CN105868793B - Classification of Polarimetric SAR Image method based on multiple dimensioned depth filter - Google Patents
Classification of Polarimetric SAR Image method based on multiple dimensioned depth filter Download PDFInfo
- Publication number
- CN105868793B CN105868793B CN201610237878.6A CN201610237878A CN105868793B CN 105868793 B CN105868793 B CN 105868793B CN 201610237878 A CN201610237878 A CN 201610237878A CN 105868793 B CN105868793 B CN 105868793B
- Authority
- CN
- China
- Prior art keywords
- eigenmatrix
- image
- layer
- pixel
- multiple dimensioned
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
- G06V10/443—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components by matching or filtering
- G06V10/449—Biologically inspired filters, e.g. difference of Gaussians [DoG] or Gabor filters
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Artificial Intelligence (AREA)
- Biodiversity & Conservation Biology (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Multimedia (AREA)
- Radar Systems Or Details Thereof (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of Classification of Polarimetric SAR Image methods based on multiple dimensioned depth filter, mainly solve the problems, such as that prior art nicety of grading is low, scheme is: inputting polarimetric SAR image to be sorted, acquire polarization scattering matrix S by polarization coherence matrix T;Pauli decomposition is carried out to polarization scattering matrix S, constitutes the eigenmatrix F based on pixel;F is normalized, and block is taken to each element in the eigenmatrix F1 after normalization, constitutes eigenmatrix F2 image block based;The eigenmatrix W1 of training dataset and the eigenmatrix W2 of test data set are obtained according to F2;Construct the disaggregated model based on multiple dimensioned depth filter;Disaggregated model is trained with the eigenmatrix W1 of training dataset;Classified using eigenmatrix W2 of the trained disaggregated model to test data set.Present invention introduces multiple dimensioned depth filters, improve the nicety of grading of polarimetric SAR image, can be used for target identification.
Description
Technical field
The invention belongs to technical field of image processing, and in particular to a kind of Classification of Polarimetric SAR Image method can be used for target
Identification.
Background technique
Polarization SAR is a kind of high-resolution active-mode active microwave remote sensing imaging radar, has round-the-clock, round-the-clock, divides
Resolution is high, can side view imaging the advantages that, the richer information of target can be obtained.The purpose of Classification of Polarimetric SAR Image is to utilize machine
Carry or borne polarization SAR sensor obtain polarization measurement data determine classification belonging to each pixel, agricultural, forestry,
Military affairs, geology, hydrology and ocean etc. have extensive research and application value.
The key of Classification of Polarimetric SAR Image is the target's feature-extraction to polarimetric SAR image, existing to be based on scattering properties
Polarimetric SAR image target's feature-extraction method, including Cloude decompose, Freeman decompose etc..
1997, Cloude et al. proposed Cloude decomposition, divides to H/ α plane, was characterized by H and α two
The characteristic value of polarization data turns to each pixel the classification of corresponding region.A defect existing for H/ alpha taxonomy is the division in region
Excessively dogmatic, when of a sort data distribution is in two classes or classes of boundary, classifier performance will be deteriorated, another deficiency
Place is, when several different atural objects coexist in the same region, it is impossible to effectively distinguish;
2004, Lee et al. proposed a kind of feature extracting method decomposed based on Freeman, and this method is able to maintain
All kinds of polarization scattering characteristics, but influence of the classification results vulnerable to Freeman decomposability, should to the polarization data of different-waveband
The universality of algorithm is poor.
These feature extracting methods are special due to multiple dimensioned, multi-direction, the more resolutions for not accounting for polarimetric SAR image
Property, thus the polarimetric SAR image of background complexity is difficult to obtain higher nicety of grading.
Summary of the invention
It is an object of the invention in view of the above-mentioned problems, proposing a kind of polarization SAR figure based on multiple dimensioned depth filter
As classification method, to improve nicety of grading.
