CN108830336A - A kind of characters of ground object screening technique towards high score image - Google Patents
A kind of characters of ground object screening technique towards high score image Download PDFInfo
- Publication number
- CN108830336A CN108830336A CN201810676325.XA CN201810676325A CN108830336A CN 108830336 A CN108830336 A CN 108830336A CN 201810676325 A CN201810676325 A CN 201810676325A CN 108830336 A CN108830336 A CN 108830336A
- Authority
- CN
- China
- Prior art keywords
- feature
- target scalar
- extraction
- accidentally
- target
- 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
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/211—Selection of the most significant subset of features
- G06F18/2113—Selection of the most significant subset of features by ranking or filtering the set of features, e.g. using a measure of variance or of feature cross-correlation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
- G06V20/13—Satellite images
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Astronomy & Astrophysics (AREA)
- Remote Sensing (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of characters of ground object screening techniques towards high score image, including according to atural object actual distribution and the remote sensing image form of expression, carry out signature analysis and feature roughing to Target scalar；Using the different images feature of selection, disposable feature extraction is carried out to Target scalar in high-resolution raw video, obtains multiple extraction results of Target scalar；Sensitivity analysis is carried out to each extraction result using the feature-sensitive degree model of building respectively, obtains different images feature to effective extraction degree of Target scalar；Different images feature-sensitive degree is ranked up, higher to Target scalar susceptibility a few feature is selected, forms the sensitive features group of Target scalar.Its remarkable result is：Synthesis examines many-sided consideration, analyzes feature to the sensitivity of Target scalar, improves the Impersonal authenticity for extracting result, to efficiently solve the problems, such as that characters of ground object selection is difficult, reduces interference of the human factor to ground object information extraction.
Description
Technical field
The present invention relates to Remote Sensing Image Processing Technology fields, specifically, being that a kind of atural object towards high score image is special
Levy screening technique.
Background technique
High-resolution remote sensing image includes atural object detailed information abundant, obtains complicated and diversified spectral information, Yi Jikong
Between, the structural information abundant such as texture, topology, semanteme, inevitably there is the phenomenon that characteristic dimension is high, quantity is big,
That is feature " dimension disaster ", how magnanimity characters of ground object to be in optimized selection is the key that realize that terrestrial object information effectively extracts institute
?.
Existing feature selection approach is mainly studied in terms of search strategy and interpretational criteria two.The former algorithms most in use
There are exhaustive search, sequential search, genetic algorithm, ant group algorithm, particle swarm optimization etc., algorithm is based on analyzing, between research characteristic
Degree of relevancy achieve the purpose that characteristic dimensionality reduction, algorithm is complex, and has certain limitation.And press interpretational criteria
Feature selecting algorithm be broadly divided into filtering type, packaged type and combined type, wherein filtering type algorithm include distance metric, information
Entropy measurement, similarity measurement and consistency metric；Packaged type algorithm is to select mistake using classification error probability as interpretational criteria
The smallest feature of probability, but the algorithm needs repeatedly to call learning algorithm, it is more time-consuming.Currently, combined type feature selection approach
It is the main direction of studying of characteristic optimization selection, this method mutually ties filtering type feature selecting and packaged type feature selecting algorithm
It closes, character subset is just chosen with filtering type algorithm, characteristic optimization selection is then realized by packaged type algorithm.
But existing high score characteristics of remote sensing image selection algorithm be mostly by analysis feature between correlation, or
Analysis feature achievees the purpose that characteristic optimization is selected to the correctly or incorrectly extraction effect of ground object sample.In practical applications,
The design of algorithm has biggish limitation, is affected by artificial subjective experience, and algorithm is complicated, analyzes workload
Greatly, it is difficult to realize a wide range of earth's surface information extraction.
Summary of the invention
In view of the deficiencies of the prior art, the object of the present invention is to provide a kind of characters of ground object screening sides towards high score image
Method, this method construct characters of ground object susceptibility from the correctness and accidentally extraction two elements of complexity of characters of ground object information extraction
Analysis model, quantization different images feature solve the problems, such as that characters of ground object selection is difficult to effective extraction degree of Target scalar.
