CN105279519A - Remote sensing image water body extraction method and system based on cooperative training semi-supervised learning - Google Patents
Remote sensing image water body extraction method and system based on cooperative training semi-supervised learning Download PDFInfo
- Publication number
- CN105279519A CN105279519A CN201510616672.XA CN201510616672A CN105279519A CN 105279519 A CN105279519 A CN 105279519A CN 201510616672 A CN201510616672 A CN 201510616672A CN 105279519 A CN105279519 A CN 105279519A
- Authority
- CN
- China
- Prior art keywords
- remote sensing
- sensing image
- sorter
- sample set
- sample
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
Abstract
The invention relates to the technical field of water body remote sensing, provides a remote sensing image water body extraction method and system based on cooperative training semi-supervised learning, and solves the problem that the water body extraction precision is poor due to an insufficient manually-selected sample size when a conventional supervised classification method is adopted for performing remote sensing image water body extraction. The system comprises a remote sensing image feature extraction module, a double-view construction module, a classifier training module and a classification module. The classifier training module comprises an initialization module, a sample marking module, a sample set updating module and an iteration control module. The technical scheme provided by the invention can still ensure the high-resolution remote sensing image water body extraction precision under a small sample condition, thereby reducing the workload and the complexity of manually marking a training sample.
Description
Technical field
The present invention relates to water body remote sensing technology field, particularly a kind of remote sensing image Clean water withdraw method and system based on coorinated training semi-supervised learning.
Background technology
Traditional remote sensing image Clean water withdraw method comprises water body index method, spectrum-photometric method, usually uses image classification method wave band number is less when image spatial resolution is higher in addition.In current remote sensing image Clean water withdraw research, many employings is supervised classification method, such as maximum likelihood method, support vector machine, neural network etc., and the basic step taking these methods to carry out Clean water withdraw is:
(1) feature (based on spectral signature and textural characteristics) the composition characteristic vector of each pixel of image is extracted;
(2) some training samples (being labeled as water body or non-water body) are chosen, and adopt certain supervised learning method to learn training sample set, the aim of learning reaches to minimize empiric risk or structure risk, finally obtains a decision function or rule;
(3) take above-mentioned decision function or the pixel of rule to unknown classifications all in image to classify, be judged as water body or non-water body, to reach the object of Clean water withdraw.
When in image when background atural object more complicated, due to the existence of the different spectrum of jljl and same object different images phenomenon, be often difficult to when causing taking traditional supervised classification obtain representative enough good training sample, thus have influence on the precision of Clean water withdraw.So, in order to improve the precision of Clean water withdraw, just must obtain abundant or even unlimited many training samples, and the quantity of training sample is The more the better naturally, therefore obtaining a large amount of training samples just becomes most important condition.But the sample of artificial mark is limited after all, and the operation complexity of mark training sample is very high, and the sample therefore obtaining a large amount of mark is unpractical, must seek other approach.
Summary of the invention
[technical matters that will solve]
The object of this invention is to provide a kind of remote sensing image Clean water withdraw method based on coorinated training semi-supervised learning, when adopting supervised classification method to carry out remote sensing image Clean water withdraw at present to solve, the problem that the sample size deficiency due to artificial selection causes Clean water withdraw precision poor.
[technical scheme]
The present invention is achieved by the following technical solutions.
First the present invention relates to a kind of remote sensing image Clean water withdraw method based on coorinated training semi-supervised learning, comprises step:
A, the spectral signature extracting remote sensing image and textural characteristics, described spectral signature at least comprises each wave band data X of remote sensing image, water body index NDWI and vegetation index NDVI, described textural characteristics at least comprises the fractal dimension FD of angle second moment ASM, the homogeneity HOM of this gray level co-occurrence matrixes of the gray level co-occurrence matrixes of remote sensing image, the entropy ENT of this gray level co-occurrence matrixes and remote sensing image fractal texture model, wherein X=[B
1, B
2..., B
n]
t, n is wave band number, B
ifor the gray-scale value of wave band i image, 1≤i≤n;
B, build according to the spectral signature of the remote sensing image extracted in steps A and textural characteristics the dual-view be shown below:
C, from remote sensing image, select initial training sample and carry out mark to initial training sample to obtain mark sample set L, from remote sensing image residue pixel, stochastic generation does not mark sample set U;
D, maximum iteration time is set, never marks and take out a sample set U' at random in sample set U;
The view V of E, use mark sample set L
1train a sorter h
1, use the view V of mark sample set L
2train a sorter h
2, use sorter h
1data in sample set U' to be classified and p the sample the highest to degree of confidence marks, use sorter h
2data in sample set U' to be classified and p the sample the highest to degree of confidence marks;
F, the sample of mark new in step e to be joined in mark sample set L, and random never mark in sample set U is selected 2p not mark sample to add in sample set U';
G, whether be empty, if iterations reaches maximum iteration time or do not mark sample set U for empty, then preserve sorter h if judging whether iterations reaches maximum iteration time and do not mark sample set U
1with sorter h
2and perform next step, otherwise then return step e and carry out next iteration;
H, the sorter h using step G to obtain
1with sorter h
2to Remote sensing image classification, extract the Water-Body Information of remote sensing image.
As one preferred embodiment, described step H specifically comprises:
Read remote sensing image, extract the proper vector of each pixel;
Use sorter h
1to the view V of pixel
1classify, obtain confidence alpha
1;
Use sorter h
2to the view V of pixel
2classify, obtain confidence alpha
2;
If α
1> α
2, then this pixel is sorter h
1the classification of prediction, on the contrary then this pixel is sorter h
2the classification of prediction.
As another preferred embodiment, described water body index
wherein Green, NIR are the reflectivity of green light band, near-infrared band respectively.
As another preferred embodiment, described water body index
wherein Red, NIR are the reflectivity of red spectral band, near-infrared band respectively.
As another preferred embodiment, the sample size marking sample set L in described step C is 1:20 ~ 1:10 with the ratio of the sample size not marking sample set U.
The invention still further relates to a kind of remote sensing image Clean water withdraw system based on coorinated training semi-supervised learning, comprise characteristics of remote sensing image extraction module, dual-view builds module, sorter training module and sort module, described sorter training module comprises initialization module, sample labeling module, sample set update module and iteration control module
Described characteristics of remote sensing image extraction module is configured to the spectral signature and the textural characteristics that extract remote sensing image, described spectral signature at least comprises each wave band data X of remote sensing image, water body index NDWI and vegetation index NDVI, described textural characteristics at least comprises the fractal dimension FD of angle second moment ASM, the homogeneity HOM of this gray level co-occurrence matrixes of the gray level co-occurrence matrixes of remote sensing image, the entropy ENT of this gray level co-occurrence matrixes and remote sensing image fractal texture model, wherein X=[B
1, B
2..., B
n]
t, n is wave band number, B
ifor the gray-scale value of wave band i image, 1≤i≤n;
Described dual-view builds module and is configured to the spectral signature of the remote sensing image extracted according to characteristics of remote sensing image extraction module and textural characteristics builds the dual-view be shown below:
Described initialization module is configured to: from remote sensing image, select initial training sample and carry out mark to initial training sample to obtain mark sample set L, from remote sensing image residue pixel, stochastic generation does not mark sample set U, maximum iteration time is set, never marks a random taking-up sample set U' in sample set U;
Described sample labeling module is configured to: the view V using mark sample set L
1train a sorter h
1, use the view V of mark sample set L
2train a sorter h
2, use sorter h
1data in sample set U' to be classified and p the sample the highest to degree of confidence marks, use sorter h
2data in sample set U' to be classified and p the sample the highest to degree of confidence marks;
Described sample set update module is configured to: joined in mark sample set L by the sample of mark new in sample labeling module, and random never mark in sample set U is selected 2p not mark sample to add in sample set U';
Described iteration control module is configured to: whether be empty, if iterations reaches maximum iteration time or do not mark sample set U for empty, then preserve sorter h if judging whether iterations reaches maximum iteration time or do not mark sample set U
1with sorter h
2and by sorter h
1with sorter h
2be sent to sort module, otherwise then return sample labeling module and carry out next iteration;
Described sort module is configured to: use the sorter h obtained
1with sorter h
2to Remote sensing image classification, extract the Water-Body Information of remote sensing image.
As one preferred embodiment, described sort module comprises:
Remote sensing image read module, it is configured to: read remote sensing image, extract the proper vector of each pixel;
First classification submodule, it is configured to: use sorter h
1to the view V of pixel
1classify, obtain confidence alpha
1;
Second classification submodule, it is configured to: use sorter h
2to the view V of pixel
2classify, obtain confidence alpha
2;
Classification judge module, it is configured to: if α
1> α
2, then this pixel is sorter h
1the classification of prediction, on the contrary then this pixel is sorter h
2the classification of prediction.
As another preferred embodiment, described water body index
wherein Green, NIR are the reflectivity of green light band, near-infrared band respectively.
As another preferred embodiment, described vegetation index
wherein Red, NIR are the reflectivity of red spectral band, near-infrared band respectively.
As another preferred embodiment, the sample size marking sample set L in described initialization module is 1:20 ~ 1:10 with the ratio of the sample size not marking sample set U.
[beneficial effect]
The technical scheme that the present invention proposes has following beneficial effect:
Technical scheme of the present invention is on the basis building image feature dual-view, the quantity having mark sample is constantly increased by coorinated training mode, thus get two all very strong sorters of generalization ability, finally be as the criterion with the predicting the outcome of that sorter that degree of confidence is higher when carrying out the identification of water body pixel, even if when original have mark sample size less, owing to there being mark sample size to be also ever-increasing in training process, the precision of remote sensing image Clean water withdraw therefore also can be ensured.Therefore, the present invention still can ensure the precision of high-resolution remote sensing image Clean water withdraw under Small Sample Size, thus can reduce workload and the complexity of artificial mark training sample.
Accompanying drawing explanation
The structured flowchart of the remote sensing image Clean water withdraw system based on coorinated training semi-supervised learning that Fig. 1 provides for embodiments of the invention one.
The process flow diagram of the remote sensing image Clean water withdraw method based on coorinated training semi-supervised learning that Fig. 2 provides for embodiments of the invention two.
Embodiment
For making the object, technical solutions and advantages of the present invention clearly, carry out clear, complete description by the specific embodiment of the present invention below.
Embodiment one
The structured flowchart of the remote sensing image Clean water withdraw system based on coorinated training semi-supervised learning that Fig. 1 provides for the embodiment of the present invention one.As shown in Figure 1, this system comprises characteristics of remote sensing image extraction module, dual-view builds module, sorter training module and sort module, and sorter training module comprises initialization module, sample labeling module, sample set update module and iteration control module.
Characteristics of remote sensing image extraction module is configured to the spectral signature and the textural characteristics that extract remote sensing image, spectral signature comprises each wave band data X of remote sensing image, water body index NDWI and vegetation index NDVI, textural characteristics at least comprises the fractal dimension FD of angle second moment ASM, the homogeneity HOM of this gray level co-occurrence matrixes of the gray level co-occurrence matrixes of remote sensing image, the entropy ENT of this gray level co-occurrence matrixes and remote sensing image fractal texture model, wherein X=[B
1, B
2..., B
n]
t, n is wave band number, B
ifor the gray-scale value of wave band i image, 1≤i≤n.Water body index
wherein Green, NIR are the reflectivity of green light band, near-infrared band respectively.Vegetation index
wherein Red, NIR are the reflectivity of red spectral band, near-infrared band respectively.
Dual-view builds module and is configured to the spectral signature of the remote sensing image extracted according to characteristics of remote sensing image extraction module and textural characteristics builds the dual-view be shown below:
Initialization module is configured to: from remote sensing image, select initial training sample and carry out mark to initial training sample to obtain mark sample set L, from remote sensing image residue pixel, stochastic generation does not mark sample set U, maximum iteration time is set, never marks a random taking-up sample set U' in sample set U.Particularly, sample class is water body and non-water body two class, and label is {-1 ,+1}, mark sample set L={x
1, x
2..., x
l, do not mark sample set U={x
l+1..., x
l+u, the sample size of mark sample set L is 1:15, i.e. l:u=1:15 with the ratio of the sample size not marking sample set U.
Sample labeling module is configured to: the view V using mark sample set L
1train a sorter h
1, use the view V of mark sample set L
2train a sorter h
2, use sorter h
1data in sample set U' to be classified and p the sample the highest to degree of confidence marks, use sorter h
2data in sample set U' to be classified and p the sample the highest to degree of confidence marks.Need to illustrate, the present invention does not limit the value of p.
Sample set update module is configured to: joined in mark sample set L by the sample of mark new in sample labeling module, and random never mark in sample set U is selected 2p not mark sample to add in sample set U'.
Iteration control module is configured to: whether be empty, if iterations reaches maximum iteration time or do not mark sample set U for empty, then preserve sorter h if judging whether iterations reaches maximum iteration time or do not mark sample set U
1with sorter h
2and by sorter h
1with sorter h
2be sent to sort module, otherwise then return sample labeling module and carry out next iteration.
Sort module is configured to: use the sorter h obtained
1with sorter h
2to Remote sensing image classification, extract the Water-Body Information of remote sensing image.Particularly, sort module comprises remote sensing image read module, the first classification submodule, the second classification submodule and classification judge module.
Remote sensing image read module is configured to: read remote sensing image, extract the proper vector of each pixel;
First classification submodule is configured to: use sorter h
1to the view V of pixel
1classify, obtain confidence alpha
1;
Second classification submodule is configured to: use sorter h
2to the view V of pixel
2classify, obtain confidence alpha
2;
Classification judge module is configured to: if α
1> α
2, then this pixel is sorter h
1the classification of prediction, on the contrary then this pixel is sorter h
2the classification of prediction.
Embodiment two
The process flow diagram of the remote sensing image Clean water withdraw method based on coorinated training semi-supervised learning that Fig. 2 provides for the embodiment of the present invention two.As shown in Figure 2, the method comprising the steps of S1, to step S7, is described in detail to above-mentioned steps below respectively.
Step S1: characteristics of remote sensing image extracts.
In step S1, extract spectral signature and the textural characteristics of remote sensing image, spectral signature comprises each wave band data X of remote sensing image, water body index NDWI and vegetation index NDVI, textural characteristics comprises the fractal dimension FD of angle second moment ASM, the homogeneity HOM of this gray level co-occurrence matrixes of the gray level co-occurrence matrixes of remote sensing image, the entropy ENT of this gray level co-occurrence matrixes and remote sensing image fractal texture model, wherein X=[B
1, B
2..., B
n]
t, n is wave band number, B
ifor the gray-scale value of wave band i image, 1≤i≤n.Water body index
wherein Green, NIR are the reflectivity of green light band, near-infrared band respectively.Vegetation index
wherein Red, NIR are the reflectivity of red spectral band, near-infrared band respectively.
Step S2: dual-view builds.
In step S2, build according to the spectral signature of the remote sensing image extracted in step S1 and textural characteristics the dual-view be shown below:
Step S3: sorter training initialization.
In step S3, from remote sensing image, select initial training sample and carry out mark to initial training sample to obtain mark sample set L, from remote sensing image residue pixel, stochastic generation does not mark sample set U, maximum iteration time is set, never marks a random taking-up sample set U' in sample set U.Particularly, sample class is water body and non-water body two class, and corresponding label is {-1 ,+1}, mark sample set L={x
1, x
2..., x
l, do not mark sample set U={x
l+1..., x
l+u, the sample size of mark sample set L is 1:15, i.e. l:u=1:15 with the ratio of the sample size not marking sample set U.
Step S4: sample marks.
In step S4, use the view V of mark sample set L
1train a sorter h
1, use the view V of mark sample set L
2train a sorter h
2, use sorter h
1data in sample set U' to be classified and p the sample the highest to degree of confidence marks, use sorter h
2data in sample set U' to be classified and p the sample the highest to degree of confidence marks.Need to illustrate, the present invention does not limit the value of p.
Step S5: sample set upgrades.
In step S5, the sample of mark new in step S4 is joined in mark sample set L, and random never mark in sample set U is selected 2p not mark sample to add in sample set U'.
Step S6: judge whether to meet stopping criterion for iteration, if meet stopping criterion for iteration, preserves sorter h
1with sorter h
2and perform step S7, otherwise then return step S4 and carry out next iteration.
In step S6, whether be empty, if iterations reaches maximum iteration time or do not mark sample set U for empty, then preserve sorter h if judging whether iterations reaches maximum iteration time and do not mark sample set U
1with sorter h
2and perform next step, otherwise then return step S4 and carry out next iteration.
Step S7: to Remote sensing image classification, extracts the Water-Body Information of remote sensing image.
In step S7, use the sorter h that step S6 obtains
1with sorter h
2to Remote sensing image classification, extract the Water-Body Information of remote sensing image.Step S7 specifically comprises: read remote sensing image, extract the proper vector of each pixel; Use sorter h
1to the view V of pixel
1classify, obtain confidence alpha
1; Use sorter h
2to the view V of pixel
2classify, obtain confidence alpha
2; If α
1> α
2, then this pixel is sorter h
1the classification of prediction, on the contrary then this pixel is sorter h
2the classification of prediction.
As can be seen from the above embodiments, the embodiment of the present invention is on the basis building image feature dual-view, the quantity having mark sample is constantly increased by coorinated training mode, thus get two all very strong sorters of generalization ability, finally be as the criterion with the predicting the outcome of that sorter that degree of confidence is higher when carrying out the identification of water body pixel, even if when original have mark sample size less, owing to there being mark sample size to be also ever-increasing in training process, the precision of remote sensing image Clean water withdraw therefore also can be ensured.Therefore, the present invention still can ensure the precision of high-resolution remote sensing image Clean water withdraw under Small Sample Size, thus can reduce workload and the complexity of artificial mark training sample.
Need to illustrate, the embodiment of foregoing description is a part of embodiment of the present invention, instead of whole embodiment, neither limitation of the present invention.Based on embodiments of the invention, those of ordinary skill in the art, not paying the every other embodiment obtained under creative work prerequisite, belong to protection scope of the present invention.
Claims (10)
1., based on a remote sensing image Clean water withdraw method for coorinated training semi-supervised learning, it is characterized in that comprising step:
A, the spectral signature extracting remote sensing image and textural characteristics, described spectral signature at least comprises each wave band data X of remote sensing image, water body index NDWI and vegetation index NDVI, described textural characteristics at least comprises the fractal dimension FD of angle second moment ASM, the homogeneity HOM of this gray level co-occurrence matrixes of the gray level co-occurrence matrixes of remote sensing image, the entropy ENT of this gray level co-occurrence matrixes and remote sensing image fractal texture model, wherein X=[B
1, B
2..., B
n]
t, n is wave band number, B
ifor the gray-scale value of wave band i image, 1≤i≤n;
B, build according to the spectral signature of the remote sensing image extracted in steps A and textural characteristics the dual-view be shown below:
C, from remote sensing image, select initial training sample and carry out mark to initial training sample to obtain mark sample set L, from remote sensing image residue pixel, stochastic generation does not mark sample set U;
D, maximum iteration time is set, never marks and take out a sample set U' at random in sample set U;
The view V of E, use mark sample set L
1train a sorter h
1, use the view V of mark sample set L
2train a sorter h
2, use sorter h
1data in sample set U' to be classified and p the sample the highest to degree of confidence marks, use sorter h
2data in sample set U' to be classified and p the sample the highest to degree of confidence marks;
F, the sample of mark new in step e to be joined in mark sample set L, and random never mark in sample set U is selected 2p not mark sample to add in sample set U';
G, whether be empty, if iterations reaches maximum iteration time or do not mark sample set U for empty, then preserve sorter h if judging whether iterations reaches maximum iteration time and do not mark sample set U
1with sorter h
2and perform next step, otherwise then return step e and carry out next iteration;
H, the sorter h using step G to obtain
1with sorter h
2to Remote sensing image classification, extract the Water-Body Information of remote sensing image.
2. the remote sensing image Clean water withdraw method based on coorinated training semi-supervised learning according to claim 1, is characterized in that described step H specifically comprises:
Read remote sensing image, extract the proper vector of each pixel;
Use sorter h
1to the view V of pixel
1classify, obtain confidence alpha
1;
Use sorter h
2to the view V of pixel
2classify, obtain confidence alpha
2;
If α
1> α
2, then this pixel is sorter h
1the classification of prediction, on the contrary then this pixel is sorter h
2the classification of prediction.
3. the remote sensing image Clean water withdraw method based on coorinated training semi-supervised learning according to claim 1 and 2, is characterized in that described water body index
wherein Green, NIR are the reflectivity of green light band, near-infrared band respectively.
4. the remote sensing image Clean water withdraw method based on coorinated training semi-supervised learning according to claim 1 and 2, is characterized in that described vegetation index
wherein Red, NIR are the reflectivity of red spectral band, near-infrared band respectively.
5. the remote sensing image Clean water withdraw method based on coorinated training semi-supervised learning according to claim 1 and 2, is characterized in that the sample size marking sample set L in described step C is 1:20 ~ 1:10 with the ratio of the sample size not marking sample set U.
6. the remote sensing image Clean water withdraw system based on coorinated training semi-supervised learning, it is characterized in that comprising characteristics of remote sensing image extraction module, dual-view builds module, sorter training module and sort module, described sorter training module comprises initialization module, sample labeling module, sample set update module and iteration control module
Described characteristics of remote sensing image extraction module is configured to the spectral signature and the textural characteristics that extract remote sensing image, described spectral signature at least comprises each wave band data X of remote sensing image, water body index NDWI and vegetation index NDVI, described textural characteristics at least comprises the fractal dimension FD of angle second moment ASM, the homogeneity HOM of this gray level co-occurrence matrixes of the gray level co-occurrence matrixes of remote sensing image, the entropy ENT of this gray level co-occurrence matrixes and remote sensing image fractal texture model, wherein X=[B
1, B
2..., B
n]
t, n is wave band number, B
ifor the gray-scale value of wave band i image, 1≤i≤n;
Described dual-view builds module and is configured to the spectral signature of the remote sensing image extracted according to characteristics of remote sensing image extraction module and textural characteristics builds the dual-view be shown below:
Described initialization module is configured to: from remote sensing image, select initial training sample and carry out mark to initial training sample to obtain mark sample set L, from remote sensing image residue pixel, stochastic generation does not mark sample set U, maximum iteration time is set, never marks a random taking-up sample set U' in sample set U;
Described sample labeling module is configured to: the view V using mark sample set L
1train a sorter h
1, use the view V of mark sample set L
2train a sorter h
2, use sorter h
1data in sample set U' to be classified and p the sample the highest to degree of confidence marks, use sorter h
2data in sample set U' to be classified and p the sample the highest to degree of confidence marks;
Described sample set update module is configured to: joined in mark sample set L by the sample of mark new in sample labeling module, and random never mark in sample set U is selected 2p not mark sample to add in sample set U';
Described iteration control module is configured to: whether be empty, if iterations reaches maximum iteration time or do not mark sample set U for empty, then preserve sorter h if judging whether iterations reaches maximum iteration time or do not mark sample set U
1with sorter h
2and by sorter h
1with sorter h
2be sent to sort module, otherwise then return sample labeling module and carry out next iteration;
Described sort module is configured to: use the sorter h obtained
1with sorter h
2to Remote sensing image classification, extract the Water-Body Information of remote sensing image.
7. the remote sensing image Clean water withdraw system based on coorinated training semi-supervised learning according to claim 6, is characterized in that described sort module comprises:
Remote sensing image read module, it is configured to: read remote sensing image, extract the proper vector of each pixel;
First classification submodule, it is configured to: use sorter h
1to the view V of pixel
1classify, obtain confidence alpha
1;
Second classification submodule, it is configured to: use sorter h
2to the view V of pixel
2classify, obtain confidence alpha
2;
Classification judge module, it is configured to: if α
1> α
2, then this pixel is sorter h
1the classification of prediction, on the contrary then this pixel is sorter h
2the classification of prediction.
8. the remote sensing image Clean water withdraw system based on coorinated training semi-supervised learning according to claim 6 or 7, is characterized in that described water body index
wherein Green, NIR are the reflectivity of green light band, near-infrared band respectively.
9. the remote sensing image Clean water withdraw system based on coorinated training semi-supervised learning according to claim 6 or 7, is characterized in that described vegetation index
wherein Red, NIR are the reflectivity of red spectral band, near-infrared band respectively.
10. the remote sensing image Clean water withdraw system based on coorinated training semi-supervised learning according to claim 6 or 7, is characterized in that the sample size marking sample set L in described initialization module is 1:20 ~ 1:10 with the ratio of the sample size not marking sample set U.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510616672.XA CN105279519B (en) | 2015-09-24 | 2015-09-24 | Remote sensing image Clean water withdraw method and system based on coorinated training semi-supervised learning |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510616672.XA CN105279519B (en) | 2015-09-24 | 2015-09-24 | Remote sensing image Clean water withdraw method and system based on coorinated training semi-supervised learning |
Publications (2)
Publication Number | Publication Date |
---|---|
CN105279519A true CN105279519A (en) | 2016-01-27 |
CN105279519B CN105279519B (en) | 2018-09-21 |
Family
ID=55148500
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201510616672.XA Active CN105279519B (en) | 2015-09-24 | 2015-09-24 | Remote sensing image Clean water withdraw method and system based on coorinated training semi-supervised learning |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN105279519B (en) |
Cited By (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107229910A (en) * | 2017-05-18 | 2017-10-03 | 北京环境特性研究所 | A kind of remote sensing images icing lake detection method and its system |
CN107451616A (en) * | 2017-08-01 | 2017-12-08 | 西安电子科技大学 | Multi-spectral remote sensing image terrain classification method based on the semi-supervised transfer learning of depth |
CN107644235A (en) * | 2017-10-24 | 2018-01-30 | 广西师范大学 | Image automatic annotation method based on semi-supervised learning |
CN108229425A (en) * | 2018-01-29 | 2018-06-29 | 浙江大学 | A kind of identifying water boy method based on high-resolution remote sensing image |
CN108805208A (en) * | 2018-06-13 | 2018-11-13 | 哈尔滨工业大学 | A kind of coorinated training method based on unlabeled exemplars consistency checking |
WO2019075771A1 (en) * | 2017-10-20 | 2019-04-25 | Huawei Technologies Co., Ltd. | Self-training method and system for semi-supervised learning with generative adversarial networks |
CN110633758A (en) * | 2019-09-20 | 2019-12-31 | 四川长虹电器股份有限公司 | Method for detecting and locating cancer region aiming at small sample or sample unbalance |
CN111144487A (en) * | 2019-12-27 | 2020-05-12 | 二十一世纪空间技术应用股份有限公司 | Method for establishing and updating remote sensing image sample library |
CN111259366A (en) * | 2020-01-22 | 2020-06-09 | 支付宝(杭州)信息技术有限公司 | Verification code recognizer training method and device based on self-supervision learning |
CN112131912A (en) * | 2019-06-25 | 2020-12-25 | 北京尚水信息技术股份有限公司 | Remote sensing image underlying surface extraction method |
WO2022214039A1 (en) * | 2021-04-07 | 2022-10-13 | 中国农业大学 | Remote sensing identification method for agricultural planting structure |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102054274A (en) * | 2010-12-01 | 2011-05-11 | 南京大学 | Method for full automatic extraction of water remote sensing information in coastal zone |
CN102855494A (en) * | 2012-08-20 | 2013-01-02 | 中国测绘科学研究院 | Method and device for extracting water body of satellite remote sensing image |
CN103034863A (en) * | 2012-12-24 | 2013-04-10 | 重庆市勘测院 | Remote-sensing image road acquisition method combined with kernel Fisher and multi-scale extraction |
CN103559500A (en) * | 2013-10-15 | 2014-02-05 | 北京航空航天大学 | Multispectral remote sensing image land feature classification method based on spectrum and textural features |
CN103793907A (en) * | 2013-12-12 | 2014-05-14 | 中国科学院深圳先进技术研究院 | Water body information extracting method and device |
-
2015
- 2015-09-24 CN CN201510616672.XA patent/CN105279519B/en active Active
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102054274A (en) * | 2010-12-01 | 2011-05-11 | 南京大学 | Method for full automatic extraction of water remote sensing information in coastal zone |
CN102855494A (en) * | 2012-08-20 | 2013-01-02 | 中国测绘科学研究院 | Method and device for extracting water body of satellite remote sensing image |
CN103034863A (en) * | 2012-12-24 | 2013-04-10 | 重庆市勘测院 | Remote-sensing image road acquisition method combined with kernel Fisher and multi-scale extraction |
CN103559500A (en) * | 2013-10-15 | 2014-02-05 | 北京航空航天大学 | Multispectral remote sensing image land feature classification method based on spectrum and textural features |
CN103793907A (en) * | 2013-12-12 | 2014-05-14 | 中国科学院深圳先进技术研究院 | Water body information extracting method and device |
Non-Patent Citations (2)
Title |
---|
X.GIGANDET 等: "REGION-BASED SATELLITE IMAGE CLASSIFICATION:METHOD AND VALIDATION", 《IMAGE PROCESSING,2005. ICIP.2005. IEEE INTERNATIONAL CONFERENCE ON》 * |
于重重 等: ""组合标记的多视图半监督协同分类算法"", 《计算机应用》 * |
Cited By (14)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107229910A (en) * | 2017-05-18 | 2017-10-03 | 北京环境特性研究所 | A kind of remote sensing images icing lake detection method and its system |
CN107451616A (en) * | 2017-08-01 | 2017-12-08 | 西安电子科技大学 | Multi-spectral remote sensing image terrain classification method based on the semi-supervised transfer learning of depth |
WO2019075771A1 (en) * | 2017-10-20 | 2019-04-25 | Huawei Technologies Co., Ltd. | Self-training method and system for semi-supervised learning with generative adversarial networks |
US11120337B2 (en) | 2017-10-20 | 2021-09-14 | Huawei Technologies Co., Ltd. | Self-training method and system for semi-supervised learning with generative adversarial networks |
CN107644235A (en) * | 2017-10-24 | 2018-01-30 | 广西师范大学 | Image automatic annotation method based on semi-supervised learning |
CN108229425A (en) * | 2018-01-29 | 2018-06-29 | 浙江大学 | A kind of identifying water boy method based on high-resolution remote sensing image |
CN108805208A (en) * | 2018-06-13 | 2018-11-13 | 哈尔滨工业大学 | A kind of coorinated training method based on unlabeled exemplars consistency checking |
CN108805208B (en) * | 2018-06-13 | 2021-12-31 | 哈尔滨工业大学 | Collaborative training method based on consistency judgment of label-free samples |
CN112131912A (en) * | 2019-06-25 | 2020-12-25 | 北京尚水信息技术股份有限公司 | Remote sensing image underlying surface extraction method |
CN110633758A (en) * | 2019-09-20 | 2019-12-31 | 四川长虹电器股份有限公司 | Method for detecting and locating cancer region aiming at small sample or sample unbalance |
CN111144487A (en) * | 2019-12-27 | 2020-05-12 | 二十一世纪空间技术应用股份有限公司 | Method for establishing and updating remote sensing image sample library |
CN111144487B (en) * | 2019-12-27 | 2023-09-26 | 二十一世纪空间技术应用股份有限公司 | Method for establishing and updating remote sensing image sample library |
CN111259366A (en) * | 2020-01-22 | 2020-06-09 | 支付宝(杭州)信息技术有限公司 | Verification code recognizer training method and device based on self-supervision learning |
WO2022214039A1 (en) * | 2021-04-07 | 2022-10-13 | 中国农业大学 | Remote sensing identification method for agricultural planting structure |
Also Published As
Publication number | Publication date |
---|---|
CN105279519B (en) | 2018-09-21 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN105279519A (en) | Remote sensing image water body extraction method and system based on cooperative training semi-supervised learning | |
CN110443143B (en) | Multi-branch convolutional neural network fused remote sensing image scene classification method | |
CN110909820B (en) | Image classification method and system based on self-supervision learning | |
CN103049763B (en) | Context-constraint-based target identification method | |
CN103984959B (en) | A kind of image classification method based on data and task-driven | |
CN110390251B (en) | Image and character semantic segmentation method based on multi-neural-network model fusion processing | |
CN107016664B (en) | A kind of bad needle flaw detection method of large circle machine | |
CN109583425A (en) | A kind of integrated recognition methods of the remote sensing images ship based on deep learning | |
CN109711448A (en) | Based on the plant image fine grit classification method for differentiating key field and deep learning | |
CN107943856A (en) | A kind of file classification method and system based on expansion marker samples | |
CN105488536A (en) | Agricultural pest image recognition method based on multi-feature deep learning technology | |
CN107766890A (en) | The improved method that identification segment learns in a kind of fine granularity identification | |
CN105184298A (en) | Image classification method through fast and locality-constrained low-rank coding process | |
CN105069481A (en) | Multi-label natural scene classification method based on spatial pyramid and sparse coding | |
CN110414616B (en) | Remote sensing image dictionary learning and classifying method utilizing spatial relationship | |
CN104268552B (en) | One kind is based on the polygonal fine classification sorting technique of part | |
Isobe et al. | Deep convolutional encoder-decoder network with model uncertainty for semantic segmentation | |
CN107480723B (en) | Texture Recognition based on partial binary threshold learning network | |
CN110807485B (en) | Method for fusing two-classification semantic segmentation maps into multi-classification semantic map based on high-resolution remote sensing image | |
CN107025442A (en) | A kind of multi-modal fusion gesture identification method based on color and depth information | |
CN113657414B (en) | Object identification method | |
CN104866850A (en) | Optimized binarization method for document images | |
Li et al. | Alter-cnn: An approach to learning from label proportions with application to ice-water classification | |
CN113408606A (en) | Semi-supervised small sample image classification method based on graph collaborative training | |
CN115410059B (en) | Remote sensing image part supervision change detection method and device based on contrast loss |
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 |