CN105160351A - Semi-monitoring high-spectral classification method based on anchor point sparse graph - Google Patents
Semi-monitoring high-spectral classification method based on anchor point sparse graph Download PDFInfo
- Publication number
- CN105160351A CN105160351A CN201510493200.XA CN201510493200A CN105160351A CN 105160351 A CN105160351 A CN 105160351A CN 201510493200 A CN201510493200 A CN 201510493200A CN 105160351 A CN105160351 A CN 105160351A
- Authority
- CN
- China
- Prior art keywords
- mrow
- msub
- sample
- msup
- data set
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 35
- 238000012544 monitoring process Methods 0.000 title abstract 2
- 239000011159 matrix material Substances 0.000 claims abstract description 36
- 238000012549 training Methods 0.000 claims abstract description 25
- 230000003595 spectral effect Effects 0.000 claims description 17
- 238000001228 spectrum Methods 0.000 claims description 10
- 238000012360 testing method Methods 0.000 claims description 7
- 238000004364 calculation method Methods 0.000 claims description 5
- 238000007635 classification algorithm Methods 0.000 description 14
- 238000004088 simulation Methods 0.000 description 8
- 238000012706 support-vector machine Methods 0.000 description 8
- 235000008331 Pinus X rigitaeda Nutrition 0.000 description 4
- 235000011613 Pinus brutia Nutrition 0.000 description 4
- 241000018646 Pinus brutia Species 0.000 description 4
- 238000002474 experimental method Methods 0.000 description 4
- 230000000694 effects Effects 0.000 description 3
- 239000000463 material Substances 0.000 description 3
- 238000012545 processing Methods 0.000 description 3
- 238000004422 calculation algorithm Methods 0.000 description 2
- 230000007547 defect Effects 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 238000002372 labelling Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 239000013598 vector Substances 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 238000003064 k means clustering Methods 0.000 description 1
- 230000005855 radiation Effects 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
- 239000000126 substance Substances 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—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
- G06F18/2155—Generating training patterns; Bootstrap methods, e.g. bagging or boosting characterised by the incorporation of unlabelled data, e.g. multiple instance learning [MIL], semi-supervised techniques using expectation-maximisation [EM] or naïve labelling
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention discloses a semi-monitoring high-spectral classification method based on an anchor point sparse graph, and mainly solves the problems that the computational complexity is high and the storage amount is large during composition in the prior art. The method comprises the steps that 1) a training data set and a mark sample set are obtained from a high-spectral data set; 2) anchor points are selected randomly; 3) a sparse spatial-spectral relation matrix of sample points and the anchor points is established; 4) a Laplacian matrix of the graph is calculated; 5) labels of the anchor points are calculated; and 6) the types of unmarked sample points are obtained according to the obtained labels of the anchor points as well as the sparse spatial-spectral relation matrix. In the composition process, a small amount of anchor points are selected, the sparse spatial-spectral relation matrix is established according to spatial-spectral relation of the sample and anchor points, the complexity in composition of greatly is reduced, and the computation time is shortened. The method can be used for classified identification of high-spectral data.
Description
Technical Field
The invention belongs to the technical field of image processing, and further relates to a classification method which can be used for classifying hyperspectral images.
Background
In recent years, hyperspectral remote sensing has become the leading edge of the remote sensing direction. The hyperspectral image contains abundant spatial, spectral and radiation information, and the information characteristics make the hyperspectral image have huge application prospects in many fields, such as geological exploration, fine agriculture, ocean detection, military reconnaissance and the like.
Besides obtaining the spectral information for determining the properties of the substances or the ground objects, the hyperspectral remote sensing technology can also reveal the spatial position relationship among the ground objects, namely the integration of maps is realized. The hyperspectral remote sensing can simultaneously realize the acquisition of the spectral characteristics of the ground objects and the reservation of the relationship between the spectral characteristics and the surrounding ground objects. In addition, compared with common remote sensing, the hyperspectral remote sensing has more spectral bands with narrower wave bands, so that the material characteristics which can be shown only in a narrow range can be revealed, and more complete spectral data is formed. Due to the characteristics of the hyperspectral image and the abundant space spectrum information contained in the hyperspectral image, the hyperspectral image has unique advantages in ground feature classification and identification, and the accuracy of classification and identification can be obviously improved.
The classification of the ground objects with the hyperspectral data is always a key research direction in the research of the hyperspectral data. The hyperspectral data ground object classification refers to a data processing technology which gives the types of ground object targets needing to be classified according to collected information, and mark images of all pixels are obtained by giving a mark of the type to each pixel in a hyperspectral image. The labeled graph after classification and identification can reflect the space distribution condition of various ground objects, and is beneficial to comprehensively and clearly recognizing the researched area, so that the hyperspectral image has application value.
However, while the spectral resolution of the hyperspectral remote sensing data is improved, the data dimension and the data volume of the hyperspectral remote sensing data are also greatly increased, so that the calculation pressure during data processing is obviously increased, and the difficulty is brought to the practical application of hyperspectral classification and identification. Many conventional multispectral data classification algorithms are no longer suitable for hyperspectral data, and a classification algorithm suitable for hyperspectral data, which can reduce the amount of computation and improve the classification accuracy, needs to be provided according to the characteristics of the hyperspectral data. The existing hyperspectral data classification algorithm can classify the classification method into the following three classes according to whether a labeled sample participates in training of a classifier:
(1) unsupervised classification algorithm
The unsupervised classification method is to classify samples only according to whether spectral characteristics in hyperspectral data are similar or not under the condition that the labels of all samples are unknown, and belongs to a segmentation mode, such as K-means clustering and ISODATA. Unsupervised classification algorithms are also called clustering. The method has the advantages that the sample is not required to be marked, and manpower and material resources can be saved. However, without prior information, the classification accuracy of the unsupervised classification method is not high, and the specific class to which the ground object belongs cannot be obtained. In addition, the calculation process of the unsupervised classification method is time-consuming.
(2) Supervised classification algorithm
The supervised classification algorithm is firstly trained according to the labeled samples, a classifier is obtained through learning, and then the classifier is used for classifying the unlabeled samples. Support Vector Machines (SVM) are currently the most common method, which belongs to a classification method based on the minimization of structural risk. In the supervised classification algorithm, a labeled sample determines the accuracy of classification, but the labeling of the sample in the hyperspectral data requires a large amount of manpower and material resources, and the acquisition of a large amount of unlabeled samples is easier. When the number of the marked samples is small, the classification effect of the supervised classification algorithm is not ideal, and the popularization of the supervised classification algorithm in the field of hyperspectral data classification is restricted by the problem.
(3) Semi-supervised classification algorithm
The semi-supervised classification algorithm trains the classifier by using the labeled samples and the unlabeled samples at the same time, and the algorithm can fully utilize the information in the unlabeled samples to improve the classification precision and improve the generalization performance of the classifier. Among them, semi-supervised classification algorithms based on graphs have been receiving attention and developed in recent years due to the clear description and solution of flow pattern assumptions, such as LapSVM. However, the number of classification samples for hyperspectral data is large, which results in too large amount of calculation and storage required for composition, making the graph-based learning method difficult to use.
Disclosure of Invention
The invention aims to provide a semi-supervised hyperspectral classification method based on an anchor point sparse graph aiming at the defects in a graph-based semi-supervised classification algorithm.
The technical scheme for realizing the purpose of the invention is as follows: a sparse spatial spectrum relation matrix of the sample points and the anchor points is constructed by selecting a small number of anchor points, a graph Laplacian matrix is obtained according to the relation matrix, so that labels of the anchor points are obtained through calculation, and data classification is realized according to the relation between the labels of the sample points and the labels of the anchor points. The method comprises the following specific steps:
(1) and obtaining a training data set X and a marked sample set Y from the hyperspectral data set.
(2) From training data setsRandomly selecting m points as anchor points, and expressing asLabel f (v) of samplek) Can be obtained by the following formula:
wherein, f (v)k) Labels representing anchor points, ZikRepresenting anchor labels f (v)k) For sample label f (x)i) The contribution of (1) is weighted.
(3) Constructing sparse space spectrum relation matrix of sample points and anchor points
3a) For each sample point xiAccording to the Euclidean distance d (x) of the sample point and the space of all anchor pointsi,vk) Selecting x as the spatial distanceiThe nearest s anchor points;
3b) calculating a sparse spatial spectral relationship matrix according to the following formula:
wherein the kernel function is Gaussian kernel <math>
<mrow>
<mi>K</mi>
<mrow>
<mo>(</mo>
<msub>
<mi>x</mi>
<mi>i</mi>
</msub>
<mo>,</mo>
<msub>
<mi>v</mi>
<mi>k</mi>
</msub>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mi>exp</mi>
<mrow>
<mo>(</mo>
<mo>-</mo>
<mo>|</mo>
<mo>|</mo>
<msub>
<mi>x</mi>
<mi>i</mi>
</msub>
<mo>-</mo>
<msub>
<mi>v</mi>
<mi>k</mi>
</msub>
<mo>|</mo>
<msubsup>
<mo>|</mo>
<mn>2</mn>
<mn>2</mn>
</msubsup>
<mo>/</mo>
<mn>2</mn>
<msup>
<mi>σ</mi>
<mn>2</mn>
</msup>
<mo>)</mo>
</mrow>
<mo>,</mo>
<mo><</mo>
<mi>i</mi>
<mo>></mo>
<mo>⋐</mo>
<mo>[</mo>
<mn>1</mn>
<mo>:</mo>
<mi>m</mi>
<mo>]</mo>
</mrow>
</math> Represents a distance sample xiThe position index of the nearest s anchors.
(4) The graph laplacian matrix is calculated according to the following formula:
wherein,represents a diagonal matrix and
(5) the label of the anchor point is calculated according to the following formula:
wherein,and the submatrix corresponding to the l marked samples in the Z is represented, and gamma & gt 0 represents a regular parameter.
(6) Anchor obtained from solutionPoint label U*And a sparse spatial spectral relationship matrix Z, calculating the classes of the unlabeled sample points by the following formula
Wherein Z isiLine i of Z, xiAnd relation weight vectors with m anchor points.
Compared with the prior art, the invention has the following advantages:
the sparse spatial spectrum relation matrix of the sample points and the anchor points is constructed by selecting a small number of anchor points, the defects of overlarge calculated amount and storage amount during composition in the existing graph-based semi-supervised algorithm are overcome, the graph Laplace matrix is calculated according to the sparse spatial spectrum relation matrix, so that the labels of the anchor points are calculated, and then the labels of the sample points are obtained according to the relation between the labels of the sample points and the labels of the anchor points, so that the classification of data is realized.
Drawings
FIG. 1 is a flow chart of the present invention;
FIG. 2 is a graph of experimental hyperspectral data IndianPines and its true tags for use in the simulation of the present invention;
fig. 3 is a comparison graph of classification results when 10 samples are labeled per class.
Detailed Description
The present invention is described in further detail with reference to fig. 1.
Step 1: and obtaining a training data set X and a marked sample set Y from the hyperspectral data set.
1a) In the hyperspectral data set, 40% of data are randomly selected to form a training sample data set X belonging to Rd×nAnd taking the residual 60 percent of data as a test sample data set T epsilon Rd×tWhere d represents the dimensions of the training set samples and the test set samples, RNRepresenting an N-dimensional real number space, N representing the total number of training set samples, and t representing the total number of test set samples; in the IndianPines dataset of the embodiment of the present invention, the sample dimension d is 200, and the total number n of training set samples is 4146;
1b) in a training data set X, each class randomly selects k samples to form a marked sample set Y e Rl×cWhere c is the number of classes, l ═ k × c is the total number of marked samples, when sample x isiLabel y ofiWhen j is equal to Yij1 is ═ 1; in the InianPines dataset, which is an example of an implementation of the present invention, c is 16 and k is {3,5,8,10 }.
Step 2: from training data setsRandomly selecting m points as anchor points, and expressing asLabel f (v) of samplek) Can be obtained by the following formula:
wherein, f (v)k) Labels representing anchor points, ZikRepresenting anchor labels f (v)k) For sample label f (x)i) The contribution of (1) is weighted. In the IndianPines dataset of the embodiment of the present invention, the number m of anchor points is 1000.
And step 3: constructing sparse space spectrum relation matrix of sample points and anchor points
3a) For each sample point xiAccording to the Euclidean distance d (x) of the sample point and the space of all anchor pointsi,vk) Selecting x as the spatial distanceiThe nearest s anchor points. In the IndianPines dataset, an example of the present invention, let s be 3.
3b) Calculating a sparse spatial spectral relationship matrix according to the following formula:
wherein the kernel function is Gaussian kernel <math>
<mrow>
<mi>K</mi>
<mrow>
<mo>(</mo>
<msub>
<mi>x</mi>
<mi>i</mi>
</msub>
<mo>,</mo>
<msub>
<mi>v</mi>
<mi>k</mi>
</msub>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mi>exp</mi>
<mrow>
<mo>(</mo>
<mo>-</mo>
<mo>|</mo>
<mo>|</mo>
<msub>
<mi>x</mi>
<mi>i</mi>
</msub>
<mo>-</mo>
<msub>
<mi>v</mi>
<mi>k</mi>
</msub>
<mo>|</mo>
<msubsup>
<mo>|</mo>
<mn>2</mn>
<mn>2</mn>
</msubsup>
<mo>/</mo>
<mn>2</mn>
<msup>
<mi>σ</mi>
<mn>2</mn>
</msup>
<mo>)</mo>
</mrow>
<mo>,</mo>
<mo><</mo>
<mi>i</mi>
<mo>></mo>
<mo>⋐</mo>
<mo>[</mo>
<mn>1</mn>
<mo>:</mo>
<mi>m</mi>
<mo>]</mo>
</mrow>
</math> Represents a distance sample xiThe position index of the nearest s anchors.
And 4, step 4: the graph laplacian matrix is calculated according to the following formula:
wherein,represents a diagonal matrix and
and 5: the label of the anchor point is calculated according to the following formula:
wherein,and the submatrix corresponding to the l marked samples in the Z is represented, and gamma & gt 0 represents a regular parameter. In the IndianPines dataset, which is an example of the present invention, let γ equal to 0.01.
Step 6: anchor point label U obtained according to solution*And a sparse spatial spectral relationship matrix Z, calculating the classes of the unlabeled sample points by the following formula
Wherein Z isiLine i of Z, xiAnd relation weight vectors with m anchor points.
The effects of the present invention can be further illustrated by the following simulation experiments.
1. And (5) simulating experimental conditions.
In the experiment, an IndianPines data set is used as experiment data, software MATLABR2008a is used as a simulation tool, and a computer is configured to be IntelCorei 5/2.4G/4G.
IndianPinnes hyperspectral data 92AV 3C: the scene is an Indian pines test site of northwest Indiana obtained by the AVIRIS sensor in 6.1992, the data size is 145 x 145, each pixel has 220 wave bands, 20 wave bands containing noise are removed, only the remaining 200 wave bands are reserved, the data totally comprise 16 types of ground objects, the Indian pines hyperspectral data is shown in figure 2(a), and a real mark map of the Indian pines hyperspectral data is shown in figure 2 (b).
2. And (5) simulating the experimental content.
Simulation 1, performing simulation experiments under different numbers of marked samples on the Indian pines hyperspectral data given in FIG. 2(a), and comparing the method of the invention with the following three classification methods under the real marks given in FIG. 2 (b): 1) a Support Vector Machine (SVM); 2) laplacian support vector machine (LapSVM); 3) the null spectrum incorporates a Laplacian support vector machine (SS-LapSVM).
In the experiment, the graph regularization parameter γ of the invention is 0.01, the anchor point number m is 1000, and the nearest anchor point number s is selected to be 3. In Table 1, OA represents the overall accuracy, AA represents the average accuracy, and Kappa represents the Kappa coefficient.
Table 1 shows the average value of 20 simulation results obtained from experiments when the number of labeled samples is {3,5,8 and 10}, respectively.
Table 1: the comparison result of the invention and other methods under different marked sample numbers
As can be seen from Table 1, the classification accuracy is highest among the four methods listed in the table when the number of labeled samples per class is {3,5,8,10 }. Compared with the SVM method with supervision, the classification effect of the other three methods is good, and the use of the unmarked samples can improve the classification precision. The spatial information has great influence on hyperspectral classification, so that the classification precision of the hyperspectral image classification method and the SS-LapSVM method added with the spatial information is greatly improved. More importantly, in the composition process, both LapSVM and SS-LapSVM need to construct images with the size of n multiplied by n, but only the images with the size of n multiplied by m need to be constructed in the invention, the number m of selected anchor points is less than n, and only s nearest anchor points are selected through the space distance, so that the images are more sparse, the required calculated amount is greatly reduced, and the calculating time is shortened.
And 2, performing simulation, namely classifying the hyperspectral data by using four methods in the simulation 1 when the number of the labeled samples of each type is 10. The labeling results of the classification are shown in fig. 3. FIGS. 3(a) - (d) are labeled graphs of the results of SVM, LapSVM, SS-LapSVM, and the method of the present invention, respectively.
As can be seen from FIG. 3, compared with other methods, the method has the advantages of better consistency of the classified space structure and high classification precision, and the effectiveness of the method is proved.
Claims (2)
1. A semi-supervised hyperspectral classification method based on anchor point sparse graphs comprises the following steps:
(1) obtaining a training sample data set X and a marked sample category matrix Y from a hyperspectral data set;
(2) from training sample data setRandomly selecting m samples as anchor points, and expressing the samples as an anchor point setWherein x isiFor the ith training sample in the training sample data set X, n is the number of training samples, vkThe k-th anchor point in the anchor point set V is defined, and m is the number of the anchor points;
for any sample X in training sample data set XiSample xiLabel f (x)i) Obtained by the following formula:
wherein, f (x)i)∈R1×cAnd f (v)k)∈R1×cRespectively represent samples xiAnd anchor point vkC represents the number of categories, ZikRepresenting anchor labels f (v)k) For sample label f (x)i) Contribution proportion weight of (1);
(3) constructing sparse space spectrum relation matrix of sample points and anchor points
3a) For each training sample xiCalculating the space Euclidean distance between the sample point and all anchor points in the anchor point set VWherein x isijRepresents a sample xiThe jth feature of (v)kjRepresenting anchor points vkD denotes a sample xiAnd anchor point vkAccording to the calculation result, the feature dimension of the sample x is selectediThe nearest s anchor points;
3b) the elements in the sparse spatial spectral relationship matrix, 1, are calculated according to:
wherein a Gaussian kernel function is selectedCalculating a sample xiAnd anchor point vkσ is a parameter of the gaussian kernel function,<i>represents a distance sample xiThe set of position indices of the nearest s anchors,<i>has a value range of [1: m]，Ζ∈Rn×mRepresenting sparse space spectrum relation matrix, n and m respectively representing the number of training samples and the number of anchor points, ZikRepresenting sparse spatial spectral relationshipsA matrix, z, the ith row and the jth column;
(4) the graph laplacian matrix is calculated according to the following formula:
wherein, Z represents a sparse spatial spectral relationship matrix,represents a diagonal matrix andrepresenting a graph laplacian matrix;
(5) the label of the anchor point is calculated according to the following formula:
wherein, U*Corresponding labels, U, representing all anchors*＝[f(v1),f(v2),…,f(vm)]T∈Rm×cThe kth row represents the label of the kth anchor point, c represents the number of classes, Y represents the labeled sample class matrix, Y ∈ Rl×cAnd l represents the total number of labeled samples,representing a submatrix formed by rows corresponding to the l marked samples in the sparse space spectrum relationship matrix Z, wherein gamma is more than 0 and represents a regular parameter;
(6) according to the obtained anchor point label U*And a sparse spatial spectral relationship matrix Z, calculating the category of the unlabeled sample by the following formula:
wherein,class label, Z, representing the ith unlabeled sampleiRow i, u representing sparse spatial spectral relationship matrix ZjRepresenting anchor labels U*Column j.
2. The semi-supervised hyperspectral classification method based on the anchor point sparse graph according to claim 1, wherein in the step (1), a training sample data set X and a labeled sample category matrix Y are obtained from a hyperspectral data set, and the method comprises the following steps:
1a) in the hyperspectral data set, 40% of data are randomly selected to form a training sample data set X, and X belongs to Rd×nAnd taking the rest 60 percent of data as a test sample data set T, and the T belongs to Rd×tWhere d represents the feature dimensions of the samples in the training sample data set X and the test sample data set T, RNRepresenting an N-dimensional real number space, wherein N represents the total number of samples in a training sample data set X, and T represents the total number of samples in a test sample data set T;
1b) in a training sample data set X, selecting l training samples to obtain a marked sample data setWherein x isiFor marking a sample data set XlThe ith marking sample in the set, l represents the number of the marking samples, and the marking sample data set XlThe samples in (1) are marked to obtain the category of each marked sample, and a category marking matrix Y belongs to Rl×cWhere c denotes the number of classes and l denotes the total number of labeled samples, when labeled sample xiBelonging to class j, i.e. marked sample xiLabel y ofiWhen j, the element marking the ith row and the jth column in the sample class matrix Y is YijThe other elements of the row are all zero, 1.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510493200.XA CN105160351B (en) | 2015-08-12 | 2015-08-12 | Semi-supervised hyperspectral classification method based on anchor point sparse graph |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510493200.XA CN105160351B (en) | 2015-08-12 | 2015-08-12 | Semi-supervised hyperspectral classification method based on anchor point sparse graph |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105160351A true CN105160351A (en) | 2015-12-16 |
| CN105160351B CN105160351B (en) | 2018-11-23 |
Family
ID=54801203
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510493200.XA Active CN105160351B (en) | 2015-08-12 | 2015-08-12 | Semi-supervised hyperspectral classification method based on anchor point sparse graph |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105160351B (en) |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106127225A (en) * | 2016-06-13 | 2016-11-16 | 西安电子科技大学 | Semi-supervised hyperspectral image classification method based on rarefaction representation |
| CN109858531A (en) * | 2019-01-14 | 2019-06-07 | 西北工业大学 | A kind of high-spectrum remote sensing quick clustering algorithm based on figure |
| CN110874465A (en) * | 2018-08-31 | 2020-03-10 | 浙江大学 | Mobile equipment entity identification method and device based on semi-supervised learning algorithm |
| CN111488923A (en) * | 2020-04-03 | 2020-08-04 | 陕西师范大学 | Enhanced anchor point image semi-supervised classification method |
| CN111753904A (en) * | 2020-06-24 | 2020-10-09 | 广东工业大学 | Rapid hyperspectral image clustering method, device, equipment and medium |
| CN111783865A (en) * | 2020-06-23 | 2020-10-16 | 西北工业大学 | Hyperspectral classification method based on space spectrum neighborhood embedding and optimal similarity graph |
| CN113869454A (en) * | 2021-10-12 | 2021-12-31 | 西北工业大学 | Hyperspectral image sparse feature selection method based on fast embedded spectral analysis |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104392251A (en) * | 2014-11-28 | 2015-03-04 | 西安电子科技大学 | Hyperspectral image classification method based on semi-supervised dictionary learning |
| CN104699781A (en) * | 2015-03-12 | 2015-06-10 | 西安电子科技大学 | Specific absorption rate image retrieval method based on double-layer anchor chart hash |
-
2015
- 2015-08-12 CN CN201510493200.XA patent/CN105160351B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104392251A (en) * | 2014-11-28 | 2015-03-04 | 西安电子科技大学 | Hyperspectral image classification method based on semi-supervised dictionary learning |
| CN104699781A (en) * | 2015-03-12 | 2015-06-10 | 西安电子科技大学 | Specific absorption rate image retrieval method based on double-layer anchor chart hash |
Non-Patent Citations (2)
| Title |
|---|
| HONGYING LIU: "SEMI-SUPERVISED CLASSIFICATION BASED ON ANCHOR-SPATIAL GRAPH FOR LARGE POLARIMETRIC SAR DATA", 《IGARSS 2015》 * |
| LIXIA YANG: "Semi-Supervised Hyperspectral Image Classification Using Spatio-Spectral Laplacian Support Vector Machine", 《IEEE GEOSCIENCE AND REMOTE SENSING LETTERS》 * |
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106127225A (en) * | 2016-06-13 | 2016-11-16 | 西安电子科技大学 | Semi-supervised hyperspectral image classification method based on rarefaction representation |
| CN106127225B (en) * | 2016-06-13 | 2019-07-02 | 西安电子科技大学 | Semi-supervised hyperspectral image classification method based on rarefaction representation |
| CN110874465A (en) * | 2018-08-31 | 2020-03-10 | 浙江大学 | Mobile equipment entity identification method and device based on semi-supervised learning algorithm |
| CN110874465B (en) * | 2018-08-31 | 2022-01-28 | 浙江大学 | Mobile equipment entity identification method and device based on semi-supervised learning algorithm |
| CN109858531A (en) * | 2019-01-14 | 2019-06-07 | 西北工业大学 | A kind of high-spectrum remote sensing quick clustering algorithm based on figure |
| CN109858531B (en) * | 2019-01-14 | 2022-04-26 | 西北工业大学 | Hyperspectral remote sensing image fast clustering algorithm based on graph |
| CN111488923A (en) * | 2020-04-03 | 2020-08-04 | 陕西师范大学 | Enhanced anchor point image semi-supervised classification method |
| CN111488923B (en) * | 2020-04-03 | 2023-02-07 | 陕西师范大学 | Enhanced anchor point image semi-supervised classification method |
| CN111783865A (en) * | 2020-06-23 | 2020-10-16 | 西北工业大学 | Hyperspectral classification method based on space spectrum neighborhood embedding and optimal similarity graph |
| CN111753904A (en) * | 2020-06-24 | 2020-10-09 | 广东工业大学 | Rapid hyperspectral image clustering method, device, equipment and medium |
| CN113869454A (en) * | 2021-10-12 | 2021-12-31 | 西北工业大学 | Hyperspectral image sparse feature selection method based on fast embedded spectral analysis |
| CN113869454B (en) * | 2021-10-12 | 2024-09-10 | 西北工业大学 | Hyperspectral image sparse feature selection method based on rapid embedded spectrum analysis |
Also Published As
| Publication number | Publication date |
|---|---|
| CN105160351B (en) | 2018-11-23 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110321963B (en) | Hyperspectral image classification method based on fusion of multi-scale and multi-dimensional space spectrum features | |
| Wu et al. | ORSIm detector: A novel object detection framework in optical remote sensing imagery using spatial-frequency channel features | |
| CN105160351B (en) | Semi-supervised hyperspectral classification method based on anchor point sparse graph | |
| Zalpour et al. | A new approach for oil tank detection using deep learning features with control false alarm rate in high-resolution satellite imagery | |
| CN104392251B (en) | Hyperspectral image classification method based on semi-supervised dictionary learning | |
| CN107145836B (en) | Hyperspectral image classification method based on stacked boundary identification self-encoder | |
| CN111639587B (en) | Hyperspectral image classification method based on multi-scale spectrum space convolution neural network | |
| CN106503739A (en) | The target in hyperspectral remotely sensed image svm classifier method and system of combined spectral and textural characteristics | |
| CN105528595A (en) | Method for identifying and positioning power transmission line insulators in unmanned aerial vehicle aerial images | |
| CN103996047B (en) | Hyperspectral image classification method based on squeezed spectra clustering ensemble | |
| Boggavarapu et al. | A new framework for hyperspectral image classification using Gabor embedded patch based convolution neural network | |
| CN108229551B (en) | Hyperspectral remote sensing image classification method based on compact dictionary sparse representation | |
| CN104182767B (en) | The hyperspectral image classification method that Active Learning and neighborhood information are combined | |
| CN107590515A (en) | The hyperspectral image classification method of self-encoding encoder based on entropy rate super-pixel segmentation | |
| CN106503727A (en) | A kind of method and device of classification hyperspectral imagery | |
| CN105760900A (en) | Hyperspectral image classification method based on affinity propagation clustering and sparse multiple kernel learning | |
| CN104268556A (en) | Hyperspectral image classification method based on nuclear low-rank representing graph and spatial constraint | |
| CN104778482A (en) | Hyperspectral image classifying method based on tensor semi-supervised scale cutting dimension reduction | |
| CN109034213B (en) | Hyperspectral image classification method and system based on correlation entropy principle | |
| CN107203779A (en) | Hyperspectral dimensionality reduction method based on spatial-spectral information maintenance | |
| CN113723492A (en) | Hyperspectral image semi-supervised classification method and device for improving active deep learning | |
| CN111191700B (en) | Hyperspectral image dimension reduction method and device based on self-adaptive collaborative image discriminant analysis | |
| CN105512622A (en) | Visible remote-sensing image sea-land segmentation method based on image segmentation and supervised learning | |
| CN106096622A (en) | Semi-supervised Classification of hyperspectral remote sensing image mask method | |
| Hu et al. | Cross-domain meta-learning under dual adjustment mode for few-shot hyperspectral image classification |
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 |