In order to achieve the above objectives, the technical solution adopted by the present invention is as follows：
A kind of characters of ground object screening technique towards high score image, key are to follow the steps below：
Step 1：High-resolution raw video is imported, according to atural object actual distribution and the remote sensing image form of expression, to target
Atural object carries out signature analysis and feature roughing；
Step 2：Disposable feature extraction is carried out to Target scalar using the different images feature of selection, obtains Target scalar
Multiple extraction results；
Step 3：Sensitivity analysis is carried out to each extraction result using the feature-sensitive degree model of building respectively, is obtained not
With image feature to the susceptibility of Target scalar, the feature-sensitive degree model is：
Wherein, α indicates that Target scalar extracts image feature used, SαFor the feature-sensitive degree of image feature α, RαFor image
The correct degree of extraction of the Target scalar of feature α, FαAtural object complexity is accidentally extracted for the Target scalar of image feature α；
Step 4：The susceptibility of different images feature is ranked up, is selected to the higher feature of Target scalar susceptibility,
Form the sensitive features group of Target scalar.
Preferably, in order to reduce interference of the artificial subjective factor to extraction effect, with the letter of complete extraction atural object as far as possible
The principle of breath, method used by feature extraction described in step 2 are carrying out image threshold segmentation method.
Further technical solution is correct degree of extraction R described in step 2αFormula of seeking be：
Wherein, α indicates that Target scalar extracts feature used, AαFor the correct extraction imagery coverage of image feature α, A is indicated
Target scalar imagery coverage.
Further technical solution is that the Target scalar imagery coverage A passes through Imaging enhanced and combines real-time map
It is obtained with related data is consulted.
Further technical solution is that atural object complexity F is accidentally extracted described in step 3αCalculating steps are as follows：
Step 3-1：The inseparable distance for accidentally extracting atural object and Target scalar is calculated, and is calculated using normalization algorithm
To each atural object that accidentally extracts to the weighing factor value of Target scalar, calculation formula is：
Wherein, i indicates that Target scalar, j indicate accidentally to extract atural object, and j ∈ [0, n], n indicate atural object species number,It indicates
Accidentally extract the inseparable degree between atural object j and Target scalar i, WijIt indicates accidentally to extract influence power of the atural object j for Target scalar i
Weight values；
Step 3-2：According to formulaIt calculates and accidentally extracts area weight Eα, whereinFor accidentally mentioning for image feature α
Imagery coverage is taken,Imagery coverage outside for Target scalar；
Step 3-3：According to resulting weighing factor value WijWith accidentally extraction area weight Eα, according to formulaIt calculates and accidentally extracts atural object complexity Fα, whereinIt indicates that feature α is extracted in result accidentally to extract
The area of atural object j.
Further technical solution is the correct degree of extraction R of Target scalarαAtural object complexity is accidentally extracted with Target scalar
FαValue range be [0,1], feature-sensitive degree Sα>0。
The present invention carries out signature analysis for Target scalar first according to atural object actual distribution and the remote sensing images form of expression
And slightly select the feature of suitable atural object expression；Then Target scalar is carried out respectively using the different images feature of selection primary
Property extract, obtain multiple Objects extraction results；Followed by building feature-sensitive degree model to each extraction result respectively into
Row sensitivity analysis obtains different images feature to effective extraction degree of Target scalar, finally using statistical method to not
It is ranked up with image feature susceptibility, selects a few feature higher to Target scalar susceptibility, form the quick of atural object
Feel feature group.
Remarkable result of the invention is：Sensitivity analysis is introduced ground object information extraction field by this programme creativeness, will
Characters of ground object from the correctness of characters of ground object information extraction and misses extraction two levels of complexity as variable, constructs atural object
Feature-sensitive degree analysis model is big to the sensitivity of Target scalar acquisition of information with modeling pattern quantization different images feature
It is small, that is, quantify different images feature to effective extraction degree of Target scalar；
From practical application, the degree that influences each other, the correct recognition effect of feature and the mistake between atural object have been comprehensively considered
Recognition effect etc. analyzes feature to the sensitivity of Target scalar, improves the Impersonal authenticity for extracting result, thus
It efficiently solves the problems, such as that characters of ground object selection is difficult, reduces interference of the human factor to ground object information extraction；And model letter
Single, convenience of calculation is applicable to the acquisition of larger range high score image information, has good universality.
