CN101826160B - Hyperspectral image classification method based on immune evolutionary strategy - Google Patents
Hyperspectral image classification method based on immune evolutionary strategy Download PDFInfo
- Publication number
- CN101826160B CN101826160B CN2010101392492A CN201010139249A CN101826160B CN 101826160 B CN101826160 B CN 101826160B CN 2010101392492 A CN2010101392492 A CN 2010101392492A CN 201010139249 A CN201010139249 A CN 201010139249A CN 101826160 B CN101826160 B CN 101826160B
- Authority
- CN
- China
- Prior art keywords
- population
- antibody
- spectrum
- classification
- affinity
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Expired - Fee Related
Links
Images
Landscapes
- Investigating Or Analysing Materials By Optical Means (AREA)
Abstract
The invention relates to a hyperspectral image classification method based on an immune evolutionary strategy and develops a corresponding simulation prototype system. The system comprises the following four functional modules: a human-machine interface module, a hyperspectral optimal band selection module, a hyperspectral terrain classification module and a classification result output module. The method comprises the following steps: 1. obtaining the initial data and related initialization operations; 2. initializing populations; 3. initially selecting the populations; 4. cloning the populations; 5. mutating the populations in a mixed manner; 6. selecting the memory populations; 7. supplementing the population antibodies; 8. carrying out iterative computations and repeating the steps from 3 to 7 until achieving the maximum evolutionary generation; 9. using the optimal antibody to carry out terrain classification on the hyperspectral data; and 10. outputting the terrain classification results of the hyperspectral images. The method can adaptively select the optimal band combination needed by different terrain classifications under different scenes, has better time complexity and good robustness and is high in classification precision and wide in applicable scope.
Description
Technical field
The present invention relates to a kind of hyperspectral image classification method, in the high-spectral data disposal system, belong to the high-spectral data process field based on immune evolutionary strategy.
Background technology
The eighties in 20th century, one of the maximum achievement of remote sensing technology was the rise of high spectrum resolution remote sensing technique, because this technology has the advantage of imaging and spectrographic detection concurrently, was used widely in the civil and military field.Along with the raising of Hyperspectral imaging spectral resolution, make the material that in conventional remote sensing, can't survey, in high-spectrum remote-sensing, can be surveyed, for the later stage carries out the atural object exhaustive division precondition is provided.Although high spectrum can provide abundant terrestrial information, its more wave band number makes its data volume huge, causes information redundancy, has increased the complexity that late time data is handled.
At atural object identification or branch time-like, the wave band number that is not use is The more the better.This be because: (1) can not simply be equal to the information channel number to the spectrum channel number.Because generally have bigger correlativity between the adjacent wave band of high spectrum image,, not only can not improve nicety of grading, on the contrary influence identification or classification results if use all wave bands to discern or classify without analyse.In addition, not all wave band all has equal importance for subsequent treatment, forms new image space through selecting optimum wave band, under the condition of not losing important information, can represent the information of other wave band; (2) select wave band too much, not only increase operand, influence computing velocity, but also need a large amount of training samples, otherwise be difficult to obtain satisfied identification or classification results.Needing under rational mathematical model instructs, to choose the best band combination according to certain criterion to high-spectral data higher-dimension, magnanimity information, carry out the data dimensionality reduction, is the effective way that improves high-spectral data disposal system processing speed and precision.
On the other hand; Research in the last few years shows that Immune System is a height profile, parallel and adaptive system; Safeguarding body health, getting rid of and to have numerous good characteristics aspect the germ invasion; It has the information processing capability of height, has characteristics such as identification, memory, adjusting, study.These characteristics have determined the applications well prospect of Immune System in field of information processing.
Artificial immune system is the immune messaging model of the mechanism of action, particularly level vertebrate animals (mainly being the people) of using for reference vertebrate immune system, is a kind of novel intelligence computation method of base configuration with immunology term and ultimate principle.In in the past several years; The application of artificial immune system has expanded to numerous areas such as information security, pattern-recognition, machine learning, data mining gradually, has demonstrated powerful information processing and problem solving ability and the wide research prospect of immune algorithm.MPP in the immune algorithm, fault-tolerance, self-organization and characteristics such as adaptive ability and association function have achieved a kind of Immune System that receives and have inspired; The learning art of natural defense mechanism through the study external substance; Evolutionary learning mechanism such as noise is restrained oneself, teacherless learning, self-organization, memory are provided; And combined the advantage of systems such as sorter, neural network and machine inference, therefore novel method of dealing with problems and approach be provided.
In the high-spectral data disposal system, from the computation complexity requirement, sorting algorithm will have relative smaller calculation, and often nicety of grading is relatively poor and reach the algorithm of this requirement.From the requirement of nicety of grading, sorting algorithm will have the good robustness to difference classification scene, and often computation complexity is higher and reach the algorithm of this requirement.Therefore, need to seek a kind of efficient algorithm of taking all factors into consideration computation complexity and this two aspects balance of nicety of grading, make it have more excellent time complexity and good robustness.
Summary of the invention
The object of the present invention is to provide a kind of hyperspectral image classification method (OBS-ICS), thereby specifically be that a kind of immune evolutionary strategy that uses is chosen the method that the optimum band combination of high spectrum is classified to different atural objects in the high spectrum image based on immune evolutionary strategy.This method can be chosen the needed best band combination of different terrain classifications in the high spectrum image adaptively, and the less relatively while of its calculated amount has kept higher classification accuracy again.The method is applied to has good characteristics such as stronger robustness and higher nicety of grading in the high-spectral data disposal system.
The inventive method is based on the emulation prototype system, and this system has Man Machine Interface module, the optimum band selection module of high spectrum, high spectrum terrain classification module, four functional modules of classification results output module.Wherein, the Man Machine Interface module is accomplished the setting with correlation parameter of reading in of high-spectral data; The optimum band selection module of high spectrum is mainly accomplished utilization immune evolutionary strategy choosing optimum wave band in the high-spectral data; High spectrum terrain classification module is mainly accomplished the classification of application of optimal wave band to different atural objects in the high-spectral data; The classification results output module is mainly accomplished the classification results of output to the different atural objects of high-spectral data.
Method flow involved in the present invention may further comprise the steps: 1, obtain primary data and relevant initialization operation; 2, initialization of population; 3, population initial selected; 4, population clonal expansion; 5, population mixovariation; 6, select the memory population; 7, population antibody replenishes; 8, iterative computation, repeating step 3~7 is until reaching maximum evolutionary generation; 9, classification uses optimum antibody that high-spectral data is carried out terrain classification; 10 results output, output high spectrum image terrain classification result.Wherein, the corresponding completing steps 1 of Man Machine Interface module, the corresponding completing steps 2~8 of the optimum band selection module of high spectrum, the corresponding completing steps 9 of high spectrum terrain classification module, the corresponding completing steps 10 of classification results output module.
Be elaborated in the face of each step of this method flow down, suppose spectral band set B in the given high spectrum image, note is done
The wave band sum that comprises in the expression high spectrum image, and antibody population set A b, note is made Ab (k)={ ab
1, ab
2..., ab
N, N representes population scale, and satisfies ab
i∈ B (i=1,2 ..., N).
Through Man Machine Interface module completing steps one;
Step 1 obtains primary data and relevant initialization operation
Obtain pending high-spectral data through the Man Machine Interface module, and following correlation parameter is set: maximum evolutionary generation genMax, initial population quantity N
Init, per generation initial selected population quantity N, mixovariation step-length δ selects wave band sum N
B
Through the optimum band selection module of high spectrum completing steps two~step 8;
The step 2 initialization of population
The random initializtion population, obtaining quantity is N
InitThe initial antibodies population, note is done
Ab wherein
i∈ B (i=1,2 ..., N
Init).
Step 3 population initial selected
Obtain each antibody ab through computing formula (1)
i(ab
i∈ Ab
Init(k)) affinity.From initial population Ab
Init(k) select top n antibody, Ab here in
Init(k) the antibody ab in
iAccording to its affinity descending sort.Through population initial selected step, form initial selected population A b (k).
Wherein, Aff representes the size of the affinity between antibody and antigen, and sepIndex representes the separability between atural object, and its value can obtain through computing formula (2):
Wherein, SCM representes the similarity between two spectrum, and BC representes the correlativity between two wave bands, and ED representes the Euclidean distance between two spectrum, and SAM representes the angle between two spectrum.The value of SCM, BC, ED, SAM can obtain through computing formula (3)~(6) respectively.
Wherein, (A B) is related coefficient between spectrum A, the B, μ to SCM
AAnd σ
AAverage and the variance of representing A respectively, μ
BAnd σ
BAverage and the variance of representing B respectively.
Wherein, (i j) is related coefficient between wave band i and wave band j, x to BC
IpAnd x
JpBe respectively the radiance value of p pixel in wave band i and the wave band j, μ
iAnd μ
jBe respectively the radiance value average of wave band i and wave band j.
Wherein, (A B) is Euclidean distance between spectrum A, the B, A to ED
kBe k vectorial component of the spectrum of atural object A, B
kBe k vectorial component of the spectrum of atural object B.
Wherein, arc cos α () is the spectrum angle between spectrum A, the B.
Step 4 population clonal expansion
Resist each the antibody ab in the former population
i(ab
i∈ Ab (k)) carries out clone operations, its clone's quantity N
CiCan obtain through computing formula (7):
Wherein, N
Ci(i=1,2 ..., N) expression ab
iClone's quantity,
Expression is bracket function down, norAff
i(i=1,2 ..., N) antibody ab after the expression normalization
iAffinity.
Through population clonal expansion step; Form total clone's population A b ' (k), population scale is
Step 5 population mixovariation
The clone population high to affinity ab '
i(ab '
i∈ Ab ' (k), i=1,2 ..., p) carry out the single-point mutation operation, and the clone population low to affinity ab '
i(ab '
i∈ Ab ' (k), i=1,2 ..., q, s.t.:p+q=N
Ci) carry out the multiple spot mutation operation.When carrying out the mutation operation of antibody, antibody
All adopt the real coding mode, here b
i∈ B.The single-point mutation operation is described below: at antibody ab
iMiddle 1 b of picked at random
k(k=1,2 ..., N
B), use random value b then
r(b
r∈ [b
k-δ, b
k+ δ]) substitute.The multiple spot mutation operation is described below: select antibody
In every bit b
k(b
k∈ ab
i, k=1,2 ..., N
B), repeat similar operation in the single-point variation.
Through population mixovariation step, form total variation antibody population Ab " (k);
Step 6 is selected the memory population
Through computing formula (1) each the antibody ab in the population that respectively made a variation "
i(ab "
i∈ Ab " (k)) affinity.Each the variation population ab "
j}
iN before middle the selection
MiIndividual antibody, the antibody in each variation population is according to the descending sort of affinity size here, and antibody is selected number N
MiCan obtain through computing formula (8):
Wherein,
Represent to round up function, mean () representes norAffM
iAverage, norAffM
i(i=1,2 ..., N) antibody ab after the expression normalization
iAffinity, N
CiExpression antibody ab
iClone's quantity.
Through selecting memory population step, form total memory antibody population A b " ' (k), population scale is N
M
Step 7 population antibody replenishes
Generate N at random
kIndividual antibody, note is done
The quantity that antibody replenishes can calculate through formula (9).
Wherein, N
k(k=1,2 ..., genMax) represent the antibody quantity that the k time iteration replenished,
The same formula of definition (7), abNum
kThe quantity that comprises antibody when being illustrated in the k time evolution in the antibody population.
Through population antibody replenish step, form follow-on antibody population Ab (k+1) (that is Ab (k+1)=Ab,
Add(k) ∪ Ab " ' (k)).
The step 8 iterative computation
Repeating step three~step 7 is until reaching maximum evolutionary generation;
Through high spectrum terrain classification module completing steps nine.
The step 9 classification
Through the optimum antibody that step 2~step 8 is chosen, promptly optimum band combination carries out terrain classification to high-spectral data;
Through classification results output module completing steps ten.
Step 10 result's output
Output high spectrum image terrain classification result.
The present invention is a kind of hyperspectral image classification method based on immune evolutionary strategy; Its advantage is: be used for the high-spectral data disposal system; Can choose the needed best band combination of different terrain classifications under the different scenes adaptively; More excellent time complexity and good robustness are arranged, and nicety of grading is high, applied widely.
Description of drawings
Shown in Figure 1 is high spectrum image OBS-ICS sorting technique process flow diagram of the present invention
Embodiment
Further specify technical scheme of the present invention below in conjunction with accompanying drawing and embodiment.
Developed the emulation prototype system based on the present invention, this system applies is in the high spectrum image classification of Data.Prototype system comprises four functional modules, is respectively: Man Machine Interface module, the optimum band selection module of high spectrum, high spectrum terrain classification module, classification results output module.In the optimum band selection module of high spectrum, comprise seven sub-function modules again, be respectively: (1) initialization of population submodule, its function are to accomplish the initialization operation of population by the correlation parameter that the Man Machine Interface module obtains; (2) population initial selected submodule, its function are the initial selected operations of the antibody population that obtains in the initialization of population submodule being accomplished population; (3) population clonal expansion submodule, its function are that the antibody population that obtains in the population initial selected submodule is carried out the operation of population clonal expansion; (4) population mixovariation submodule, its function are that the antibody population that obtains in the population clonal expansion submodule is carried out the operation of population mixovariation; (5) select memory population submodule, its function is that the antibody population that obtains in the population mixovariation submodule is carried out the operation of population mixovariation; (6) population antibody replenishes submodule, and its function is that the antibody population of selecting to obtain in the memory population submodule is carried out population antibody complement operation; (7) iterative computation submodule, its function are successively through calculating that the antibody population that obtains in behind the submodule in above-mentioned (2)~(6) is iterated, until reaching maximum iteration time.
Below in conjunction with specific embodiment, the embodiment of analogue system is described in further detail.
The first, obtain pending high-spectral data through the Man Machine Interface module.Present embodiment adopts Washington D.C.Mall high-spectral data; Size is 1280 * 307; Wavelength coverage is 0.4~2.4 μ m, remove water vapor absorption wave band and low signal-to-noise ratio wave band after, keep wherein 191 wave bands; And intercepting wherein a subgraph size be 562 * 307, subgraph comprises 7 types of atural objects altogether: be respectively: roof, meadow, trees, path, street, water, shade.Following correlation parameter is set: maximum evolutionary generation genMax=35, initial population quantity N
Init=80, per generation initial selected population quantity N=50, wave band sum N is selected in mixovariation step-length δ=5
B=11.
The second, the best band that is used for terrain classification through the optimum band selection module acquisition of high spectrum makes up, and promptly embodiment is handled successively as follows:
(i) initialization of population
Use the initialization of population submodule, the random initializtion population, obtaining quantity is N
InitThe initial antibodies population.
(ii) population initial selected
Use population initial selected submodule, obtain each antibody ab through computing formula (1)
i(ab
i∈ Ab
Init(k)) affinity.From initial population Ab
Init(k) select top n antibody, Ab here in
Init(k) the antibody ab in
iAccording to its affinity descending sort.Through the population initial selected, form initial selected population A b (k).
(iii) population clonal expansion
Use population clonal expansion submodule, resist each the antibody ab in the former population
i(ab
i∈ Ab (k)) carries out clone operations, its clone's quantity N
CiCan obtain through computing formula (7).
Through the population clonal expansion, form total clone's population A b ' (k).
(iv) population mixovariation
Use population mixovariation submodule, the clone population high to affinity ab '
i(ab '
i∈ Ab ' (k), i=1,2 ..., p) carry out the single-point mutation operation, and the clone population low to affinity ab '
i(ab '
i∈ Ab ' (k), i=1,2 ..., q, s.t.:p+q=N
Ci) carry out the multiple spot mutation operation.Through the population mixovariation, form variation antibody population Ab " (k).
(v) select the memory population
Use and select memory population submodule, calculate each the antibody ab in the population that respectively makes a variation " through formula (1)
i(ab "
i∈ Ab " (k)) affinity.Each the variation population ab "
j}
iN before middle the selection
MiIndividual antibody, the antibody in each variation population is according to the descending sort of affinity size here, and antibody is selected number N
MiCan obtain through computing formula (8):
Through selecting the memory population, form memory antibody population A b " ' (k).
(vi) population antibody replenishes
Use population antibody to replenish submodule, generate N at random
kIndividual antibody, note is done
(Ab
Add(k) ∈ B), the additional quantity of antibody can calculate through formula (9).
Replenish through population antibody, form follow-on antibody population Ab (k+1) (that is Ab (k+1)=Ab,
Add(k) ∪ Ab " ' (k)).
(vii) iterative computation
Use the iterative computation submodule, repeat (iii)~(vi), until reaching maximum evolutionary generation, obtain the best band combination of high spectrum terrain classification, the best band combination that present embodiment is selected is: 2,39,48,68,69,70,79,80,137,138,191.
The 3rd, through high spectrum terrain classification module, use the best band combination that embodiment is carried out terrain classification.
The 4th, through the classification results output module, output is to the classification results of embodiment.
The inventive method can be chosen the needed best band combination of different terrain classifications under the different scenes adaptively through the practical implementation of analogue system, can accomplish the classification to different atural objects in the high-spectral data.The inventive method is used for the high-spectral data disposal system, has more excellent time complexity and good robustness, and nicety of grading is high, applied widely.
Claims (1)
1. hyperspectral image classification method based on immune evolutionary strategy; And having developed corresponding emulation prototype system, this system has Man Machine Interface module, the optimum band selection module of high spectrum, high spectrum terrain classification module, four functional modules of classification results output module;
This method flow step is following: suppose spectral band set B in the given high spectrum image, note is done
The wave band sum that comprises in the expression high spectrum image, and antibody population set A b, note is made Ab (k)={ ab
1, ab
2..., av
N, N representes population scale, and satisfies ab
i∈ B, wherein, i=1,2 ..., N;
Through Man Machine Interface module completing steps one;
Step 1 obtains primary data and relevant initialization operation
Obtain pending high-spectral data through the Man Machine Interface module, and following correlation parameter is set: maximum evolutionary generation genMax, initial population quantity N
Init, per generation initial selected population quantity N, mixovariation step-length δ selects wave band sum N
B
Through the optimum band selection module of high spectrum completing steps two~step 8;
The step 2 initialization of population
The random initializtion population, obtaining quantity is N
InitThe initial antibodies population, note is done
Ab wherein
i∈ B, i=1,2 ..., N
Init
Step 3 population initial selected
Obtain each antibody ab through computing formula (1)
iAffinity; From initial population Ab
Init(k) select top n antibody, Ab here in
Init(k) the antibody ab in
iAccording to its affinity descending sort; Through population initial selected step, form initial selected population A b (k);
Wherein, Aff representes the size of the affinity between antibody and antigen, and sepIndex representes the separability between atural object, ab
i∈ Ab
Init(k) its value obtains through computing formula (2):
Wherein, SCM representes the similarity between two spectrum, and BC representes the correlativity between two wave bands, and ED representes the Euclidean distance between two spectrum, and SAM representes the angle between two spectrum; The value of SCM, BC, ED, SAM obtains through computing formula (3)~(6) respectively;
Wherein, (A B) is related coefficient between spectrum A, the B, μ to SCM
AAnd σ
AAverage and the variance of representing A respectively, μ
BAnd σ
BAverage and the variance of representing B respectively; N
BThe wave band sum is selected in expression;
Wherein, (i j) is related coefficient between wave band i and wave band j, x to BC
IpAnd x
JpBe respectively the radiance value of p pixel in wave band i and the wave band j, μ
iAnd μ
jBe respectively the radiance value average of wave band i and wave band j;
Wherein, (A B) is Euclidean distance between spectrum A, the B, A to ED
kBe k vectorial component of the spectrum of atural object A, B
kBe k vectorial component of the spectrum of atural object B;
Wherein, arccos α () is the spectrum angle between spectrum A, the B;
Step 4 population clonal expansion
Resist each the antibody ab in the former population
iCarry out clone operations, its clone's quantity N
CiObtain through computing formula (7):
Wherein, N
CiExpression ab
iClone's quantity,
Expression is bracket function down, norAff
iAntibody ab after the expression normalization
iAffinity; Wherein, i=1,2 ..., N; Ab
i∈ Ab (k);
Through population clonal expansion step, form total clone's population A b ' (k), population scale is N
C, wherein
Step 5 population mixovariation
The clone population high to affinity ab '
iCarry out the single-point mutation operation, wherein, ab '
j∈ Ab ' (k), i=1,2 ..., p, and the clone population low to affinity ab '
iCarry out the multiple spot mutation operation; Wherein, ab '
i∈ Ab ' (k), i=1,2 ..., q, s.t.:p+q=N
CiWhen carrying out the mutation operation of antibody, antibody
All adopt the real coding mode, here b
i∈ B; The single-point mutation operation is described below: at antibody ab
iMiddle 1 b of picked at random
k, wherein, k=1,2 ..., N
B, b
k∈ ab
i, use random value b then
rSubstitute; Wherein, b
r∈ [b
k-δ, b
k+ δ], the multiple spot mutation operation is described below: select antibody ab
iIn every bit b
k, repeat similar operation in the single-point variation;
Through population mixovariation step, form total variation antibody population Ab " (k);
Step 6 is selected the memory population
Through computing formula (1) each the antibody ab in the population that respectively made a variation "
iAffinity; Each the variation population ab "
j}
iN before middle the selection
MiIndividual antibody, the antibody in each variation population is according to the descending sort of affinity size here, and antibody is selected number N
MiObtain through computing formula (8):
Wherein,
Represent to round up function, mean () representes norAffM
iAverage, norAffM
iAntibody ab after the expression normalization
iAffinity, N
CiExpression antibody ab
iClone's quantity; Wherein, ab "
i∈ Ab " (k), i=1,2 ..., N
Through selecting memory population step, form total memory antibody population A b " ' (k), population scale is N
M, wherein,
Step 7 population antibody replenishes
Generate N at random
kIndividual antibody, note is done
Wherein, Ab
Add(k) ∈ B, the quantity that antibody replenishes calculates through formula (9);
Wherein, N
kRepresent the antibody quantity that the k time iteration replenished, k=1,2 ..., genMax,
The same formula of definition (7),
AbNum
kThe quantity that comprises antibody when being illustrated in the k time evolution in the antibody population;
Through population antibody replenish step, form follow-on antibody population Ab (k+1) promptly, Ab (k+1)=Ab
Add(k) ∪ Ab " ' (k);
The step 8 iterative computation
Repeating step three~step 7 is until reaching maximum evolutionary generation;
Through high spectrum terrain classification module completing steps nine;
The step 9 classification
Through the optimum antibody that step 2~step 8 is chosen, promptly optimum band combination carries out terrain classification to high-spectral data;
Through classification results output module completing steps ten;
Step 10 result's output
Output high spectrum image terrain classification result.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN2010101392492A CN101826160B (en) | 2010-03-31 | 2010-03-31 | Hyperspectral image classification method based on immune evolutionary strategy |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN2010101392492A CN101826160B (en) | 2010-03-31 | 2010-03-31 | Hyperspectral image classification method based on immune evolutionary strategy |
Publications (2)
Publication Number | Publication Date |
---|---|
CN101826160A CN101826160A (en) | 2010-09-08 |
CN101826160B true CN101826160B (en) | 2012-11-14 |
Family
ID=42690069
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN2010101392492A Expired - Fee Related CN101826160B (en) | 2010-03-31 | 2010-03-31 | Hyperspectral image classification method based on immune evolutionary strategy |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN101826160B (en) |
Families Citing this family (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104820840B (en) * | 2015-05-08 | 2018-03-06 | 西安电子科技大学 | The arest neighbors hyperspectral image classification method recombinated based on dictionary and wave band |
CN106485709A (en) * | 2016-10-17 | 2017-03-08 | 哈尔滨工业大学 | EO-1 hyperion band selection method based on entropy redundancy and Immune Clone Selection |
CN108509840B (en) * | 2018-02-02 | 2021-10-01 | 哈尔滨工程大学 | Hyperspectral remote sensing image waveband selection method based on quantum memory optimization mechanism |
CN108615048B (en) * | 2018-04-04 | 2020-06-23 | 浙江工业大学 | Defense method for image classifier adversity attack based on disturbance evolution |
CN109948693B (en) * | 2019-03-18 | 2021-09-28 | 西安电子科技大学 | Hyperspectral image classification method based on superpixel sample expansion and generation countermeasure network |
CN113011589B (en) * | 2021-03-29 | 2024-03-08 | 湖北工业大学 | Co-evolution-based hyperspectral image band selection method and system |
CN112801881B (en) * | 2021-04-13 | 2021-06-22 | 湖南大学 | High-resolution hyperspectral calculation imaging method, system and medium |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US5751850A (en) * | 1993-06-30 | 1998-05-12 | International Business Machines Corporation | Method for image segmentation and classification of image elements for documents processing |
US7249116B2 (en) * | 2002-04-08 | 2007-07-24 | Fiske Software, Llc | Machine learning |
CN101061510A (en) * | 2004-11-19 | 2007-10-24 | 皇家飞利浦电子股份有限公司 | False positive reduction in computer-assisted detection (CAD) |
CN101110103A (en) * | 2006-07-20 | 2008-01-23 | 中国科学院自动化研究所 | Image registration self-verifying method based on learning |
-
2010
- 2010-03-31 CN CN2010101392492A patent/CN101826160B/en not_active Expired - Fee Related
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US5751850A (en) * | 1993-06-30 | 1998-05-12 | International Business Machines Corporation | Method for image segmentation and classification of image elements for documents processing |
US7249116B2 (en) * | 2002-04-08 | 2007-07-24 | Fiske Software, Llc | Machine learning |
CN101061510A (en) * | 2004-11-19 | 2007-10-24 | 皇家飞利浦电子股份有限公司 | False positive reduction in computer-assisted detection (CAD) |
CN101110103A (en) * | 2006-07-20 | 2008-01-23 | 中国科学院自动化研究所 | Image registration self-verifying method based on learning |
Also Published As
Publication number | Publication date |
---|---|
CN101826160A (en) | 2010-09-08 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Sameen et al. | Application of convolutional neural networks featuring Bayesian optimization for landslide susceptibility assessment | |
CN101826160B (en) | Hyperspectral image classification method based on immune evolutionary strategy | |
Chen et al. | Mapping landslide susceptibility at the Three Gorges Reservoir, China, using gradient boosting decision tree, random forest and information value models | |
Nourani et al. | Daily and monthly suspended sediment load predictions using wavelet based artificial intelligence approaches | |
CN108038445B (en) | SAR automatic target identification method based on multi-view deep learning framework | |
Kanellopoulos et al. | Land-cover discrimination in SPOT HRV imagery using an artificial neural network—a 20-class experiment | |
Zhong et al. | Remote sensing image subpixel mapping based on adaptive differential evolution | |
Zhong et al. | Sub-pixel mapping based on artificial immune systems for remote sensing imagery | |
CN112733725A (en) | Hyperspectral image change detection method based on multistage cyclic convolution self-coding network | |
Xu et al. | Downscaling and projection of multi‐CMIP5 precipitation using machine learning methods in the upper Han River basin | |
CN111695467A (en) | Spatial spectrum full convolution hyperspectral image classification method based on superpixel sample expansion | |
CN102073879A (en) | Method for identifying characteristic land categories of ocean remote sensing images of coast on basis of semi-supervised learning | |
CN111814607A (en) | Deep learning model suitable for small sample hyperspectral image classification | |
Han et al. | Spatial-spectral unsupervised convolutional sparse auto-encoder classifier for hyperspectral imagery | |
Lim et al. | Morphology for jet classification | |
Amini et al. | Effects of Automatic Hyperparameter Tuning on the Performance of Multi‐variate Deep Learning‐based Rainfall Nowcasting | |
Dang et al. | U-shaped deep-learning models for island ecosystem type classification, a case study in Con Dao Island of Vietnam | |
Singh et al. | Mono-and dual-regulated contractive-expansive-contractive deep convolutional networks for classification of multispectral remote sensing images | |
Tian et al. | A deep‐learning ensemble method to detect atmospheric rivers and its application to projected changes in precipitation regime | |
Montanaro et al. | Stack-CNN algorithm: A new approach for the detection of space objects | |
CN104463210A (en) | Polarization SAR image classification method based on object orienting and spectral clustering | |
Teodoro | Applicability of data mining algorithms in the identification of beach features/patterns on high-resolution satellite data | |
Ian et al. | Novel prediction in storm surge using ensemble machine learning algorithms | |
Luwanga et al. | Automatic spread-F detection using deep learning | |
Mohamed et al. | Assessment of ensemble classifiers using the bagging technique for improved land cover classification of multispectral satellite images |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C14 | Grant of patent or utility model | ||
GR01 | Patent grant | ||
C17 | Cessation of patent right | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20121114 Termination date: 20140331 |