CN103839075B - SAR image classification method based on united sparse representation - Google Patents
SAR image classification method based on united sparse representation Download PDFInfo
- Publication number
- CN103839075B CN103839075B CN201410062506.5A CN201410062506A CN103839075B CN 103839075 B CN103839075 B CN 103839075B CN 201410062506 A CN201410062506 A CN 201410062506A CN 103839075 B CN103839075 B CN 103839075B
- Authority
- CN
- China
- Prior art keywords
- sar image
- coefficients
- vector
- representing
- sar
- 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.)
- Expired - Fee Related
Links
- 238000000034 method Methods 0.000 title claims abstract description 33
- 239000013598 vector Substances 0.000 claims abstract description 66
- 238000012549 training Methods 0.000 claims description 37
- 238000012360 testing method Methods 0.000 claims description 14
- 239000011159 matrix material Substances 0.000 claims description 8
- 238000005457 optimization Methods 0.000 claims description 3
- 230000000694 effects Effects 0.000 abstract description 6
- 230000008569 process Effects 0.000 abstract description 3
- 238000002474 experimental method Methods 0.000 description 6
- 239000002131 composite material Substances 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 238000004519 manufacturing process Methods 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 2
- 238000004422 calculation algorithm Methods 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000007635 classification algorithm Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000009826 distribution Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000035515 penetration Effects 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Landscapes
- Image Analysis (AREA)
- Radar Systems Or Details Thereof (AREA)
Abstract
The invention discloses an SAR image classification method based on united sparse representation. The SAR image classification effect is improved based on an existing sparse representation method. According to the implementation process, the SAR image classification method comprises the steps that (1) SAR images to be trained are input, the features of the SAR images are extracted, and similar sets are classified; (2) united sparse representation is conducted on the similar set of each class of the SAR images, and a small dictionary and sparse coefficients of each similar set are obtained correspondingly; (4) SAR images to be tested are input, the features of the SAR images to be tested are extracted, feature vectors are projected on the small dictionary, and coefficients of the tested images are obtained; (5) the coefficients of the tested images and the sparse coefficients of all the trained images are matched, a set of most matched coefficients in the sparse coefficients are found, and the marked category of the set of most matched coefficients serves as the category of the SAR images to be tested. Compared with a traditional KNN and classic sparse representation classification method, the accuracy of even texture image and SAR image classification is greatly improved.
Description
Technical Field
The invention belongs to the technical field of image processing, and relates to an SAR image classification method based on joint sparse representation, which can be used for SAR image classification.
Background
The classification of the SAR image is a key step for realizing the automatic processing of the SAR image, is a premise for further interpretation of the SAR image, and is a typical example for extracting the front end part in an interpretation system independently as a specific application. The synthetic aperture radar SAR is a high-resolution remote sensing imaging radar, and has the characteristics of all-weather, all-time, multi-band and multi-polarization working modes, variable measurement visual angle, strong penetration capability, high resolution and the like. The system can observe the terrain and the landform more accurately and more in detail to obtain the land surface information, and can collect the underground information through a certain land surface and natural vegetation, so that the system is rapidly developed in recent years, is widely applied to various aspects of military affairs, civil affairs and national economy, and has great research value and very wide application prospect. Because the SAR image forming mechanism is different from that of a common visible light image, the classification algorithm suitable for the common optical image can not be directly applied to a radar image generally, the inherent multiplicative speckle noise of the SAR image increases the difficulty of image classification, and the effect of classifying the original information of the SAR image by adopting the traditional method is poor.
Wright et al constructs a sample dictionary-based Sparse Representation Classification (SRC) for face recognition, and defines a training sample as a fixed dictionary, and then finds out the Sparse Representation of the test sample according to the fixed dictionary, and then uses reconstruction criteria and the like to classify the classes. Due to the difference between the SAR image and the face image and the difference between the distribution rules, the effect of the SRC method directly applied to SAR image classification is not very prominent. The KNN (K-adjacent) algorithm is used as a theoretically mature supervised learning classification method, has the characteristics of simplicity, intuition, easy realization and low error rate, and has the basic idea that: and calculating the distance between the sample to be classified and each training sample according to the distance function, selecting k training samples with the minimum distance with the sample to be classified as k nearest neighbors, and judging the category of the sample to be classified according to the k nearest neighbors.
The SRC method is applied to SAR image classification after feature extraction, the obtained effect is good, but the class information of the training sample is not fully utilized in the training process, namely the process of calculating the sparse representation coefficient. The SAR images are classified by adopting a traditional KNN method, and the achieved effect is not ideal.
Disclosure of Invention
The invention aims to provide an SAR image classification method based on joint sparse representation aiming at the defects of the prior art, so that the classification of the SAR image is more accurate, and the classification result is further improved.
The invention provides an SAR image classification method based on joint sparse representation, which comprises the following steps:
the method comprises the following steps: performing 4-level non-downsampling wavelet transform on an input SAR image to be trained, extracting energy characteristic E of each sub-band, and extracting three characteristics of correlation Cor, local similarity Hom and entropy Ent of a gray level co-occurrence matrix to obtain a characteristic vector set E of all training images; defining the characteristic vector of each training sample as a column vector formed by combining the energy characteristic of 13 wavelet transform sub-bands and 3 gray level co-occurrence matrix characteristics, and defining the combination of all the training sample characteristic vectors as a characteristic vector set E;
step two: the set of feature vectors E for the training samples is normalized according to the following formula:
wherein u isqMean, σ, of line q representing the set of feature vectors EqRepresenting the variance of the q-th row of the characteristic vector set E, p representing the column of the characteristic vector set E, C representing the number of classes of input training samples, and Num being the number of each class of training samples;
step three: according to the category information of the training sample labels, calculating a similarity set S of each category of feature vectors according to the following formula:
wherein k represents the number of the similar sets of each class of feature vectors, and we take k as 3, SiRepresenting the direction of each kind of featuresIth similar set of quantities, xjRepresents a similar set SiCharacteristic vector of uiDenotes SiThe mean vector of (2);
step four: for each similarity set S of each class of feature vectorsiPerforming joint sparse representation, solving the following optimization problem to obtain a sparse coefficient AiAnd according to the sparse coefficient AiFinding out atoms in corresponding dictionary D to form small dictionary Pi:
Wherein,d is a fixed DCT dictionary, xjTo belong to the similar set SiIs determined by the feature vector of (a),ito minimize the error, 10 is taken-6;
Step five: inputting an SAR image y to be tested, extracting features according to the step one to obtain a feature vector t, normalizing the feature vector t according to the step two to obtain a feature vector t in a small dictionary PiThe test coefficient β is calculated according to the following formulai:
βi=(Pi TPi)-1Pi Tt
Step six, testing coefficients βiAnd the sparse coefficient AiAnd (3) carrying out nearest neighbor matching, and obtaining the category of the SAR image y to be tested according to the following formula:
idendity(y)=minjdj,j=1,2,...C.
wherein d isjRepresenting test coefficients and sparse coefficients AiC is the total number of classes of input training samples.
The SAR image classification method based on the joint sparse representation further comprises the following steps:
a: performing 4-level non-downsampling wavelet transform on an input SAR image to be trained, and calculating a wavelet energy characteristic e on each sub-band according to the following formula:
where M × N denotes the sub-band size, wm,nRepresents wavelet coefficients located at (m, n) in a subband;
b: for an input SAR image to be trained, extracting three characteristics of correlation Cor, local similarity Hom and entropy Ent of gray level co-occurrence moment characteristics according to the following formula:
where p (i, j) represents the number of occurrences of a pair of pixel gray levels (i, j) in an image separated by a distance s in the direction θ, where θ takes four discrete directions: 0 °, 45 °, 90 °, 135 °, s ═ 1, uhAnd σhRespectively representing the mean and variance, u, of the gray pair (i, j) in the horizontal directionvAnd σvRespectively representing the mean and variance of the gray level pair (i, j) in the vertical direction;
c: and defining the feature vector of each training sample as a column vector formed by combining the energy features of 13 wavelet transform sub-bands and the 3 gray level co-occurrence matrix features, and defining the combination of all the training sample feature vectors as a feature vector set E.
Compared with the prior art, the SAR image classification method based on joint sparse representation provided by the invention has the following advantages:
1. the invention achieves the purpose of original data dimension reduction by extracting the wavelet energy characteristic and the gray level co-occurrence matrix characteristic, and can be better applied to a combined sparse representation model.
2. The method extracts the public components of the similar set by performing joint sparse representation on the similar set, and takes the proportion of each training sample in the public components as the basis for classifying the test samples.
3. The joint sparse representation model classification method provided by the invention does not need dictionary updating learning, is quick and simple, and has high classification accuracy.
Drawings
FIG. 1 is a flow chart of the present invention;
FIG. 2 is a class 77 uniform texture image used by the present invention;
fig. 3 is a SAR terrain image used by the present invention.
Detailed Description
The invention is described in detail below with reference to the figures and examples.
Example 1:
the invention discloses an SAR image classification method based on joint sparse representation, which comprises the following steps of:
as shown in fig. 1, in the first step, a SAR image to be trained, which is labeled with a category, is input, and feature extraction is performed on the SAR image to obtain a feature vector set E of a training sample:
a: performing 4-level non-downsampling wavelet transform on an input SAR image to be trained, and calculating a wavelet energy characteristic e on each sub-band according to the following formula:
where M × N denotes the sub-band size, wm,nRepresents wavelet coefficients located at (m, n) in a subband;
b: for an input SAR image to be trained, extracting three characteristics of correlation Cor, local similarity Hom and entropy Ent of gray level co-occurrence moment characteristics according to the following formula:
where p (i, j) represents the number of occurrences of a pair of pixel gray levels (i, j) in an image separated by a distance s in the direction θ, where θ takes four discrete directions: 0 °, 45 °, 90 °, 135 °, s ═ 1, uhAnd σhRespectively representing the mean and variance, u, of the gray pair (i, j) in the horizontal directionvAnd σvRespectively representing the mean and variance of the gray level pair (i, j) in the vertical direction;
c: and defining the feature vector of each training sample as a column vector formed by combining the energy features of 13 wavelet transform sub-bands and the 3 gray level co-occurrence matrix features, and defining the combination of all the training sample feature vectors as a feature vector set E.
Step two: the set of feature vectors E for the training samples is normalized according to the following formula:
wherein u isqMean, σ, of line q representing the set of feature vectors EqRepresenting the variance of the q-th row of the characteristic vector set E, p representing the column of the characteristic vector set E, C representing the number of classes of input training samples, and Num being the number of each class of training samples;
step three: according to the category information of the training sample labels, calculating a similarity set S of each category of feature vectors according to the following formula:
wherein k represents the number of the similar sets of each class of feature vectors, and we take k as 3, SiThe ith similarity set, x, representing each class of feature vectorjRepresents a similar set SiCharacteristic vector of uiDenotes SiThe mean vector of (2);
step four: for each similarity set S of each class of feature vectorsiPerforming joint sparse representation, solving the following optimization problem to obtain a sparse coefficient AiAnd according to the sparse coefficient AiFinding out atoms in corresponding dictionary D to form small dictionary Pi:
Wherein,d is a fixed DCT wordDictionary, xjTo belong to the similar set SiIs determined by the feature vector of (a),ito minimize the error, 10 is taken-6;
Step five: inputting an SAR image y to be tested, extracting features according to the step (1) to obtain a feature vector t, normalizing the feature vector t according to the step (2), and placing the feature vector t in a small dictionary PiThe test coefficient β is calculated according to the following formulai:
βi=(Pi TPi)-1Pi Tt
Step six, testing coefficients βiAnd the sparse coefficient AiAnd (3) carrying out nearest neighbor matching, and obtaining the category of the SAR image y to be tested according to the following formula:
idendity(y)=minjdj,j=1,2,...C.
wherein d isjRepresenting test coefficients and sparse coefficients AiC is the total number of classes of input training samples.
Example 2:
experimental conditions and contents
The experimental conditions are as follows: the texture input images used in the experiments are 77 types of uniform texture images in a Brodatz database, the size of the images is 640 x 640, 25 sub-images of 128 x 128 are obtained by carrying out non-overlapping division on each large image, 13 images are randomly taken for training in each experiment, and 12 images are tested. The SAR images used in the experiment are six types of cities and towns, farmlands, mountains, bridges, water areas and sea ice, each type has 100 images of 128 × 128, the three types of experiments are divided, one type of experiment is divided into 30 images at random for training, and 70 types of experiments are carried out; 40 images of each type are randomly taken for training, and 60 tests are carried out; one type randomly takes 50 trains per class and 50 tests. The above experimental results are the average of 10 experimental results. The experimental contents are as follows: under the above experimental conditions, the method of the present invention is first applied to the uniform texture image, and 24 types, 41 types, and 77 types are respectively tested, where fig. 2 shows all 77 types of uniform texture images used in the test. The method of embodiment 1 of the present invention is implemented for six types of cities, towns, farmlands, mountains, bridges, water areas, and sea ices in the SAR surface feature image, wherein fig. 3 is a six types of SAR surface feature images used by the present invention.
Second, experimental results
The classification accuracy CR is taken as an evaluation index of a classification effect, and the calculation method comprises the following steps:
CR=CN/N
in the formula, CN is the number of images with correct classification, and N is the total number of images with correct classification. The results of the uniform texture image classification processed by the present invention are listed in table 1. The classification results of the SAR terrain image processed by the present invention are listed in table 2.
TABLE 1 Uniform texture image Classification results
SRC | Method for producing a composite material | KNN | |
24classes | 99.69 | 99.92 | 99.77 |
41classes | 97.88 | 99.27 | 98.85 |
77classes | 90.75 | 96.19 | 95.85 |
As can be seen from Table 1, the classification accuracy of the uniform texture image by the method of the invention is greatly improved compared with the SRC and KNN methods.
TABLE 2SAR image classification results
SRC | Method for producing a composite material | KNN | |
30traing70testing | 97.24 | 98.40 | 97.88 |
40training60testing | 98.33 | 98.69 | 97.31 |
50training50testing | 98.60 | 98.88 | 98.33 |
As can be seen from Table 2, the SAR terrain image classification accuracy rate is greatly improved compared with that of the SRC and KNN methods.
Claims (2)
1. A SAR image classification method based on joint sparse representation is characterized in that: the method comprises the following steps:
the method comprises the following steps: performing 4-level non-downsampling wavelet transform on an input SAR image to be trained, extracting energy characteristic E of each sub-band, and extracting three characteristics of correlation Cor, local similarity Hom and entropy Ent of a gray level co-occurrence matrix to obtain a characteristic vector set E of all training images; defining the characteristic vector of each training sample as a column vector formed by combining the energy characteristic of 13 wavelet transform sub-bands and 3 gray level co-occurrence matrix characteristics, and defining the combination of all the training sample characteristic vectors as a characteristic vector set E;
step two: the set of feature vectors E for the training samples is normalized according to the following formula:
wherein u isqMean, σ, of line q representing the set of feature vectors EqRepresenting the variance of the q-th row of the characteristic vector set E, p representing the column of the characteristic vector set E, C representing the number of classes of input training samples, and Num being the number of each class of training samples;
step three: according to the category information of the training sample labels, calculating a similarity set S of each category of feature vectors according to the following formula:
wherein k represents the number of the similar sets of each class of feature vectors, and we take k as 3, SiThe ith similarity set, x, representing each class of feature vectorjRepresents a similar set SiCharacteristic vector of uiDenotes SiThe mean vector of (2);
step four: for each similarity set S of each class of feature vectorsiPerforming joint sparse representation, solving the following optimization problem to obtain a sparse coefficient AiAnd according to the sparse coefficient AiFinding out atoms in corresponding dictionary D to form small dictionary Pi:
Wherein,d is a fixed DCT dictionary, xjTo belong to the similar set SiIs determined by the feature vector of (a),ito minimize the error, 10 is taken-6;
Step five: inputting an SAR image y to be tested, extracting features according to the step one to obtain a feature vector t, normalizing the feature vector t according to the step two to obtain a feature vector t in a small dictionary PiThe test coefficient β is calculated according to the following formulai:
βi=(Pi TPi)-1Pi Tt
Step six, testing coefficients βiAnd the sparse coefficient AiAnd (3) carrying out nearest neighbor matching, and obtaining the category of the SAR image y to be tested according to the following formula:
idendity(y)=minjdj,j=1,2,...C.
wherein d isjRepresenting test coefficients and sparse coefficients AiC is the total number of classes of input training samples.
2. The SAR image classification method based on joint sparse representation according to claim 1, characterized in that: the method for obtaining the feature vector set E in the first step is as follows:
a: performing 4-level non-downsampling wavelet transform on an input SAR image to be trained, and calculating a wavelet energy characteristic e on each sub-band according to the following formula:
where M × N denotes the sub-band size, wm,nRepresents wavelet coefficients located at (m, n) in a subband;
b: for an input SAR image to be trained, extracting three characteristics of correlation Cor, local similarity Hom and entropy Ent of gray level co-occurrence moment characteristics according to the following formula:
where p (i, j) represents the number of occurrences of a pair of pixel gray levels (i, j) in an image separated by a distance s in the direction θ, where θ takes four discrete directions: 0 °, 45 °, 90 °, 135 °, s ═ 1, ukAnd σkRespectively representing the mean and variance, u, of the gray pair (i, j) in the horizontal directionvAnd σvRespectively representing the mean and variance of the gray level pair (i, j) in the vertical direction;
c: and defining the feature vector of each training sample as a column vector formed by combining the energy features of 13 wavelet transform sub-bands and the 3 gray level co-occurrence matrix features, and defining the combination of all the training sample feature vectors as a feature vector set E.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201410062506.5A CN103839075B (en) | 2014-02-24 | 2014-02-24 | SAR image classification method based on united sparse representation |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201410062506.5A CN103839075B (en) | 2014-02-24 | 2014-02-24 | SAR image classification method based on united sparse representation |
Publications (2)
Publication Number | Publication Date |
---|---|
CN103839075A CN103839075A (en) | 2014-06-04 |
CN103839075B true CN103839075B (en) | 2017-04-12 |
Family
ID=50802553
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201410062506.5A Expired - Fee Related CN103839075B (en) | 2014-02-24 | 2014-02-24 | SAR image classification method based on united sparse representation |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN103839075B (en) |
Families Citing this family (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105095913B (en) * | 2015-08-11 | 2017-07-11 | 湖北大学 | The Classifying Method in Remote Sensing Image and system represented based on neighbour's canonical joint sparse |
CN106096652B (en) * | 2016-06-12 | 2019-05-24 | 西安电子科技大学 | Classification of Polarimetric SAR Image method based on sparse coding and small echo self-encoding encoder |
CN106203316A (en) * | 2016-07-04 | 2016-12-07 | 中国人民解放军电子工程学院 | The communication radiation source fine feature extracting method embedded based on manifold and system |
CN106778811B (en) * | 2016-11-21 | 2020-12-25 | 西安电子科技大学 | Image dictionary generation method, image processing method and device |
CN108875445B (en) * | 2017-05-08 | 2020-08-25 | 深圳荆虹科技有限公司 | Pedestrian re-identification method and device |
CN111797744B (en) * | 2020-06-28 | 2022-06-10 | 武汉大学 | Multimode remote sensing image matching method based on co-occurrence filtering algorithm |
CN117978271B (en) * | 2024-04-02 | 2024-06-21 | 浙江大学 | Optical fiber communication strong interference suppression method, system, equipment and storage medium |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101841932A (en) * | 2010-05-10 | 2010-09-22 | 南京邮电大学 | Distributed compression sensing method based on dynamic clustering in wireless sensor network |
US8081209B2 (en) * | 2008-06-26 | 2011-12-20 | Toyota Motor Engineering & Manufacturing North America, Inc. | Method and system of sparse code based object classification with sensor fusion |
CN103020647A (en) * | 2013-01-08 | 2013-04-03 | 西安电子科技大学 | Image classification method based on hierarchical SIFT (scale-invariant feature transform) features and sparse coding |
CN103093445A (en) * | 2013-01-17 | 2013-05-08 | 西安电子科技大学 | Unified feature space image super-resolution reconstruction method based on joint sparse constraint |
CN103164713A (en) * | 2011-12-12 | 2013-06-19 | 阿里巴巴集团控股有限公司 | Image classification method and device |
-
2014
- 2014-02-24 CN CN201410062506.5A patent/CN103839075B/en not_active Expired - Fee Related
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US8081209B2 (en) * | 2008-06-26 | 2011-12-20 | Toyota Motor Engineering & Manufacturing North America, Inc. | Method and system of sparse code based object classification with sensor fusion |
CN101841932A (en) * | 2010-05-10 | 2010-09-22 | 南京邮电大学 | Distributed compression sensing method based on dynamic clustering in wireless sensor network |
CN103164713A (en) * | 2011-12-12 | 2013-06-19 | 阿里巴巴集团控股有限公司 | Image classification method and device |
CN103020647A (en) * | 2013-01-08 | 2013-04-03 | 西安电子科技大学 | Image classification method based on hierarchical SIFT (scale-invariant feature transform) features and sparse coding |
CN103093445A (en) * | 2013-01-17 | 2013-05-08 | 西安电子科技大学 | Unified feature space image super-resolution reconstruction method based on joint sparse constraint |
Non-Patent Citations (3)
Title |
---|
Jumei Li 等.SAR Image Segmentation Based on Multiresolution GLCP in Overcomplete Brushlet Domain.《2006 CIE International Conference on Radar》.2006, * |
张娟 等.基于 Gabor 小波和稀疏表示的人脸表情识别.《人工智能及识别技术》.2012,第38卷(第6期),207-209、212. * |
王成语 等.基于双密度双树复小波变换多字典的人脸特征稀疏分类方法.《计算机应用》.2011,第31卷(第8期),2115-2118. * |
Also Published As
Publication number | Publication date |
---|---|
CN103839075A (en) | 2014-06-04 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN103839075B (en) | SAR image classification method based on united sparse representation | |
CN109934166A (en) | Unmanned plane image change detection method based on semantic segmentation and twin neural network | |
CN105809198B (en) | SAR image target recognition method based on depth confidence network | |
CN103679675B (en) | Remote sensing image fusion method oriented to water quality quantitative remote sensing application | |
CN102402685B (en) | Method for segmenting three Markov field SAR image based on Gabor characteristic | |
CN105335975B (en) | Polarization SAR image segmentation method based on low-rank decomposition and statistics with histogram | |
CN103208011B (en) | Based on average drifting and the hyperspectral image space-spectral domain classification method organizing sparse coding | |
CN104123555A (en) | Super-pixel polarimetric SAR land feature classification method based on sparse representation | |
CN107832797B (en) | Multispectral image classification method based on depth fusion residual error network | |
CN101520894A (en) | Method for extracting significant object based on region significance | |
CN102509120B (en) | Supervised image segmentation method for hyperspectral image based migration dictionary learning | |
CN105787937A (en) | OSM-based high-resolution remote sensing image road change detection method | |
CN111008644B (en) | Ecological change monitoring method based on local dynamic energy function FCN-CRF model | |
CN104331698A (en) | Remote sensing type urban image extracting method | |
CN103984746B (en) | Based on the SAR image recognition methodss that semisupervised classification and region distance are estimated | |
CN113312993B (en) | Remote sensing data land cover classification method based on PSPNet | |
CN103824088A (en) | SAR target variant recognition method based on multi-information joint dynamic sparse representation | |
CN102208017A (en) | Ship detection method based on high-resolution synthetic aperture radar images | |
CN107358214A (en) | Polarization SAR terrain classification method based on convolutional neural networks | |
Chen et al. | Agricultural remote sensing image cultivated land extraction technology based on deep learning | |
CN107392237A (en) | A kind of cross-domain ground cloud atlas sorting technique based on migration visual information | |
CN101894368A (en) | Method for dividing semi-monitoring SAR image water area based on geodesic distance | |
CN111325158B (en) | CNN and RFC-based integrated learning polarized SAR image classification method | |
CN113421198A (en) | Hyperspectral image denoising method based on subspace non-local low-rank tensor decomposition | |
CN102902982B (en) | Based on the SAR image texture classifying method of observation vector difference |
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 | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20170412 |
|
CF01 | Termination of patent right due to non-payment of annual fee |