Technological core of the invention is to introduce multiple dimensioned depth filter in convolutional neural networks to extract feature, technology
Scheme includes the following:
(1) polarimetric SAR image to be sorted is inputted, Polarization scattering is acquired by the polarization coherence matrix T of the polarimetric SAR image
Matrix S;
(2) Pauli decomposition is carried out to polarization scattering matrix S, obtains odd times scattering, even scattering, volume scattering coefficient, uses this
3 d image feature of 3 coefficients as polarimetric SAR image constitutes the eigenmatrix F based on pixel;
(3) element value in the eigenmatrix F based on pixel is normalized between [0,1], is denoted as F1;
(4) around 22 × 22 block is taken to represent original element value with each element in the eigenmatrix F1 after normalization,
Constitute eigenmatrix F2 image block based;
(5) the eigenmatrix W2 of the eigenmatrix W1 and test data set T' of training dataset D are constructed;
(6) disaggregated model based on multiple dimensioned depth filter is constructed:
(6a) selection one by input layer → convolutional layer → pond layer → convolutional layer → pond layer → full articulamentum → Quan Lian
8 layers of convolutional neural networks of layer → softmax classifier composition are connect, give the Feature Mapping figure of each layer, and determine convolutional layer
Filter size and random initializtion filter;
(6b) constructs multiple dimensioned depth filtering with the scaling filter in Gabor filter and non-down sampling contourlet transform
Device, and the filter of random initializtion in the convolutional layers of convolutional neural networks is replaced, it obtains based on multiple dimensioned depth filter
Disaggregated model are as follows: input layer → multiple dimensioned depth filter layer → pond layer → convolutional layer → pond layer → full articulamentum → Quan Lian
Connect this 8 layers of structures of layer → softmax classifier;
(7) disaggregated model is trained with training dataset, obtains trained model;
(8) classified using trained model to test data set, obtain test data and concentrate each pixel
Classification.
The present invention has the advantage that compared with prior art
1. the present invention can obtain spectral coverage information and space letter since Pixel-level feature is extended to image block characteristics simultaneously
Breath;
2. the present invention can obtain having multiple dimensioned, more due to introducing multiple dimensioned depth filter in convolutional neural networks
Direction, more resolution characteristics characteristics of image, the generalization ability of model is enhanced, so that still may be used in the case where training sample is less
To reach very high nicety of grading.
Detailed description of the invention
Fig. 1 is implementation flow chart of the invention;
Fig. 2 is to scheme in the present invention to the handmarking of image to be classified;
Fig. 3 is the classification results figure with the present invention to image to be classified.
Specific embodiment
Realization step of the invention and experiment effect are described in further detail below in conjunction with attached drawing:
It is referring to Fig.1, of the invention that the specific implementation steps are as follows:
Step 1, polarimetric SAR image to be sorted is inputted, polarization is acquired by the polarization coherence matrix T of the polarimetric SAR image
Collision matrix S.
Polarimetric SAR image to be sorted selects the ESAR sensor of Germany DLR near Munich, Germany
The L-band full polarimetric SAR data that the small town Oberpfaffenhofen obtains, image size are 1300 × 1200.
The polarization coherence matrix T of image to be classified known to (1a) obtains three on its diagonal line by polarization coherence matrix T
A element T11、T22、T33, that is, the member of element, polarization the 2nd row the 2nd of coherence matrix T column that the 1st row the 1st of coherence matrix T that polarizes arranges
The element of element, polarization the 3rd row the 3rd of coherence matrix T column;
(1b) is by T11、T22、T33, find out SHH、SVV、SHV:
Wherein, SHHFor horizontal emission and horizontal received scattering component, SVVFor the scattering point of Vertical Launch and vertical reception
Amount, SHVFor horizontal emission and the scattering component of vertical reception;
The S that (1c) is obtained according to step (1b)HH、SVV、SHV, form polarization scattering matrix S:
Step 2, Pauli decomposition is carried out to polarization scattering matrix S, obtains odd times scattering, even scattering, volume scattering coefficient,
It uses this 3 coefficients as the 3 d image feature of polarimetric SAR image, constitutes the eigenmatrix F based on pixel.
(2a) defines Pauli base { S1,S2,S3Formula it is as follows:
Wherein S1Indicate odd times scattering, S2Indicate even scattering, S3Indicate volume scattering;
(2b) is defined to obtain following equation by Pauli decomposition:
Wherein a is odd times scattering coefficient, and b is even scattering coefficient, and c is volume scattering coefficient;
(2c) solves formula<4>, the S that substitution formula<1>acquiresHH、SVV、SHV, obtain 3 scattering coefficients a, b, c:
(2d) defines the matrix F that size is M1 × M2 × 3, and by odd times scattering coefficient a, even scattering coefficient b, body
Scattering coefficient c is assigned to matrix F, obtains the eigenmatrix F based on pixel, and wherein M1 is the length of polarimetric SAR image to be sorted, M2
For the width of polarimetric SAR image to be sorted.
Step 3, the eigenmatrix F based on pixel is normalized.
Common method for normalizing has: characteristic line pantography, feature normalization and feature albefaction.
This example uses characteristic line pantography, i.e., first finds out the maximum value max (F) of the eigenmatrix F based on pixel;
Feature square by each element in the eigenmatrix F based on pixel divided by maximum value max (F), after being normalized again
Battle array F1.
Step 4, around 22 × 22 block is taken to represent original element with each element in the eigenmatrix F1 after normalization
Value, constitutes eigenmatrix F2 image block based.
Step 5, the eigenmatrix W2 of the eigenmatrix W1 and test data set T' of training dataset D are constructed.
Polarimetric SAR image atural object is divided into 3 classes by (5a), records the corresponding pixel of each classification in image to be classified
Position generates position A1, A2, A3 of 3 kinds of correspondences heterogeneously image vegetarian refreshments, wherein the corresponding 1st class atural object pixel of A1 to
The position classified in image, position of the corresponding 2nd class atural object pixel of A2 in image to be classified, A3 corresponding 3rd class ground image
Position of the vegetarian refreshments in image to be classified;
(5b) randomly selects 5% element from described A1, A2, A3, generates 3 kinds of corresponding inhomogeneity atural objects and is selected as training
Position B1, B2, B3 of the pixel of data set, wherein B1 is the pixel that training dataset is selected as in corresponding 1st class atural object
Position in image to be classified, B2 are to be selected as the pixel of training dataset in image to be classified in corresponding 2nd class atural object
In position, B3 is that position of the pixel of training dataset in image to be classified is selected as in corresponding 3rd class atural object, and will
Element in B1, B2, B3 merges position L1 of all pixels point of composition training dataset in image to be classified;
(5c) is selected as test data with 3 kinds of remaining 95% Element generation corresponding inhomogeneity atural objects in described A1, A2, A3
Position C1, C2, C3 of the pixel of collection, wherein C1 be selected as in corresponding 1st class atural object the pixel of test data set to
Position in classification image, C2 are to be selected as the pixel of test data set in image to be classified in corresponding 2nd class atural object
Position, C3 be position of the pixel of test data set in image to be classified is selected as in corresponding 3rd class atural object, and by C1,
Element in C2, C3 merges position L2 of all pixels point of composition test data set in image to be classified;
(5d) defines the eigenmatrix W1 of training dataset D, takes pair in eigenmatrix F2 image block based according to L1
The value on position is answered, and is assigned to the eigenmatrix W1 of training dataset D;
(5e) defines the eigenmatrix W2 of test data set T', takes pair in eigenmatrix F2 image block based according to L2
The value on position is answered, and is assigned to the eigenmatrix W2 of test data set T'.
Step 6, the disaggregated model based on multiple dimensioned depth filter is constructed.
(6a) selection one by input layer → convolutional layer → pond layer → convolutional layer → pond layer → full articulamentum → Quan Lian
8 layers of convolutional neural networks of layer → softmax classifier composition are connect, the Feature Mapping figure of each layer is given, determines the filter of convolutional layer
Wave device size and random initializtion filter;
(6b) constructs multiple dimensioned depth filtering with the scaling filter in Gabor filter and non-down sampling contourlet transform
Device, and the filter of random initializtion in the convolutional layers of convolutional neural networks is replaced, it obtains based on multiple dimensioned depth filter
Disaggregated model are as follows: input layer → multiple dimensioned depth filter layer → pond layer → convolutional layer → pond layer → full articulamentum → Quan Lian
This 8 layers of structures of layer → softmax classifier are connect, every layer of parameter is as follows:
For the 1st layer of input layer, it is 3 that Feature Mapping map number, which is arranged,;
Depth filter layer multiple dimensioned for the 2nd layer, setting Feature Mapping map number are 9;
For the 3rd layer of pond layer, down-sampling is set having a size of 2;
For the 4th layer of convolutional layer, it is 20 that Feature Mapping map number, which is arranged, and setting filter size is 4;
For the 5th layer of pond layer, down-sampling is set having a size of 2;
For the 6th layer of full articulamentum, it is 100 that Feature Mapping map number, which is arranged,;
For the 7th layer of full articulamentum, it is 64 that Feature Mapping map number, which is arranged,;
For the 8th layer of softmax classifier, it is 3 that Feature Mapping map number, which is arranged,.
Step 7, disaggregated model is trained with training dataset, obtains trained disaggregated model.
Using the eigenmatrix W1 of training dataset D as the input of disaggregated model, each pixel in training dataset D
Output of the classification as disaggregated model, by solving the error between above-mentioned classification and the correct classification of handmarking and to error
Backpropagation is carried out, carrys out the network parameter of Optimum Classification model, obtains trained disaggregated model, the correct category of handmarking
As shown in Figure 2.
Step 8, classified using trained disaggregated model to test data set.
Using the eigenmatrix W2 of test data set T' as the input of trained disaggregated model, trained disaggregated model
Output be the class categories for concentrating each pixel to be classified test data.
Effect of the invention can be further illustrated by following emulation experiment:
1. simulated conditions:
Hardware platform are as follows: Intel (R) Xeon (R) CPU E5-2630,2.40GHz*16 inside saves as 64G.
Software platform are as follows: Keras.
2. emulation content and result:
It is tested under above-mentioned simulated conditions with the method for the present invention, i.e., respectively from each classification of polarization SAR data
5% markd pixel is randomly selected as training sample, remaining markd pixel is obtained as test sample such as figure
3 classification results.
As can be seen from Figure 3: the region consistency of classification results is preferable, and the edge after different zones divide is also very clear
It is clear, and maintain detailed information.
Training sample is successively reduced again, training sample is made to account for 4%, 3%, the 2% of total sample number, by the present invention and convolution mind
Test data set nicety of grading through network compares, and the results are shown in Table 1:
Table 1
Training sample proportion | Convolutional neural networks | The present invention |
5% | 98.0384% | 98.1865% |
4% | 96.2356% | 97.6791% |
3% | 95.8376% | 97.3895% |
2% | 94.3743% | 96.1405% |
As seen from Table 1, when training sample accounts for 5%, 4%, 3%, the 2% of total sample number, test data set of the invention point
Class precision is above convolutional neural networks.
To sum up, the present invention effectively increases image spy by introducing multiple dimensioned depth filter in convolutional neural networks
The ability to express of sign enhances the generalization ability of model, so that still can achieve in the case where training sample is less very high
Nicety of grading.
Claims (6)
1. a kind of Classification of Polarimetric SAR Image method based on multiple dimensioned depth filter, comprising:
(1) polarimetric SAR image to be sorted is inputted, polarization scattering matrix is acquired by the polarization coherence matrix T of the polarimetric SAR image
S;
(2) Pauli decomposition is carried out to polarization scattering matrix S, odd times scattering, even scattering, volume scattering coefficient is obtained, with this 3
3 d image feature of the coefficient as polarimetric SAR image constitutes the eigenmatrix F based on pixel;
(3) element value in the eigenmatrix F based on pixel is normalized between [0,1], is denoted as F1;
(4) it takes around 22 × 22 block to represent original element value with each element in the eigenmatrix F1 after normalization, constitutes
Eigenmatrix F2 image block based;
(5) the eigenmatrix W2 of the eigenmatrix W1 and test data set T' of training dataset D are constructed:
Polarimetric SAR image atural object is divided into 3 classes by (5a), records position of the corresponding pixel of each classification in image to be classified
Set, generate position A1, A2, A3 of 3 kinds of correspondences heterogeneously image vegetarian refreshments, wherein the corresponding 1st class atural object pixel of A1 to point
Position in class image, position of the corresponding 2nd class atural object pixel of A2 in image to be classified, the corresponding 3rd class atural object pixel of A3
Position of the point in image to be classified;
(5b) randomly selects 5% element from described A1, A2, A3, generates 3 kinds of corresponding inhomogeneity atural objects and is selected as training data
Position B1, B2, B3 of the pixel of collection, wherein B1 be selected as in corresponding 1st class atural object the pixel of training dataset to
Position in classification image, B2 are to be selected as the pixel of training dataset in image to be classified in corresponding 2nd class atural object
Position, B3 be position of the pixel of training dataset in image to be classified is selected as in corresponding 3rd class atural object, and by B1,
Element in B2, B3 merges position L1 of all pixels point of composition training dataset in image to be classified;
(5c) is selected as test data set with 3 kinds of remaining 95% Element generation corresponding inhomogeneity atural objects in described A1, A2, A3
Position C1, C2, C3 of pixel, wherein C1 is to be selected as the pixel of test data set to be sorted in corresponding 1st class atural object
Position in image, C2 are to be selected as position of the pixel of test data set in image to be classified in corresponding 2nd class atural object,
C3 is that position of the pixel of test data set in image to be classified is selected as in corresponding 3rd class atural object, and by C1, C2, C3
In element merge position L2 of all pixels point in image to be classified of composition test data set;
(5d) defines the eigenmatrix W1 of training dataset D, takes corresponding position according to L1 in eigenmatrix F2 image block based
The value set, and it is assigned to the eigenmatrix W1 of training dataset D;
(5e) defines the eigenmatrix W2 of test data set T', takes corresponding position according to L2 in eigenmatrix F2 image block based
The value set, and it is assigned to the eigenmatrix W2 of test data set T';
(6) disaggregated model based on multiple dimensioned depth filter is constructed:
(6a) selection one by input layer → convolutional layer → pond layer → convolutional layer → pond layer → full articulamentum → full articulamentum
8 layers of convolutional neural networks of → softmax classifier composition give the Feature Mapping figure of each layer, and determine the filtering of convolutional layer
Device size and random initializtion filter;
(6b) constructs multiple dimensioned depth filter with the scaling filter in Gabor filter and non-down sampling contourlet transform,
And the filter of random initializtion in the convolutional layers of convolutional neural networks is replaced, obtain the classification based on multiple dimensioned depth filter
Model are as follows: input layer → multiple dimensioned depth filter layer → pond layer → convolutional layer → pond layer → full articulamentum → full articulamentum
This 8 layers of structures of → softmax classifier;
(7) disaggregated model is trained with training dataset, obtains trained model;
(8) classified using trained model to test data set, obtain the classification that test data concentrates each pixel.
2. the Classification of Polarimetric SAR Image method according to claim 1 based on multiple dimensioned depth filter, wherein step
(1) polarization scattering matrix S is acquired by the polarization coherence matrix T of polarimetric SAR image in, is carried out as follows:
(2a) known coherence matrix T that polarizes, can find out S according to formula<1>HH、SVV、SHV:
Wherein, T11、T22、T33For element on the diagonal line for the coherence matrix T that polarizes, SHHFor horizontal emission and horizontal received scattering
Component, SVVFor the scattering component of Vertical Launch and vertical reception, SHVFor horizontal emission and the scattering component of vertical reception;
The S that (2b) is obtained according to step (2a)HH、SVV、SHV, form polarization scattering matrix S:
3. the Classification of Polarimetric SAR Image method according to claim 2 based on multiple dimensioned depth filter, wherein step
(2) Pauli decomposition is carried out to polarization scattering matrix S in, steps are as follows:
(3a) defines Pauli base { S1,S2,S3, formula is as follows:
Wherein S1Indicate odd times scattering, S2Indicate even scattering, S3Indicate volume scattering;
(3b) is defined to obtain following equation by Pauli decomposition:
Wherein a is odd times scattering coefficient, and b is even scattering coefficient, and c is volume scattering coefficient;
(3c) solves formula<4>, the S that substitution formula<1>acquiresHH、SVV、SHV, obtain 3 scattering coefficients a, b, c:
4. the Classification of Polarimetric SAR Image method according to claim 1 based on multiple dimensioned depth filter, wherein step
(2) the eigenmatrix F based on pixel is constituted in, is the eigenmatrix F that first one size of definition is M1 × M2 × 3, then will be odd
Secondary scattering coefficient a, even scattering coefficient b, volume scattering coefficient c are assigned to eigenmatrix F, and wherein M1 is polarimetric SAR image to be sorted
Length, M2 be polarimetric SAR image to be sorted width.
5. the Classification of Polarimetric SAR Image method according to claim 1 based on multiple dimensioned depth filter, wherein step
(3) the eigenmatrix F based on pixel is normalized in, using characteristic line pantography, i.e., first finds out the spy based on pixel
Levy the maximum value max (F) of matrix F;Again by each element in the eigenmatrix F based on pixel divided by maximum value max
(F), the eigenmatrix F1 after being normalized.
6. the Classification of Polarimetric SAR Image method according to claim 1 based on multiple dimensioned depth filter, wherein step
Disaggregated model based on multiple dimensioned depth filter in (6b), parameter are as follows:
For the 1st layer of input layer, it is 3 that Feature Mapping map number, which is arranged,;
Depth filter layer multiple dimensioned for the 2nd layer, setting Feature Mapping map number are 9;
For the 3rd layer of pond layer, down-sampling is set having a size of 2;
For the 4th layer of convolutional layer, it is 20 that Feature Mapping map number, which is arranged, and setting filter size is 4;
For the 5th layer of pond layer, down-sampling is set having a size of 2;
For the 6th layer of full articulamentum, it is 100 that Feature Mapping map number, which is arranged,;
For the 7th layer of full articulamentum, it is 64 that Feature Mapping map number, which is arranged,;
For the 8th layer of softmax classifier, it is 3 that Feature Mapping map number, which is arranged,.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610237878.6A CN105868793B (en) | 2016-04-18 | 2016-04-18 | Classification of Polarimetric SAR Image method based on multiple dimensioned depth filter |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610237878.6A CN105868793B (en) | 2016-04-18 | 2016-04-18 | Classification of Polarimetric SAR Image method based on multiple dimensioned depth filter |
Publications (2)
Publication Number | Publication Date |
---|---|
CN105868793A CN105868793A (en) | 2016-08-17 |
CN105868793B true CN105868793B (en) | 2019-04-19 |
Family
ID=56633133
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201610237878.6A Active CN105868793B (en) | 2016-04-18 | 2016-04-18 | Classification of Polarimetric SAR Image method based on multiple dimensioned depth filter |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN105868793B (en) |
Families Citing this family (17)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106407986B (en) * | 2016-08-29 | 2019-07-19 | 电子科技大学 | A kind of identification method of image target of synthetic aperture radar based on depth model |
CN107154054B (en) * | 2017-05-22 | 2020-12-08 | 西安电子科技大学 | Polarization SAR image change detection method based on depth confidence network |
CN107133653B (en) * | 2017-05-23 | 2020-11-03 | 西安电子科技大学 | High-resolution SAR image classification method based on deep ladder network |
CN107239797A (en) * | 2017-05-23 | 2017-10-10 | 西安电子科技大学 | Polarization SAR terrain classification method based on full convolutional neural networks |
CN107239757B (en) * | 2017-05-23 | 2020-12-08 | 西安电子科技大学 | Polarized SAR image target detection method based on deep ladder network |
CN107169492B (en) * | 2017-05-26 | 2020-07-07 | 西安电子科技大学 | Polarized SAR target detection method based on FCN-CRF master-slave network |
CN107239799A (en) * | 2017-05-26 | 2017-10-10 | 西安电子科技大学 | Polarization SAR image classification method with depth residual error net is decomposed based on Pauli |
CN107292336A (en) * | 2017-06-12 | 2017-10-24 | 西安电子科技大学 | A kind of Classification of Polarimetric SAR Image method based on DCGAN |
CN107358202A (en) * | 2017-07-13 | 2017-11-17 | 西安电子科技大学 | Polarization SAR remote sensing imagery change detection method based on depth curve ripple heap stack network |
CN107358260B (en) * | 2017-07-13 | 2020-09-29 | 西安电子科技大学 | Multispectral image classification method based on surface wave CNN |
CN107368852A (en) * | 2017-07-13 | 2017-11-21 | 西安电子科技大学 | A kind of Classification of Polarimetric SAR Image method based on non-down sampling contourlet DCGAN |
CN107358205A (en) * | 2017-07-13 | 2017-11-17 | 西安电子科技大学 | A kind of polarization SAR remote sensing imagery change detection method based on NSCT DBN |
CN107944347A (en) * | 2017-11-03 | 2018-04-20 | 西安电子科技大学 | Polarization SAR object detection method based on multiple dimensioned FCN CRF |
CN107833216B (en) * | 2017-11-10 | 2019-10-25 | 西安电子科技大学 | Polarimetric SAR image change detecting method based on depth song wave difference DSN |
CN108416318A (en) * | 2018-03-22 | 2018-08-17 | 电子科技大学 | Diameter radar image target depth method of model identification based on data enhancing |
CN109063760B (en) * | 2018-07-22 | 2021-06-08 | 西安电子科技大学 | Polarization SAR classification method based on random forest multi-scale convolution model |
CN110516552B (en) * | 2019-07-29 | 2021-02-05 | 南京航空航天大学 | Multi-polarization radar image classification method and system based on time sequence curve |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104463193A (en) * | 2014-11-04 | 2015-03-25 | 西安电子科技大学 | Polarization SAR image classifying method based on depth sparsity ICA |
CN104700116A (en) * | 2015-03-13 | 2015-06-10 | 西安电子科技大学 | Polarized SAR (synthetic aperture radar) image object classifying method based on multi-quantum ridgelet representation |
CN105139028A (en) * | 2015-08-13 | 2015-12-09 | 西安电子科技大学 | SAR image classification method based on hierarchical sparse filtering convolutional neural network |
-
2016
- 2016-04-18 CN CN201610237878.6A patent/CN105868793B/en active Active
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104463193A (en) * | 2014-11-04 | 2015-03-25 | 西安电子科技大学 | Polarization SAR image classifying method based on depth sparsity ICA |
CN104700116A (en) * | 2015-03-13 | 2015-06-10 | 西安电子科技大学 | Polarized SAR (synthetic aperture radar) image object classifying method based on multi-quantum ridgelet representation |
CN105139028A (en) * | 2015-08-13 | 2015-12-09 | 西安电子科技大学 | SAR image classification method based on hierarchical sparse filtering convolutional neural network |
Non-Patent Citations (1)
Title |
---|
基于目标相干散射特性的极化SAR图像分解分类方法_;李晓玮等;《遥感技术与应用》;20070630;第22卷(第3期);全文 |
Also Published As
Publication number | Publication date |
---|---|
CN105868793A (en) | 2016-08-17 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN105868793B (en) | Classification of Polarimetric SAR Image method based on multiple dimensioned depth filter | |
CN107563428B (en) | Based on the Classification of Polarimetric SAR Image method for generating confrontation network | |
CN107239751B (en) | High-resolution SAR image classification method based on non-subsampled contourlet full convolution network | |
CN107944370B (en) | Classification of Polarimetric SAR Image method based on DCCGAN model | |
CN108460342A (en) | Hyperspectral image classification method based on convolution net and Recognition with Recurrent Neural Network | |
CN107358260B (en) | Multispectral image classification method based on surface wave CNN | |
CN105844279A (en) | Depth learning and SIFT feature-based SAR image change detection method | |
CN105718957A (en) | Polarized SAR image classification method based on nonsubsampled contourlet convolutional neural network | |
CN102982338B (en) | Classification of Polarimetric SAR Image method based on spectral clustering | |
CN105930772A (en) | City impervious surface extraction method based on fusion of SAR image and optical remote sensing image | |
CN108460408B (en) | Polarization SAR image classification method based on residual learning and conditional GAN | |
CN106886760B (en) | A kind of EO-1 hyperion Ship Detection combined based on empty spectrum information | |
CN104318246B (en) | Classification of Polarimetric SAR Image based on depth adaptive ridge ripple network | |
CN109389080A (en) | Hyperspectral image classification method based on semi-supervised WGAN-GP | |
CN110516728B (en) | Polarized SAR terrain classification method based on denoising convolutional neural network | |
CN107992891A (en) | Based on spectrum vector analysis multi-spectral remote sensing image change detecting method | |
CN104299232B (en) | SAR image segmentation method based on self-adaptive window directionlet domain and improved FCM | |
CN110533077A (en) | Form adaptive convolution deep neural network method for classification hyperspectral imagery | |
CN107844751A (en) | The sorting technique of guiding filtering length Memory Neural Networks high-spectrum remote sensing | |
CN108460391A (en) | Based on the unsupervised feature extracting method of high spectrum image for generating confrontation network | |
CN109493338A (en) | Hyperspectral image abnormal detection method based on combined extracting sky spectrum signature | |
CN104732244A (en) | Wavelet transform, multi-strategy PSO (particle swarm optimization) and SVM (support vector machine) integrated based remote sensing image classification method | |
CN104700116B (en) | The sorting technique of the Polarimetric SAR Image atural object represented based on multi-layer quantum ridge ripple | |
CN108764357A (en) | Polymerization residual error network hyperspectral image classification method based on compression-excitation | |
CN105138966B (en) | Classification of Polarimetric SAR Image method based on fast density peak value cluster |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |