CN101814148A - Remote sensing hyperspectral image classification method based on semi-supervised kernel adaptive learning - Google Patents
Remote sensing hyperspectral image classification method based on semi-supervised kernel adaptive learning Download PDFInfo
- Publication number
- CN101814148A CN101814148A CN 201010160203 CN201010160203A CN101814148A CN 101814148 A CN101814148 A CN 101814148A CN 201010160203 CN201010160203 CN 201010160203 CN 201010160203 A CN201010160203 A CN 201010160203A CN 101814148 A CN101814148 A CN 101814148A
- Authority
- CN
- China
- Prior art keywords
- training sample
- optimum
- remote sensing
- kernel function
- function
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Abstract
The invention discloses a remote sensing hyperspectral image classification method based on semi-supervised kernel adaptive learning, relating to a classification method of remote sensing hyperspectral images and solving the problem of low resolution ratio existing in the traditional remote sensing hyperspectral image classification method. The remote sensing hyperspectral image classification method comprises the steps of: judging a labeling form of a hyperspectral image training sample set, acquiring an optimized target function, and then acquiring an optimal parameter or a data dependency kernel parameter; obtaining an optimal kernel function with an invariant structure or a variable structure according to the acquired parameter, further obtaining an optimal semi-supervised classifier, and realizing the classification of the remote sensing hyperspectral images by utilizing the classifier. The invention can accurately classify the end members of the remote hyperspectral images, improves the resolution ratio of the remote sensing hyperspectral images, and can be applied to the technical field of terrain military target reconnaissance, high-efficient warfare striking effect estimation, navy submarine real-time maritime environment monitoring and emergency responses of emergent natural disasters.
Description
Technical field
The present invention relates to a kind of sorting technique of remote sensing hyperspectral image.
Background technology
High spectrum image not only wave band is more, generally observes the type of ground objects complexity, can be divided into a plurality of atural object classifications such as vegetation, field, buildings, road, waters, swampland, exposed soil.If each pixel is represented a kind of type of ground objects, this pixel is called end member (End-member) so.For improving image resolution ratio, need end member is classified to distinguish different atural objects.High optical spectrum image end member sorting algorithm commonly used at present can be divided into to be had supervision and not to have supervise algorithm, the former known atural object generic is judged the sorting algorithm of the atural object classification of each end member representative, and the latter then unknown ground species relies on spectrum statistical discrepancy to classify purely.Supervised classification method commonly used comprises methods such as spectrum angle charting method, binary coding method, parallelepipedon method, minimum distance method and maximum likelihood method; No supervised classification method commonly used comprises methods such as IsoData method and K-Means method.
Except above-mentioned traditional sorting technique, also have some new sorting techniques, as sorting technique based on various neural networks, decision tree, support vector machine and expert system etc.
Yet, the high optical spectrum image end member sorting technique is because the restriction that is subjected to gathering sample at present, face the small sample problem that the machine learning field extensively exists and waits to solve, the sorter that is used for the high optical spectrum image end member classification can not obtain maximum generalization ability, can not obtain optimum image resolution.
Summary of the invention
The objective of the invention is to solve present remote sensing hyperspectral image classification method and have the low problem of resolution, a kind of remote sensing hyperspectral image classification method based on semi-supervised kernel adaptive learning is provided.
Based on the remote sensing hyperspectral image classification method of semi-supervised kernel adaptive learning, its process is as follows:
Step 1, the labeling form of high spectrum image training sample set is judged: if labeling form is the class label information, then execution in step two; If labeling form is a side information, then execution in step three;
Step 2, all samples that the high spectrum image training sample is concentrated mark, utilize Fisher criterion and largest interval criterion to obtain the optimization aim function then, by self-adaptation pursuit algorithm the optimization aim function that obtains is calculated then based on genetic algorithm, obtain optimized parameter, execution in step four then;
Step 3, all samples that the high spectrum image training sample is concentrated mark, utilize global flow shape to keep design criteria to obtain the optimization aim function then, by self-adaptation pursuit algorithm the optimization aim function that obtains is calculated then based on Lagrangian method, obtain data and rely on nuclear parameter, execution in step four then;
Step 4, the concrete applicable cases of basis are judged the structure type of kernel function: if structure type is constant structure, then execution in step five; If structure type is for becoming structure, then execution in step six;
The optimum kernel function of step 5, the constant structure of acquisition, execution in step seven then;
Step 6, acquisition become the optimum kernel function of structure, and execution in step seven then;
Step 7, according to the optimum kernel function that obtains, obtain optimum semi-supervised sorter; Utilize the optimum semi-supervised sorter that obtains, the actual measurement remote sensing hyperspectral image is classified, obtain the classification of this remote sensing hyperspectral image.
The present invention can classify to the end member of remote sensing hyperspectral image exactly, improve the resolution of remote sensing hyperspectral image, can be applied to the emergency response technical field of the scouting of landform military target, the assessment of war strike effect efficiently, the real-time maritime environment monitoring of navy submarine, burst disaster.
Description of drawings
Fig. 1 is the process flow diagram of the remote sensing hyperspectral image classification method based on semi-supervised kernel adaptive learning of the present invention; Fig. 2 is the process flow diagram of the detailed process of step 2; Fig. 3 is the process flow diagram of the detailed process of step 3; Fig. 4 is the process flow diagram of the detailed process of step 5; Fig. 5 and Fig. 6 are respectively the process flow diagram of two kinds of processes of the optimum kernel function that obtains the change structure described in the step 6.
Embodiment
Embodiment one:The remote sensing hyperspectral image classification method based on semi-supervised kernel adaptive learning of present embodiment, its process is as follows:
Step 1, the labeling form of high spectrum image training sample set is judged: if labeling form is the class label information, then execution in step two; If labeling form is a side information, then execution in step three;
Step 2, all samples that the high spectrum image training sample is concentrated mark, utilize Fisher criterion and largest interval criterion to obtain the optimization aim function then, by self-adaptation pursuit algorithm the optimization aim function that obtains is calculated then based on genetic algorithm, obtain optimized parameter, execution in step four then;
Step 3, all samples that the high spectrum image training sample is concentrated mark, utilize global flow shape to keep design criteria to obtain the optimization aim function then, by self-adaptation pursuit algorithm the optimization aim function that obtains is calculated then based on Lagrangian method, obtain data and rely on nuclear parameter, execution in step four then;
Step 4, the concrete applicable cases of basis are judged the structure type of kernel function: if structure type is constant structure, then execution in step five; If structure type is for becoming structure, then execution in step six;
The optimum kernel function of step 5, the constant structure of acquisition, execution in step seven then;
Step 6, acquisition become the optimum kernel function of structure, and execution in step seven then;
Step 7, according to the optimum kernel function that obtains, obtain optimum semi-supervised sorter; Utilize the optimum semi-supervised sorter that obtains, the actual measurement remote sensing hyperspectral image is classified, obtain the classification of this remote sensing hyperspectral image.
Step 1 to the flow process of step 7 sees Fig. 1 for details.
The detailed process that described all samples that the high spectrum image training sample is concentrated of step 2 mark can for:
Step 2 one, according to the labeling form of class label information, each sample that the high spectrum image training sample is concentrated marks, and then exists not mark sample;
Wherein, concentrate at described high spectrum image training sample, the classification of part sample is known, then can mark this part sample, and classification the unknown of a part of sample in addition then can't mark described other a part of sample.In addition, in the present embodiment, all categories that the sample of unknown classification is related all has been contained in the classification of all samples of known class.
Step 2 two, with the production model as sorter, be not considered as one group of disappearance parameter with marking the probability that sample belongs to each known class, adopt the EM algorithm not mark estimation then, with results estimated the described sample that do not mark is marked again marking sample.
The detailed step of step 2 is seen Fig. 2.
The detailed process that described all samples that the high spectrum image training sample is concentrated of step 3 mark can for:
The positive constraints graph and the k arest neighbors figure of step 3 one, the described high spectrum image training sample set of acquisition, according to the positive constraints graph and the k arest neighbors figure that obtain, construct an inherent figure, make that the part of this inherence figure is corresponding with described positive constraints graph, make that another part of this inherence figure is corresponding with described k arest neighbors figure, simultaneously, make the compactedness of data in the empirical features space in the figure of this inherence the strongest;
The negative constraints graph and the non-k arest neighbors figure of step 3 two, the described high spectrum image training sample set of acquisition, according to the negative constraints graph and the non-k arest neighbors figure that obtain, construct a punishment figure, make that the part of this punishment figure is corresponding with described negative constraints graph, make that another part of this punishment figure is corresponding with described non-k arest neighbors figure, simultaneously, make between class in this punishment figure the dispersiveness of data in the empirical features space the strongest;
Step 3 three, according to the inherence figure and the punishment figure of structure, all samples that the high spectrum image training sample is concentrated mark.
The detailed step of step 3 is seen Fig. 3.
The detailed process of the optimum kernel function of the constant structure of the described acquisition of step 5 can for:
The data that step May Day, the optimized parameter that step 2 is obtained or step 3 obtain rely on nuclear parameter and carry out initialization, and the parameter after the initialization is encoded the back as candidate solution;
Step 5 two, according to the fitness of the described candidate solution of optimization aim function calculation, keep the candidate solution of fitness more than or equal to threshold value, give up other candidate solution;
Step 5 three is duplicated, is intersected and operation such as variation the candidate solution that keeps, and generates new candidate solution;
The step the May 4th, new candidate solution is decoded, promptly obtain optimum nuclear parameters optimization, and then obtain the optimum kernel function of constant structure.
Wherein, in the process of " data that optimized parameter that step 2 is obtained or step 3 obtain rely on nuclear parameter and carry out initialization " described in step May Day, it is as follows to select described optimized parameter still be that described data dependence and function carry out initialized standard:
If the labeling form in the step 1 is the class label information, the optimized parameter of then selecting step 2 to obtain carries out initialization; If the labeling form in the step 1 is a side information, the data of then selecting step 3 to obtain rely on nuclear parameter and carry out initialization.
In step 5 two, the given value of described threshold value for setting according to actual conditions.
The detailed step of step 5 is seen Fig. 4.
In step 6,
When the labeling form of judging in the step 1 is the class label information, described acquisition become structure optimum kernel function detailed process can into steps A 1 to steps A 3:
Steps A 1, utilize the markup information of high spectrum image training sample set, expansion obtains data and relies on kernel function, try to achieve the relation of described data dependence kernel function and known basic kernel function then, and try to achieve the Fisher tolerance expression formula that described data rely on kernel function, setting up then and relying on the kernel function parameter with these data is the Fisher metric function of independent variable;
Steps A 2, be optimized the objective function design, set up one and be used to find the solution the equation of constraint that optimal data relies on nuclear parameter according to the markup information of described training sample set;
Steps A 3, utilize the method for loop iteration that described equation of constraint is calculated, acquisition is the optimum solution expression formula of variable with the training sample set, with the described optimum solution expression formula of training sample set substitution, calculate and obtain auto-adaptive parameter then, and then obtain to become the optimum kernel function of structure.
Steps A 1 to the detailed step of steps A 3 is seen Fig. 5.
When the labeling form of judging in the step 1 is side information, described acquisition become structure optimum kernel function detailed process can into step B1 to step B3:
Step B1, the training sample set that utilizes side information to mark, the expression formula of sample interval in the acquisition experience mapping space is utilized side information to set up data then and is relied on the kernel function expression formula, is optimized the objective function design;
Step B2, rely on kernel function expression formula and objective function, set up one and be used to find the solution the equation of constraint that optimal data relies on nuclear parameter according to sample interval expression formula, data;
Step B3, utilize method of Lagrange multipliers that described equation of constraint is calculated, acquisition is the optimum solution expression formula of variable with the training sample set, with the described optimum solution expression formula of training sample set substitution, calculate and obtain auto-adaptive parameter then, and then obtain to become the optimum kernel function of structure.
Step B1 sees Fig. 6 to the detailed step of step B3.
The present invention can classify to the end member of remote sensing hyperspectral image exactly, improve the resolution of remote sensing hyperspectral image, can be applied to the emergency response technical field of the scouting of landform military target, the assessment of war strike effect efficiently, the real-time maritime environment monitoring of navy submarine, burst disaster.
Claims (5)
1. based on the remote sensing hyperspectral image classification method of semi-supervised kernel adaptive learning, it is characterized in that its process is as follows:
Step 1, the labeling form of high spectrum image training sample set is judged: if labeling form is the class label information, then execution in step two; If labeling form is a side information, then execution in step three;
Step 2, all samples that the high spectrum image training sample is concentrated mark, utilize Fisher criterion and largest interval criterion to obtain the optimization aim function then, by self-adaptation pursuit algorithm the optimization aim function that obtains is calculated then based on genetic algorithm, obtain optimized parameter, execution in step four then;
Step 3, all samples that the high spectrum image training sample is concentrated mark, utilize global flow shape to keep design criteria to obtain the optimization aim function then, by self-adaptation pursuit algorithm the optimization aim function that obtains is calculated then based on Lagrangian method, obtain data and rely on nuclear parameter, execution in step four then;
Step 4, the concrete applicable cases of basis are judged the structure type of kernel function: if structure type is constant structure, then execution in step five; If structure type is for becoming structure, then execution in step six;
The optimum kernel function of step 5, the constant structure of acquisition, execution in step seven then;
Step 6, acquisition become the optimum kernel function of structure, and execution in step seven then;
Step 7, according to the optimum kernel function that obtains, obtain optimum semi-supervised sorter; Utilize the optimum semi-supervised sorter that obtains, the actual measurement remote sensing hyperspectral image is classified, obtain the classification of this remote sensing hyperspectral image.
2. the remote sensing hyperspectral image classification method based on semi-supervised kernel adaptive learning according to claim 1 is characterized in that the detailed process that described all samples that the high spectrum image training sample is concentrated of step 2 mark is:
Step 2 one, according to the labeling form of class label information, each sample that the high spectrum image training sample is concentrated marks, and then exists not mark sample;
Step 2 two, with the production model as sorter, be not considered as one group of disappearance parameter with marking the probability that sample belongs to each known class, adopt the EM algorithm not mark estimation then, with results estimated the described sample that do not mark is marked again marking sample.
3. the remote sensing hyperspectral image classification method based on semi-supervised kernel adaptive learning according to claim 1 is characterized in that the detailed process that described all samples that the high spectrum image training sample is concentrated of step 3 mark is:
The positive constraints graph and the k arest neighbors figure of step 3 one, the described high spectrum image training sample set of acquisition, according to the positive constraints graph and the k arest neighbors figure that obtain, construct an inherent figure, make that the part of this inherence figure is corresponding with described positive constraints graph, make that another part of this inherence figure is corresponding with described k arest neighbors figure, simultaneously, make the compactedness of data in the empirical features space in the figure of this inherence the strongest;
The negative constraints graph and the non-k arest neighbors figure of step 3 two, the described high spectrum image training sample set of acquisition, according to the negative constraints graph and the non-k arest neighbors figure that obtain, construct a punishment figure, make that the part of this punishment figure is corresponding with described negative constraints graph, make that another part of this punishment figure is corresponding with described non-k arest neighbors figure, simultaneously, make between class in this punishment figure the dispersiveness of data in the empirical features space the strongest;
Step 3 three, according to the inherence figure and the punishment figure of structure, all samples that the high spectrum image training sample is concentrated mark.
4. the remote sensing hyperspectral image classification method based on semi-supervised kernel adaptive learning according to claim 1 is characterized in that the detailed process of the optimum kernel function of the constant structure of the described acquisition of step 5 is:
The data that step May Day, the optimized parameter that step 2 is obtained or step 3 obtain rely on nuclear parameter and carry out initialization, and the parameter after the initialization is encoded the back as candidate solution;
Step 5 two, according to the fitness of the described candidate solution of optimization aim function calculation, keep the candidate solution of fitness more than or equal to threshold value, give up other candidate solution;
Step 5 three is duplicated, is intersected and operation such as variation the candidate solution that keeps, and generates new candidate solution;
The step the May 4th, new candidate solution is decoded, promptly obtain optimum nuclear parameters optimization, and then obtain the optimum kernel function of constant structure.
5. the remote sensing hyperspectral image classification method based on semi-supervised kernel adaptive learning according to claim 1, it is characterized in that in step 6, when the labeling form of judging in the step 1 is the class label information, described acquisition become structure optimum kernel function detailed process can into steps A 1 to steps A 3:
Steps A 1, utilize the markup information of high spectrum image training sample set, expansion obtains data and relies on kernel function, try to achieve the relation of described data dependence kernel function and known basic kernel function then, and try to achieve the Fisher tolerance expression formula that described data rely on kernel function, setting up then and relying on the kernel function parameter with these data is the Fisher metric function of independent variable;
Steps A 2, be optimized the objective function design, set up one and be used to find the solution the equation of constraint that optimal data relies on nuclear parameter according to the markup information of described training sample set;
Steps A 3, utilize the method for loop iteration that described equation of constraint is calculated, acquisition is the optimum solution expression formula of variable with the training sample set, with the described optimum solution expression formula of training sample set substitution, calculate and obtain auto-adaptive parameter then, and then obtain to become the optimum kernel function of structure;
When the labeling form of judging in the step 1 is side information, described acquisition become structure optimum kernel function detailed process can into step B1 to step B3:
Step B1, the training sample set that utilizes side information to mark, the expression formula of sample interval in the acquisition experience mapping space is utilized side information to set up data then and is relied on the kernel function expression formula, is optimized the objective function design;
Step B2, rely on kernel function expression formula and objective function, set up one and be used to find the solution the equation of constraint that optimal data relies on nuclear parameter according to sample interval expression formula, data;
Step B3, utilize method of Lagrange multipliers that described equation of constraint is calculated, acquisition is the optimum solution expression formula of variable with the training sample set, with the described optimum solution expression formula of training sample set substitution, calculate and obtain auto-adaptive parameter then, and then obtain to become the optimum kernel function of structure.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN 201010160203 CN101814148A (en) | 2010-04-30 | 2010-04-30 | Remote sensing hyperspectral image classification method based on semi-supervised kernel adaptive learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN 201010160203 CN101814148A (en) | 2010-04-30 | 2010-04-30 | Remote sensing hyperspectral image classification method based on semi-supervised kernel adaptive learning |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN101814148A true CN101814148A (en) | 2010-08-25 |
Family
ID=42621399
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN 201010160203 Pending CN101814148A (en) | 2010-04-30 | 2010-04-30 | Remote sensing hyperspectral image classification method based on semi-supervised kernel adaptive learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN101814148A (en) |
Cited By (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102073879A (en) * | 2010-12-02 | 2011-05-25 | 南京大学 | Method for identifying characteristic land categories of ocean remote sensing images of coast on basis of semi-supervised learning |
| CN102096825A (en) * | 2011-03-23 | 2011-06-15 | 西安电子科技大学 | Graph-based semi-supervised high-spectral remote sensing image classification method |
| CN102184423A (en) * | 2011-05-17 | 2011-09-14 | 中国科学院遥感应用研究所 | Full-automatic method for precisely extracting regional impervious surface remote sensing information |
| CN102496027A (en) * | 2011-11-10 | 2012-06-13 | 西安电子科技大学 | Semi-supervised image classification method based on constrained adaptive transmission |
| CN102540271A (en) * | 2011-12-27 | 2012-07-04 | 南京理工大学 | Semi-supervised hyperspectral sub-pixel target detection method based on enhanced constraint sparse regression method |
| CN102938072A (en) * | 2012-10-20 | 2013-02-20 | 复旦大学 | Dimension reducing and sorting method of hyperspectral imagery based on blocking low rank tensor analysis |
| CN104036298A (en) * | 2013-09-23 | 2014-09-10 | 苏州工业职业技术学院 | High-spectrum remote sensing image end-member classification method based on Fisher self-adaptive learning |
| CN104778482A (en) * | 2015-05-05 | 2015-07-15 | 西安电子科技大学 | Hyperspectral image classifying method based on tensor semi-supervised scale cutting dimension reduction |
| CN105023239A (en) * | 2015-08-18 | 2015-11-04 | 西安电子科技大学 | Hyperspectral data dimensionality reduction method based on ultra-pixel and maximum boundary distribution |
| CN107506799A (en) * | 2017-09-01 | 2017-12-22 | 北京大学 | A kind of opener classification based on deep neural network is excavated and extended method and device |
| CN109685105A (en) * | 2018-11-16 | 2019-04-26 | 中国矿业大学 | A kind of high spectrum image clustering method based on the study of unsupervised width |
| CN112329654A (en) * | 2020-11-10 | 2021-02-05 | 中国地震局地震预测研究所 | Hyperspectral image data classification method and system based on multi-manifold learning algorithm |
| CN113902717A (en) * | 2021-10-13 | 2022-01-07 | 自然资源部国土卫星遥感应用中心 | Satellite-borne hyperspectral farmland bare soil target identification method based on spectrum library |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20050234955A1 (en) * | 2004-04-15 | 2005-10-20 | Microsoft Corporation | Clustering based text classification |
| CN1987892A (en) * | 2005-12-23 | 2007-06-27 | 北京海鑫科金高科技股份有限公司 | Human face automatic identifying method based on data flow shape |
| US20070239642A1 (en) * | 2006-03-31 | 2007-10-11 | Yahoo!, Inc. | Large scale semi-supervised linear support vector machines |
| CN101520847A (en) * | 2008-02-29 | 2009-09-02 | 富士通株式会社 | Pattern identification device and method |
-
2010
- 2010-04-30 CN CN 201010160203 patent/CN101814148A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20050234955A1 (en) * | 2004-04-15 | 2005-10-20 | Microsoft Corporation | Clustering based text classification |
| CN1987892A (en) * | 2005-12-23 | 2007-06-27 | 北京海鑫科金高科技股份有限公司 | Human face automatic identifying method based on data flow shape |
| US20070239642A1 (en) * | 2006-03-31 | 2007-10-11 | Yahoo!, Inc. | Large scale semi-supervised linear support vector machines |
| CN101520847A (en) * | 2008-02-29 | 2009-09-02 | 富士通株式会社 | Pattern identification device and method |
Non-Patent Citations (2)
| Title |
|---|
| 《复旦学报(自然科学版)》 20041031 祁亨年等 基于多类支持向量机的遥感图像分类及其半监督式改进策略 781-784 1-5 第43卷, 第5期 2 * |
| 《计算机应用研究》 20090531 崔鹏,张汝波 基于核自调整进行半监督聚类 1719-1722 1-5 第26卷, 第5期 2 * |
Cited By (20)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102073879A (en) * | 2010-12-02 | 2011-05-25 | 南京大学 | Method for identifying characteristic land categories of ocean remote sensing images of coast on basis of semi-supervised learning |
| CN102096825A (en) * | 2011-03-23 | 2011-06-15 | 西安电子科技大学 | Graph-based semi-supervised high-spectral remote sensing image classification method |
| CN102096825B (en) * | 2011-03-23 | 2014-04-16 | 西安电子科技大学 | Graph-based semi-supervised high-spectral remote sensing image classification method |
| CN102184423A (en) * | 2011-05-17 | 2011-09-14 | 中国科学院遥感应用研究所 | Full-automatic method for precisely extracting regional impervious surface remote sensing information |
| CN102184423B (en) * | 2011-05-17 | 2014-07-30 | 中国科学院遥感与数字地球研究所 | Full-automatic method for precisely extracting regional impervious surface remote sensing information |
| CN102496027A (en) * | 2011-11-10 | 2012-06-13 | 西安电子科技大学 | Semi-supervised image classification method based on constrained adaptive transmission |
| CN102540271A (en) * | 2011-12-27 | 2012-07-04 | 南京理工大学 | Semi-supervised hyperspectral sub-pixel target detection method based on enhanced constraint sparse regression method |
| CN102540271B (en) * | 2011-12-27 | 2014-03-19 | 南京理工大学 | Semi-supervised hyperspectral sub-pixel target detection method based on enhanced constraint sparse regression method |
| CN102938072B (en) * | 2012-10-20 | 2016-05-11 | 复旦大学 | A kind of high-spectrum image dimensionality reduction and sorting technique based on the tensor analysis of piecemeal low-rank |
| CN102938072A (en) * | 2012-10-20 | 2013-02-20 | 复旦大学 | Dimension reducing and sorting method of hyperspectral imagery based on blocking low rank tensor analysis |
| CN104036298A (en) * | 2013-09-23 | 2014-09-10 | 苏州工业职业技术学院 | High-spectrum remote sensing image end-member classification method based on Fisher self-adaptive learning |
| CN104778482A (en) * | 2015-05-05 | 2015-07-15 | 西安电子科技大学 | Hyperspectral image classifying method based on tensor semi-supervised scale cutting dimension reduction |
| CN104778482B (en) * | 2015-05-05 | 2018-03-13 | 西安电子科技大学 | The hyperspectral image classification method that dimension about subtracts is cut based on the semi-supervised scale of tensor |
| CN105023239A (en) * | 2015-08-18 | 2015-11-04 | 西安电子科技大学 | Hyperspectral data dimensionality reduction method based on ultra-pixel and maximum boundary distribution |
| CN105023239B (en) * | 2015-08-18 | 2018-03-13 | 西安电子科技大学 | The high-spectral data dimension reduction method being distributed based on super-pixel and maximum boundary |
| CN107506799A (en) * | 2017-09-01 | 2017-12-22 | 北京大学 | A kind of opener classification based on deep neural network is excavated and extended method and device |
| CN107506799B (en) * | 2017-09-01 | 2020-04-24 | 北京大学 | Deep neural network-based mining and expanding method and device for categories of development |
| CN109685105A (en) * | 2018-11-16 | 2019-04-26 | 中国矿业大学 | A kind of high spectrum image clustering method based on the study of unsupervised width |
| CN112329654A (en) * | 2020-11-10 | 2021-02-05 | 中国地震局地震预测研究所 | Hyperspectral image data classification method and system based on multi-manifold learning algorithm |
| CN113902717A (en) * | 2021-10-13 | 2022-01-07 | 自然资源部国土卫星遥感应用中心 | Satellite-borne hyperspectral farmland bare soil target identification method based on spectrum library |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN101814148A (en) | Remote sensing hyperspectral image classification method based on semi-supervised kernel adaptive learning | |
| Sameen et al. | Application of convolutional neural networks featuring Bayesian optimization for landslide susceptibility assessment | |
| Gao et al. | Vectornet: Encoding hd maps and agent dynamics from vectorized representation | |
| Kusumam et al. | 3D‐vision based detection, localization, and sizing of broccoli heads in the field | |
| Polewski et al. | Detection of fallen trees in ALS point clouds using a Normalized Cut approach trained by simulation | |
| JP2019125340A (en) | Systems and methods for automated inferencing of changes in spatiotemporal images | |
| US11003908B2 (en) | Remote-sensing-based detection of soybean aphid induced stress in soybean | |
| CN106096627A (en) | The Polarimetric SAR Image semisupervised classification method that considering feature optimizes | |
| CN113095442A (en) | Hail identification method based on semi-supervised learning under multi-dimensional radar data | |
| CN113297572B (en) | Deep learning sample-level anti-attack defense method and device based on neuron activation mode | |
| Alkan et al. | A smart agricultural application: automated detection of diseases in vine leaves usinghybrid deep learning | |
| Almero et al. | An aquaculture-based binary classifier for fish detection using multilayer artificial neural network | |
| WO2021043126A1 (en) | System and method for event recognition | |
| CN104077610B (en) | The method of the SAR image target recognition of two-dimension non linearity projection properties | |
| Ma et al. | Vessel motion pattern recognition based on one-way distance and spectral clustering algorithm | |
| Aversano et al. | Water stress classification using Convolutional Deep Neural Networks. | |
| Eeti et al. | A modified class-specific weighted soft voting for bagging ensemble | |
| Wong et al. | Probabilistic multi svm weed species classification for weed scouting and selective spot weeding | |
| CN104036298A (en) | High-spectrum remote sensing image end-member classification method based on Fisher self-adaptive learning | |
| CN104463230B (en) | A kind of band selection method using the target in hyperspectral remotely sensed image of constraint piecemeal in pairs | |
| Pu et al. | Image change detection based on the minimum mean square error | |
| Shenoy et al. | Anamoly detection in wireless sensor networks | |
| Li et al. | Weed identification based on shape features and ant colony optimization algorithm | |
| Sahay et al. | An uncertainty quantification framework for counter unmanned aircraft systems using deep ensembles | |
| Liu et al. | A method for missing data recovery of waste gas monitoring in animal building based on GA-SVM |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| C02 | Deemed withdrawal of patent application after publication (patent law 2001) | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Open date: 20100825 |