Detailed description of the invention
Fig. 1 is flow chart of the method for the present invention.
Specific embodiment
Specific embodiment and working principle of the present invention will be described in further detail with reference to the accompanying drawing.
As shown in Figure 1, a kind of characters of ground object screening technique towards high score image, specific step is as follows：
Step 1：High-resolution raw video is imported, and according to atural object actual distribution and the remote sensing image form of expression, to mesh
Mark atural object carries out signature analysis and slightly selects the feature of suitable atural object expression；
Step 2：Disposable feature extraction is carried out to Target scalar based on different images feature, obtains the multiple of Target scalar
Extract result.
In the present embodiment, in order to reduce interference of the artificial subjective factor to extraction effect, with complete extraction as far as possible
The principle of object information carries out ground object information extraction by the way of Threshold segmentation classification.
Step 3：Sensitivity analysis is carried out to each extraction result using the feature-sensitive degree model of building respectively, is obtained not
With image feature to effective extraction degree, that is, susceptibility of Target scalar, detailed process is as follows：
Firstly, calculating the correct degree of extraction R of Target scalarα；
In characters of ground object information extraction, the final goal that correct Target scalar information is ground object information extraction is obtained.And
The correct degree of extraction of Target scalar has a positive effect in ground object information extraction, therefore the present invention passes through Imaging enhanced, and combines real
When Baidu map and consult related data, obtain the practical image capturing range of Target scalar, and count its area, as correct degree of extraction
The ratio of the foundation of evaluation, the imagery coverage for then using feature correctly to extract and Target scalar influence area is as feature to target
The correct degree of extraction of atural object, is indicated with R, and value range is between 0~1, shown in expression such as formula (1)：
Wherein, α indicates that Target scalar extracts feature used, AαFor the correct extraction imagery coverage of image feature α, A is indicated
Target scalar imagery coverage.
Then, it calculates and accidentally extracts atural object complexity Fα：
During remote sensing image ground object information extraction, due to foreign matter with spectrum, different spectrum jljl phenomena such as it is generally existing, lead
Cause is usually associated with much noise during the extraction process.The present embodiment is carrying out characters of ground object to pass through list in sensitivity analysis
One feature disposably extracts Target scalar, while to extract Target scalar as far as possible, as a result can mix other cultural noises,
And noise type, distribution situation and the area of all kinds of feature extractions also can difference, work is extracted to the essence of succeeding target atural object
Have large effect.
Therefore, in order to which the influence degree extracted to Target scalar part is accidentally extracted in quantitative study, the present embodiment passes through analysis
Partial noise atural object is accidentally extracted to the weighing factor of Target scalar, and combines the area statistics of noise atural object, proposes accidentally to extract ground
Object complexity, seeks that steps are as follows：
Step 3-1：Sample selection is carried out to survey region, calculates the separability distance of other atural objects and Target scalar,
Inseparable distance is obtained, then obtains each atural object to the influence degree of Target scalar, as atural object using normalization algorithm
Weighing factor value.
There are many determination method of topographical features separability, such as J-M distance (Jeffries-Matusita Distance), sample
Relative distance etc. between this average distance, sample dispersion degree, classification.Domestic and foreign scholars generally believe J-M distance more suitable for
The expression of classification separability, J-M distance is the Spectral divisibility index based on conditional probability theory, and specific calculating is as follows：
In formula, i, j respectively indicate Target scalar and accidentally extract atural object, and j ∈ [0, n], μ, V respectively indicate ground object sample classification
The mean value and variance matrix of characteristic value, T representing matrix transposition.JMijValue between 0~2, size represent Target scalar i and
The separable degree between atural object j is accidentally extracted, the separability being worth between two atural objects of bigger explanation is better.
Therefore, the present embodiment usesIndicate Target scalar i and accidentally extract atural object j between can not
Separation degree, i.e. influence degree between atural object obtain accidentally extracting atural object to the shadow of Target scalar then by normalized
Ring weighted value Wij, and its value range is [0,1], specific formula is as follows：
Step 3-2：Mistake based on image feature α extracts the imagery coverage outside imagery coverage and Target scalar, and calculating accidentally mentions
Take area weight Eα：
This example is using the ratio of imagery coverage outside accidentally extraction imagery coverage and Target scalar as accidentally extraction area weight.On ground
Object extracts in analysis of complexity, accidentally extracts the similarity degree that imagery coverage embodies Target scalar and other atural objects, accidentally extracts image
Area is bigger, illustrates that the two information under feature of the same race is more similar, is also more unfavorable for Target scalar information extraction.
Therefore, described accidentally to extract area weight EαCalculation formula it is as follows：
In formula,Indicate that the mistake of feature α extracts imagery coverage,Indicate the imagery coverage outside Target scalar, EαIt is characterized
The mistake of α extracts area weight, and value range is [0,1].EαThe extraction result of=0 expression feature α is all Target scalar, without it
His atural object, the result are perfect condition；Eα=1 expression feature extraction result contains entire image capturing range, i.e. shadow outside target
As all accidentally being extracted, this category feature cannot function as ground object information extraction feature, and rejecting has been carried out in Feature Selection for this example.
Step 3-3：According to resulting weighing factor value WijWith accidentally extraction area weight Eα, calculate and accidentally extract atural object complexity
Fα, calculation formula is：
In formula, j indicates that jth class accidentally extracts atural object, and value range is [0, n], and n indicates accidentally to extract atural object species number,Table
Show that feature α extracts the area that atural object j is accidentally extracted in result,Indicate that the mistake of feature α extracts imagery coverage, EαIndicate that feature α is missed
Extract area weight, WijIndicate accidentally to extract atural object j for the weighing factor value of Target scalar i.
Finally, construction feature sensitivity model：
Since correct degree of extraction is directly proportional to feature-sensitive degree, i.e., correct extraction degree is higher, then feature sensitivity is better；
And accidentally extract atural object complexity and be inversely proportional with feature-sensitive degree, i.e., accidentally extraction atural object complexity is higher, then feature sensitivity is poorer.
According to above-mentioned principle, and combines above-mentioned correct degree of extraction and accidentally extract the description of atural object complexity formula, building
Shown in feature-sensitive degree model such as formula (7)：
Wherein, α indicates that Target scalar extracts image feature used, SαFor the feature-sensitive degree of image feature α, RαFor image
The correct degree of extraction of the Target scalar of feature α, FαAtural object complexity, R are accidentally extracted for the Target scalar of image feature ααAnd FαValue is equal
Between 0~1, SαFor the value greater than 0, SαValue is bigger, and the sensibility for illustrating that feature α extracts Target scalar is better, otherwise poorer.
Step 4：Using statistical method, different images feature-sensitive degree is ranked up, obtains characters of ground object sequence, and
A few feature higher to Target scalar susceptibility is selected from characters of ground object sequence, forms the sensitive features of Target scalar
Group.
Sensitivity analysis is introduced ground object information extraction field by the invention, using characters of ground object as variable, and
From the correctness and accidentally extraction two levels of complexity of characters of ground object information extraction, characters of ground object sensitivity analysis model is constructed,
With modeling pattern quantization different images feature to the sensitivity size of Target scalar acquisition of information, that is, different images spy is quantified
Effective extraction degree to Target scalar is levied, efficiently solves the problems, such as that characters of ground object selection is difficult.
Claims (6)
1. a kind of characters of ground object screening technique towards high score image, it is characterised in that follow the steps below：
Step 1：High-resolution raw video is imported, according to atural object actual distribution and the remote sensing image form of expression, to Target scalar
Carry out signature analysis and feature roughing；
Step 2：Disposable feature extraction is carried out to Target scalar using the different images feature of selection, obtains the more of Target scalar
A extraction result；
Step 3：Sensitivity analysis is carried out to each extraction result using the feature-sensitive degree model of building respectively, obtains different shadows
As feature is to the susceptibility of Target scalar, the feature-sensitive degree model is：
Wherein, α indicates that Target scalar extracts image feature used, SαFor the feature-sensitive degree of image feature α, RαFor image feature α
The correct degree of extraction of Target scalar, FαAtural object complexity is accidentally extracted for the Target scalar of image feature α；
Step 4：The susceptibility of different images feature is ranked up, selection forms the higher feature of Target scalar susceptibility
The sensitive features group of Target scalar.
2. the characters of ground object screening technique according to claim 1 towards high score image, it is characterised in that：Institute in step 2
Stating method used by feature extraction is carrying out image threshold segmentation method.
3. the characters of ground object screening technique according to claim 1 towards high score image, it is characterised in that：Institute in step 3
State correct degree of extraction RαFormula of seeking be：
Wherein, α indicates that Target scalar extracts feature used, AαFor the correct extraction imagery coverage of image feature α, A is with indicating target
Object imagery coverage.
4. the characters of ground object screening technique according to claim 3 towards high score image, it is characterised in that：The target
Object imagery coverage A, which passes through Imaging enhanced and combines real-time map and consult related data, to be obtained.
5. the characters of ground object screening technique according to claim 1 towards high score image, it is characterised in that：Institute in step 3
It states and accidentally extracts atural object complexity FαCalculating steps are as follows：
Step 3-1：The inseparable distance for accidentally extracting atural object and Target scalar is calculated, and is calculated respectively using normalization algorithm
To the weighing factor value of Target scalar, calculation formula is a atural object that accidentally extracts：
Wherein, i indicates that Target scalar, j indicate accidentally to extract atural object, and j ∈ [0, n], n indicate atural object species number,Expression accidentally mentions
Take the inseparable degree between atural object j and Target scalar i, WijIndicate accidentally to extract atural object j for the weighing factor of Target scalar i
Value；
Step 3-2：According to formulaIt calculates and accidentally extracts area weight Eα, whereinShadow is extracted for the mistake of image feature α
Image planes product,Imagery coverage outside for Target scalar；
Step 3-3：According to resulting weighing factor value WijWith accidentally extraction area weight Eα, according to formulaIt calculates and accidentally extracts atural object complexity Fα, whereinIt indicates that feature α is extracted in result accidentally to extract
The area of atural object j.
6. described in any item characters of ground object screening techniques towards high score image according to claim 1~5, it is characterised in that：
The correct degree of extraction R of Target scalarαAtural object complexity F is accidentally extracted with Target scalarαValue range be [0,1], feature is quick
Sensitivity Sα>0。
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810676325.XA CN108830336B (en) | 2018-06-27 | 2018-06-27 | Ground feature screening method for high-resolution images |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810676325.XA CN108830336B (en) | 2018-06-27 | 2018-06-27 | Ground feature screening method for high-resolution images |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108830336A true CN108830336A (en) | 2018-11-16 |
| CN108830336B CN108830336B (en) | 2021-08-13 |
Family
ID=64138780
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810676325.XA Active CN108830336B (en) | 2018-06-27 | 2018-06-27 | Ground feature screening method for high-resolution images |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108830336B (en) |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20100076723A1 (en) * | 2008-09-23 | 2010-03-25 | Microsoft Corporation | Tensor linear laplacian discrimination for feature extraction |
| CN101706806A (en) * | 2009-11-11 | 2010-05-12 | 北京航空航天大学 | Text classification method by mean shift based on feature selection |
| CN103093182A (en) * | 2011-11-03 | 2013-05-08 | 北京师范大学 | Remote-sensing image gradation classification and identification method based on multisensibility strategy |
| CN103778467A (en) * | 2014-01-16 | 2014-05-07 | 天津大学 | Power system transient state stability estimation inputted characteristic quantity selection method |
| CN105069463A (en) * | 2015-07-17 | 2015-11-18 | 重庆交通大学 | Object-oriented multiple scale mountainous city land coverage information obtaining method |
| CN105243387A (en) * | 2015-07-30 | 2016-01-13 | 王植 | Open-pit mine typical ground object classification method based on UAV image |
| CN105320963A (en) * | 2015-10-21 | 2016-02-10 | 哈尔滨工业大学 | High resolution remote sensing image oriented large scale semi-supervised feature selection method |
| CN105608293A (en) * | 2016-01-28 | 2016-05-25 | 武汉大学 | Forest aboveground biomass inversion method and system fused with spectrum and texture features |
-
2018
- 2018-06-27 CN CN201810676325.XA patent/CN108830336B/en active Active
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20100076723A1 (en) * | 2008-09-23 | 2010-03-25 | Microsoft Corporation | Tensor linear laplacian discrimination for feature extraction |
| CN101706806A (en) * | 2009-11-11 | 2010-05-12 | 北京航空航天大学 | Text classification method by mean shift based on feature selection |
| CN103093182A (en) * | 2011-11-03 | 2013-05-08 | 北京师范大学 | Remote-sensing image gradation classification and identification method based on multisensibility strategy |
| CN103778467A (en) * | 2014-01-16 | 2014-05-07 | 天津大学 | Power system transient state stability estimation inputted characteristic quantity selection method |
| CN105069463A (en) * | 2015-07-17 | 2015-11-18 | 重庆交通大学 | Object-oriented multiple scale mountainous city land coverage information obtaining method |
| CN105243387A (en) * | 2015-07-30 | 2016-01-13 | 王植 | Open-pit mine typical ground object classification method based on UAV image |
| CN105320963A (en) * | 2015-10-21 | 2016-02-10 | 哈尔滨工业大学 | High resolution remote sensing image oriented large scale semi-supervised feature selection method |
| CN105608293A (en) * | 2016-01-28 | 2016-05-25 | 武汉大学 | Forest aboveground biomass inversion method and system fused with spectrum and texture features |
Non-Patent Citations (3)
| Title |
|---|
| KAMILA PAWLUSZEK等: "《Sensitivity analysis of automatic landslide mapping: numerical experiments towards the best solution》", 《LANDSLIDES》 * |
| 潘建平 等: "《基于遥感分类与数学形态学的道路信息提取》", 《基于遥感分类与数学形态学的道路信息提取》 * |
| 王馨爽 等: "《多时相双极化合成孔径雷达干涉测量土地覆盖分类方法》", 《测绘学报》 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108830336B (en) | 2021-08-13 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110399909B (en) | Hyperspectral image classification method based on label constraint elastic network graph model | |
| CN105608474B (en) | Region adaptivity plant extraction method based on high resolution image | |
| CN107229917B (en) | A kind of several remote sensing image general character well-marked target detection methods based on iteration cluster | |
| Yun et al. | PNN based crop disease recognition with leaf image features and meteorological data | |
| CN104598908B (en) | A kind of crops leaf diseases recognition methods | |
| CN109325431B (en) | Method and device for detecting vegetation coverage in feeding path of grassland grazing sheep | |
| CN106503739A (en) | The target in hyperspectral remotely sensed image svm classifier method and system of combined spectral and textural characteristics | |
| CN107239514A (en) | A kind of plants identification method and system based on convolutional neural networks | |
| CN111666617A (en) | Shear wall structure arrangement method and device based on countermeasure generation network | |
| CN108241871A (en) | Laser point cloud and visual fusion data classification method based on multiple features | |
| CN108732172A (en) | Chinese medicine performance rating method, equipment and medium | |
| CN105005784B (en) | A kind of time series remote sensing image land cover classification method based on CD-DTW distance | |
| CN101556600B (en) | Method for retrieving images in DCT domain | |
| CN109389080A (en) | Hyperspectral image classification method based on semi-supervised WGAN-GP | |
| CN109284786A (en) | The SAR image terrain classification method of confrontation network is generated based on distribution and structure matching | |
| CN105320970B (en) | A kind of potato disease diagnostic device, diagnostic system and diagnostic method | |
| CN109829494A (en) | A kind of clustering ensemble method based on weighting similarity measurement | |
| CN110276746A (en) | A kind of robustness method for detecting change of remote sensing image | |
| CN108364011A (en) | PolSAR image multi-stage characteristics extract and unsupervised segmentation method | |
| CN108898096A (en) | A kind of quick accurate extracting method of the information towards high score image | |
| CN109886267A (en) | A kind of soft image conspicuousness detection method based on optimal feature selection | |
| CN107798345A (en) | Based on the diagonal EO-1 hyperion camouflaged target detection method with low-rank representation of block | |
| Qi et al. | Global–local 3-D convolutional transformer network for hyperspectral image classification | |
| CN109447111A (en) | A kind of remote sensing supervised classification method based on subclass training sample | |
| CN109784401A (en) | A kind of Classification of Polarimetric SAR Image method based on ACGAN |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |