/A/ α decompose general power, Freeman resolution parameter, polarimetric SAR image Freeman decompose general power and parametric texture into
Row fusion, obtains multidimensional characteristic polarimetric SAR image.
Further, 9 dimensional vector I are expressed as follows:
I=[T11,T22,T33,Re(T12),Im(T12),Re(T13),Im(T13),Re(T23),Im(T23)]
Wherein, Re () expression takes real part to operate, and Im () expression takes imaginary part to operate.
Specifically, carrying out deep learning to long short-term memory recirculating network using multidimensional characteristic polarimetric SAR image specifically:
A certain proportion of pixel is randomly selected from multidimensional characteristic polarimetric SAR image as tape label data, and utilizes the band mark chosen
Label data learn the parameter of long short-term memory recirculating network, obtain trained long short-term memory recirculating network.
Further, comprising the following steps:
S401, the value being initialized as the stochastic parameter of long short-term memory Recognition with Recurrent Neural Network between (0,1)；
S402,1~15% pixel is randomly selected from multidimensional characteristic polarimetric SAR image as tape label data；
S403, the data using tape label have carried out supervision to long short-term memory Recognition with Recurrent Neural Network parameter { i, f, o, c }
Training, obtains trained long short-term memory Recognition with Recurrent Neural Network.
Specifically, polarimetric SAR image pixel to be sorted is input in trained long short-term memory recirculating network, obtain
Class label belonging to each pixel, and use red, three colors of green and blue as three primary colours, it paints to class label,
Obtain final color classification result figure.
Compared with prior art, the present invention at least has the advantages that
A kind of Classification of Polarimetric SAR Image method based on long short-term memory circulation nerve net of the present invention, due in realization pair
During polarimetric SAR image is classified, the polarization coherence matrix of polarimetric SAR image is combined, H/A/ α is decomposed and polarization
The space characteristics of SAR data take full advantage of Polarization scattering information and spatial texture information, it is richer to be extracted polarimetric SAR image
Rich feature, effectively improves the accuracy rate of image classification.
Further, the pixel number magnitude of remote sensing images is not generally low, if handled one by one according to single pixel, that is by ten
Divide time-consuming, calculation amount is also very big, but if processed again after being divided into super-pixel block, problem above will be obtained significantly
Improve.The present invention is split polarimetric SAR image data using simple linear iterative clustering methods, is current fast speed
One of super-pixel algorithm, while also ensuring also there is preferable segmentation to non-uniform dielectric region.After super-pixel is handled, energy
More easily calculate the spatial information of polarimetric SAR image.
Further, it after reading polarization SAR image, first has to pre-process data, main includes more views, coherent spot
Filtering etc..When more view imagings, the data that different satellites obtains are regarded than different more.The data that the present invention uses belong to L-band phase
Array synthetic-aperture radar image is controlled, it is regard ratio as 7:1 more.After multiple look processing, the polarization coherence matrix of obtained polarization SAR data
Information includes: { T11, Re (T12), Im (T12), Re (T13), Im (T13), T22, Re (T23), Im (T23), T33 } (wherein Re
() expression takes real part to operate, and Im () expression takes imaginary part to operate), it is 9 dimension datas, therefore transforms it into the vector I shape of 9 dimensions
Formula, and merged with the parameter obtained in S2, obtain multidimensional characteristic polarimetric SAR image.
Further, the present invention studies the expression of polarimetric SAR image pixel by sequence visual angle, and long short-term memory net
Network has a very big advantage to the processing of sequence data analysis, the present invention using randomly select 1 in multidimensional characteristic polarimetric SAR image~
15% tape label pixel carries out Training to long short-term memory Recognition with Recurrent Neural Network parameter, obtains trained length
When remember Recognition with Recurrent Neural Network.
Further, polarimetric SAR image pixel to be sorted is input in trained long short-term memory recirculating network,
Class label belonging to each pixel is obtained, and uses red, three colors of green and blue as three primary colours, on class label
Color obtains nicety of grading convenient for being compared with the label figure of original graph.
In conclusion obtaining the polarization SAR figure of multidimensional characteristic present invention incorporates the spatial information of polarimetric SAR image
Picture, and the polarimetric SAR image data of multidimensional characteristic is used to effectively increase polarization SAR as the input of long memory network in short-term
The accuracy rate of image classification can be used for terrain classification and the identification of polarimetric SAR image.
Below by drawings and examples, technical scheme of the present invention will be described in further detail.
Detailed description of the invention
Fig. 1 is flow diagram of the invention；
Fig. 2 is simulation result diagram of the present invention in two class complexity atural object background polarization SAR image of a width, wherein (a) is
The simulation result diagram of the prior art, (b) is original tag figure, (c) is the simulation result diagram of the prior art, is (d) of the invention
Simulation result diagram.
Specific embodiment
The present invention provides a kind of Classification of Polarimetric SAR Image methods based on long short-term memory circulation nerve net, to polarization
SAR data carries out exquisiteness Lee filtering processing；Super-pixel segmentation is carried out to filtered polarization coherence matrix；Utilize polarization SAR figure
The spatial information of picture obtains multidimensional characteristic polarimetric SAR image；Obtain the sample data and test number of polarimetric SAR image to be sorted
According to；Deep learning is carried out to long memory network in short-term using sample data；Using trained length in short-term memory network to test
Data are classified；Tag along sort is obtained, and obtains color classification result figure.
Referring to Fig. 1, a kind of Classification of Polarimetric SAR Image method based on long short-term memory circulation nerve net of the present invention, benefit
With the spatial information of polarimetric SAR image, in conjunction with the scattering properties and texture features of polarimetric SAR image, to obtain multidimensional characteristic
Polarimetric SAR image realize polarization SAR figure and using the long memory network in short-term of the polarimetric SAR image pixel of multidimensional characteristic training
The classification of picture.Itself the specific implementation process is as follows:
S1, polarization SAR initial data is pre-processed, using exquisite Lee filtering method, to institute in polarization coherence matrix
There is element to carry out noise suppressed, obtains filtered polarization coherence matrix T；
S2, according to filtered polarization coherence matrix T, calculate the spatial information of polarimetric SAR image:
S201, linear transformation is carried out to filtered polarization coherence matrix T, obtains polarization covariance matrix C, and according to pole
Change covariance matrix C, calculates polarimetric SAR image general power；
SPAN1=C11+C22+C33 (1)
Wherein, SPAN1 is polarimetric SAR image general power, and C11, C22, C33 respectively indicate the master of polarization covariance matrix C
Diagonal entry, C=A-1TA, A are transformation matrix, A-1Indicate the inverse matrix of transformation matrix A, specifically:
S202, Eigenvalues Decomposition is carried out to filtered polarization coherence matrix T, and calculate H/A/ α points using decomposition result
It solves parameter and polarimetric SAR image H/A/ α decomposes general power；
T=U Λ U* (2)
Wherein, U is unitary matrice, and U* indicates the associate matrix of unitary matrice U, and Λ is diagonal matrix, can be indicated are as follows:
Wherein, λ1, λ2And λ3Respectively indicate three characteristic values of filtered polarization coherence matrix T；
S203, the H/A/ α resolution parameter for calculating polarimetric SAR image and H/A/ α decompose general power, and calculation formula is respectively as follows:
SPAN2=λ1+λ2+λ3 (6)
Wherein, Σ indicates sum operation,λi、λjFor the characteristic value of filtered polarization coherence matrix T, i, j
Respectively correspond which characteristic value, cos-1Indicate anticosine operation, | | it is modulo operation, U1iIndicate unitary matrice U the first row the
I column element, SPAN2 are that polarimetric SAR image H/A/ α decomposes general power.
S3, the spatial information using polarimetric SAR image obtain multidimensional characteristic polarimetric SAR image；
S301, vectorization is carried out to filtered polarization coherence matrix T, the vector for obtaining polarimetric SAR image indicates, refers to
Filtered polarization coherence matrix T is transformed to the vector I form of 9 dimensions；
I=[T11,T22,T33,Re(T12),Im(T12),Re(T13),Im(T13),Re(T23),Im(T23)] (7)
Wherein, Re () expression takes real part to operate, and Im () expression takes imaginary part to operate.
S302, the vector of polarimetric SAR image is indicated and polarimetric SAR image general power, H/A/ α resolution parameter, polarization SAR
Image H/A/ α decompose general power, Freeman resolution parameter, polarimetric SAR image Freeman decompose general power and parametric texture into
Row fusion, obtains multidimensional characteristic polarimetric SAR image；
S4, deep learning is carried out to long short-term memory recirculating network using multidimensional characteristic polarimetric SAR image: from multidimensional characteristic
A certain proportion of pixel is randomly selected in polarimetric SAR image as tape label data, and using the tape label data chosen to length
The parameter of short-term memory recirculating network is learnt, and trained long short-term memory recirculating network is obtained；
S401, the value being initialized as the stochastic parameter of long short-term memory Recognition with Recurrent Neural Network between (0,1)；
S402, the pixel of 1%-15% is randomly selected from multidimensional characteristic polarimetric SAR image as tape label data；
S403, the data using tape label have carried out supervision to long short-term memory Recognition with Recurrent Neural Network parameter { i, f, o, c }
Training, obtains trained long short-term memory Recognition with Recurrent Neural Network.
S5, polarimetric SAR image pixel to be sorted is input in trained long short-term memory recirculating network, is obtained each
Class label belonging to pixel, and use red, three colors of green and blue as three primary colours, it paints, obtains to class label
Final color classification result figure.
In order to make the object, technical scheme and advantages of the embodiment of the invention clearer, below in conjunction with the embodiment of the present invention
In attached drawing, technical scheme in the embodiment of the invention is clearly and completely described, it is clear that described embodiment is
A part of the embodiment of the present invention, instead of all the embodiments.The present invention being described and shown in usually here in attached drawing is real
The component for applying example can be arranged and be designed by a variety of different configurations.Therefore, below to the present invention provided in the accompanying drawings
The detailed description of embodiment be not intended to limit the range of claimed invention, but be merely representative of of the invention selected
Embodiment.Based on the embodiments of the present invention, those of ordinary skill in the art are obtained without creative efforts
The every other embodiment obtained, shall fall within the protection scope of the present invention.
Emulation content
Use the present invention and it is existing based on polarization coherence matrix for the polarization SAR sorting algorithm of input data, to full pole
Change the emulation that SAR image is classified to compare, result is as shown in Figure 2.The full-polarization SAR data used are RadarSAT-2 systems
What system obtained under C-band, image size is 260x259, and this area mainly includes land, two class atural object of river.
The simulation experiment result
Referring to Fig. 2, Fig. 2 (a) is the RGB composite diagram that Pauli is decomposed.
Fig. 2 (b) is original tag figure, and wherein brown part is land, and white is that part is river.
Fig. 2 (c) is the simulation result diagram of the prior art, it can be seen that there are many desultory points, mistake divides phenomenon tighter
Weight.
Simulation result diagram Fig. 2 (d) of the invention, it can be seen that wrong branch is less, and classification results edge is smoother.
Fig. 2 (d) efficiently solves mistake point phenomenon caused by spatial information deficiency, has obtained higher compared with Fig. 2 (c)
Classification accuracy.
From classification results as can be seen that being utilized more since this method has used the polarimetric SAR image data of multidimensional characteristic
Spatial information abundant effectively improves the accuracy rate of image classification.In the prior art, there are many inseparable pixels
Point, this method can be effectively improved.
The above content is merely illustrative of the invention's technical idea, and this does not limit the scope of protection of the present invention, all to press
According to technical idea proposed by the present invention, any changes made on the basis of the technical scheme each falls within claims of the present invention
Protection scope within.
Claims (10)
1. a kind of Classification of Polarimetric SAR Image method based on long short-term memory circulation nerve net, which is characterized in that polarization SAR
Data carry out exquisiteness Lee filtering processing；Super-pixel segmentation is carried out to filtered polarization coherence matrix；Utilize polarimetric SAR image
Spatial information obtain multidimensional characteristic polarimetric SAR image；Obtain the sample data and test data of polarimetric SAR image to be sorted；
Deep learning is carried out to long memory network in short-term using sample data；Using trained length in short-term memory network to test data
Classify；Tag along sort is obtained, and obtains color classification result figure.
2. the Classification of Polarimetric SAR Image method according to claim 1 based on long short-term memory circulation nerve net, feature
It is, super-pixel segmentation is carried out to filtered polarization coherence matrix, calculates the specific steps of the spatial information of polarimetric SAR image
It is as follows:
S201, linear transformation is carried out to filtered polarization coherence matrix T, obtains polarization covariance matrix C, and assist according to polarization
Variance matrix C calculates polarimetric SAR image general power；
S202, Eigenvalues Decomposition is carried out to filtered polarization coherence matrix T, and calculate H/A/ α using decomposition result and decompose ginseng
Several and polarimetric SAR image H/A/ α decomposes general power；
S203, the H/A/ α resolution parameter for calculating polarimetric SAR image and H/A/ α decompose general power.
3. the Classification of Polarimetric SAR Image method according to claim 2 based on long short-term memory circulation nerve net, feature
It is, in step S201, polarimetric SAR image general power；
SPAN1=C11+C22+C33
Wherein, SPAN1 is polarimetric SAR image general power, and the master that C11, C22, C33 respectively indicate polarization covariance matrix C is diagonal
Line element, C=A-1TA, A are transformation matrix, A-1Indicate the inverse matrix of transformation matrix A, specifically:
4. the Classification of Polarimetric SAR Image method according to claim 2 based on long short-term memory circulation nerve net, feature
It is, in step S202, it is as follows that Eigenvalues Decomposition is carried out to filtered polarization coherence matrix T:
T=U Λ U*
Wherein, U is unitary matrice, and U* indicates the associate matrix of unitary matrice U, and Λ is diagonal matrix, λ1, λ2And λ3It respectively indicates
Three characteristic values of filtered polarization coherence matrix T.
5. the Classification of Polarimetric SAR Image method according to claim 2 based on long short-term memory circulation nerve net, feature
It is, in step S203, scattering entropy H calculates as follows:
Negative entropy A calculates as follows:
Average scattering angleIt calculates as follows:
It is as follows that polarimetric SAR image H/A/ α decomposes general power SPAN2 calculating:
SPAN2=λ1+λ2+λ3
Wherein, H is scattering entropy, and A is negative entropy, and Σ indicates sum operation,λi、λjFor the filtered relevant square that polarizes
The characteristic value of battle array T, i, j respectively correspond which characteristic value, cos-1Indicate anticosine operation, | | it is modulo operation, U1iIt indicates
Unitary matrice U the i-th column element of the first row.
6. the Classification of Polarimetric SAR Image method according to claim 1 based on long short-term memory circulation nerve net, feature
The step of being, obtaining multidimensional characteristic polarimetric SAR image using the spatial information of polarimetric SAR image is as follows:
S301, vectorization is carried out to filtered polarization coherence matrix T, the vector for obtaining polarimetric SAR image indicates that referring to will filter
Polarization coherence matrix T after wave is transformed to the vector I form of 9 dimensions；
S302, the vector of polarimetric SAR image is indicated and polarimetric SAR image general power, H/A/ α resolution parameter, polarimetric SAR image
H/A/ α decomposes general power, Freeman resolution parameter, polarimetric SAR image Freeman decomposition general power and parametric texture and is melted
It closes, obtains multidimensional characteristic polarimetric SAR image.
7. the Classification of Polarimetric SAR Image method according to claim 6 based on long short-term memory circulation nerve net, feature
It is, 9 dimensional vector I are expressed as follows:
I=[T11,T22,T33,Re(T12),Im(T12),Re(T13),Im(T13),Re(T23),Im(T23)]
Wherein, Re () expression takes real part to operate, and Im () expression takes imaginary part to operate.
8. the Classification of Polarimetric SAR Image method according to claim 1 based on long short-term memory circulation nerve net, feature
It is, deep learning is carried out to long short-term memory recirculating network using multidimensional characteristic polarimetric SAR image specifically: from multidimensional characteristic
A certain proportion of pixel is randomly selected in polarimetric SAR image as tape label data, and using the tape label data chosen to length
The parameter of short-term memory recirculating network is learnt, and trained long short-term memory recirculating network is obtained.
9. the Classification of Polarimetric SAR Image method according to claim 8 based on long short-term memory circulation nerve net, feature
It is, comprising the following steps:
S401, the value being initialized as the stochastic parameter of long short-term memory Recognition with Recurrent Neural Network between (0,1)；
S402,1~15% pixel is randomly selected from multidimensional characteristic polarimetric SAR image as tape label data；
S403, the data using tape label have carried out supervision instruction to long short-term memory Recognition with Recurrent Neural Network parameter { i, f, o, c }
Practice, obtains trained long short-term memory Recognition with Recurrent Neural Network.
10. the Classification of Polarimetric SAR Image method according to claim 1 based on long short-term memory circulation nerve net, special
Sign is, polarimetric SAR image pixel to be sorted is input in trained long short-term memory recirculating network, each pixel is obtained
Affiliated class label, and use red, three colors of green and blue as three primary colours, it paints, obtains final to class label
Color classification result figure.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910065078.4A CN109784413A (en) | 2019-01-23 | 2019-01-23 | A kind of Classification of Polarimetric SAR Image method based on long short-term memory circulation nerve net |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910065078.4A CN109784413A (en) | 2019-01-23 | 2019-01-23 | A kind of Classification of Polarimetric SAR Image method based on long short-term memory circulation nerve net |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109784413A true CN109784413A (en) | 2019-05-21 |
Family
ID=66502156
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910065078.4A Pending CN109784413A (en) | 2019-01-23 | 2019-01-23 | A kind of Classification of Polarimetric SAR Image method based on long short-term memory circulation nerve net |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109784413A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111860356A (en) * | 2020-07-23 | 2020-10-30 | 中国电子科技集团公司第五十四研究所 | Polarization SAR image classification method based on nonlinear projection dictionary pair learning |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103903012A (en) * | 2014-04-09 | 2014-07-02 | 西安电子科技大学 | Polarimetric SAR data classifying method based on orientation object and support vector machine |
| WO2015041295A1 (en) * | 2013-09-18 | 2015-03-26 | 国立大学法人東京大学 | Terrain category classification method, terrain category classification program, and terrain category classification device |
| CN106778821A (en) * | 2016-11-25 | 2017-05-31 | 西安电子科技大学 | Classification of Polarimetric SAR Image method based on SLIC and improved CNN |
| CN107145860A (en) * | 2017-05-05 | 2017-09-08 | 西安电子科技大学 | Classification of Polarimetric SAR Image method based on spatial information and deep learning |
| CN107563422A (en) * | 2017-08-23 | 2018-01-09 | 西安电子科技大学 | A kind of polarization SAR sorting technique based on semi-supervised convolutional neural networks |
| CN108846426A (en) * | 2018-05-30 | 2018-11-20 | 西安电子科技大学 | Polarization SAR classification method based on the twin network of the two-way LSTM of depth |
-
2019
- 2019-01-23 CN CN201910065078.4A patent/CN109784413A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2015041295A1 (en) * | 2013-09-18 | 2015-03-26 | 国立大学法人東京大学 | Terrain category classification method, terrain category classification program, and terrain category classification device |
| CN103903012A (en) * | 2014-04-09 | 2014-07-02 | 西安电子科技大学 | Polarimetric SAR data classifying method based on orientation object and support vector machine |
| CN106778821A (en) * | 2016-11-25 | 2017-05-31 | 西安电子科技大学 | Classification of Polarimetric SAR Image method based on SLIC and improved CNN |
| CN107145860A (en) * | 2017-05-05 | 2017-09-08 | 西安电子科技大学 | Classification of Polarimetric SAR Image method based on spatial information and deep learning |
| CN107563422A (en) * | 2017-08-23 | 2018-01-09 | 西安电子科技大学 | A kind of polarization SAR sorting technique based on semi-supervised convolutional neural networks |
| CN108846426A (en) * | 2018-05-30 | 2018-11-20 | 西安电子科技大学 | Polarization SAR classification method based on the twin network of the two-way LSTM of depth |
Non-Patent Citations (1)
| Title |
|---|
| LEI WANG ETC.: "EXPLORING CONVOLUTIONAL LSTM FOR POLSAR IMAGE CLASSIFICATION", 《2018 IEEE INTERNATIONAL GEOSCIENCE AND REMOTE SENCING SYMPOSIUM》 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111860356A (en) * | 2020-07-23 | 2020-10-30 | 中国电子科技集团公司第五十四研究所 | Polarization SAR image classification method based on nonlinear projection dictionary pair learning |
| CN111860356B (en) * | 2020-07-23 | 2022-07-01 | 中国电子科技集团公司第五十四研究所 | Polarization SAR image classification method based on nonlinear projection dictionary pair learning |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Li et al. | Adaptive multiscale deep fusion residual network for remote sensing image classification | |
| He et al. | Texture classification of PolSAR data based on sparse coding of wavelet polarization textons | |
| Li et al. | Noise removal from hyperspectral image with joint spectral–spatial distributed sparse representation | |
| CN106296695B (en) | Adaptive threshold natural target image segmentation extraction algorithm based on conspicuousness | |
| CN104077599B (en) | Polarization SAR image classification method based on deep neural network | |
| Neagoe et al. | Concurrent self-organizing maps for pattern classification | |
| CN107527023B (en) | Polarized SAR image classification method based on superpixels and topic models | |
| CN104331707A (en) | Polarized SAR (synthetic aperture radar) image classification method based on depth PCA (principal component analysis) network and SVM (support vector machine) | |
| CN106897679A (en) | A kind of semantic change detecting method and system based on improvement fuzzy C-means clustering | |
| Ding et al. | An approach for visual attention based on biquaternion and its application for ship detection in multispectral imagery | |
| Wu et al. | Multiscale CNN with autoencoder regularization joint contextual attention network for SAR image classification | |
| Zhong et al. | Change detection based on pulse-coupled neural networks and the NMI feature for high spatial resolution remote sensing imagery | |
| CN105335975B (en) | Polarization SAR image segmentation method based on low-rank decomposition and statistics with histogram | |
| CN107330457B (en) | A kind of Classification of Polarimetric SAR Image method based on multi-feature fusion | |
| CN105894018B (en) | Classification of Polarimetric SAR Image method based on depth multi-instance learning | |
| Xie et al. | PolSAR image classification via D-KSVD and NSCT-domain features extraction | |
| CN111680579B (en) | Remote sensing image classification method for self-adaptive weight multi-view measurement learning | |
| CN109615008A (en) | Hyperspectral image classification method and system based on stack width learning | |
| CN106446935A (en) | Kernel sparse representation and spatial constraint-based polarimetric SAR image classification method | |
| Kowkabi et al. | Enhancing hyperspectral endmember extraction using clustering and oversegmentation-based preprocessing | |
| Gao et al. | Biologically inspired progressive enhancement target detection from heavy cluttered SAR images | |
| CN107341511A (en) | Classification of Polarimetric SAR Image method based on super-pixel Yu sparse self-encoding encoder | |
| CN104751173A (en) | Polarized SAR (Synthetic Aperture Radar) image classifying method based on cooperative representation and deep learning. | |
| CN109784413A (en) | A kind of Classification of Polarimetric SAR Image method based on long short-term memory circulation nerve net | |
| Guan et al. | Fishing vessel classification in SAR images using a novel deep learning model |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190521 |