CN108446582A - Hyperspectral image classification method based on textural characteristics and affine propagation clustering algorithm - Google Patents
Hyperspectral image classification method based on textural characteristics and affine propagation clustering algorithm Download PDFInfo
- Publication number
- CN108446582A CN108446582A CN201810074087.5A CN201810074087A CN108446582A CN 108446582 A CN108446582 A CN 108446582A CN 201810074087 A CN201810074087 A CN 201810074087A CN 108446582 A CN108446582 A CN 108446582A
- Authority
- CN
- China
- Prior art keywords
- wave band
- band
- image
- training
- group
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
- G06V20/194—Terrestrial scenes using hyperspectral data, i.e. more or other wavelengths than RGB
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A40/00—Adaptation technologies in agriculture, forestry, livestock or agroalimentary production
- Y02A40/10—Adaptation technologies in agriculture, forestry, livestock or agroalimentary production in agriculture
Abstract
The invention discloses a kind of hyperspectral image classification method based on textural characteristics and affine propagation clustering algorithm, mainly solves the problems, such as that prior art classifying quality is poor.Its implementation is:1) hyperspectral image data is read;2) similarity matrix is calculated according to wave band textural characteristics;3) wave band is clustered using AP algorithms according to similarity matrix, obtains the wave band cluster result of high spectrum image;4) the optimal wave band per group cluster result medium wave band subset is calculated;5) training set for calibrating image and test set sample coordinate and corresponding label are obtained;6) training set and test set in optimal wave band are obtained according to training set and test set sample coordinate;7) SVM is trained with training set, obtains training pattern;8) test set type is predicted according to training pattern, obtains the classification results of high spectrum image.The present invention improves the precision of classification hyperspectral imagery, can be used for the identification to remote sensing images in agriculture fine, ECOLOGICAL ENVIRONMENTAL MONITORING and urban planning.
Description
Technical field
The invention belongs to technical field of remote sensing image processing, more particularly to a kind of hyperspectral image classification method can be used for
Identification in agriculture fine, ECOLOGICAL ENVIRONMENTAL MONITORING and urban planning to remote sensing images.
Background technology
Hyperspectral imagery processing has become one of the cutting edge technology of remote sensing fields at present.High-spectrum remote-sensing refers to imaging
Spectrometer is tens of to hundreds of very narrow and continuous in the ultraviolet of electromagnetic spectrum, visible light, near-infrared and mid infrared region
The technology that image data is obtained in spectrum segment, related data is obtained using these spectral bands from interesting target.And bloom
On the one hand modal data can distinguish these terrestrial materials with enough spectral resolutions, on the other hand due to the clutter reflections of its inverting
Spectrum can shine compared with Land Surface Temperatures, therefore research and analyse the spectral signature of atural object and tagsort identification technology and just seem
It is particularly important.
Compared with other remotely sensed image technologies, high-spectrum remote-sensing is with wave band quantity is more, waveband width is narrow, spectral response model
Enclose that wide, spectral resolution is high, can provide the features such as spatial-domain information and spectrum domain information.It will reflect that the spectrum of target emanation is believed
The image information of breath and reflection target two-dimensional space combines in one, realizes " collection of illustrative plates ", i.e. the base in two-dimensional space information
One-dimensional spectral information is added on plinth.Hyperspectral image data is the image cube of two-dimensional space and one-dimensional spectrum, in image
Each wave band is a width two dimensional image in space;It is bent to be then reflected as a continuous spectrum response for each pixel in spectral space
Line, different substances show as different radiation intensity in high spectrum image, thus can to same Target scalar continuous imaging,
It can reflect the diagnostic spectral signature of Target scalar.However, since high-spectral data wave band is numerous, data volume is huge, phase
The characteristics such as correlation between adjacent wave section is strong so that information redundance increases, and causes " dimension disaster ", i.e., image classification accuracy with
The increase for dimension shows as first increasing the Hughes phenomenons reduced afterwards, and this not only adds the operands of data processing, go back shadow
Ring the identification of the precision and target of terrain classification.Therefore, Classification of hyperspectral remote sensing image is generally required first to carry out dimensionality reduction etc. pre-
Processing.
There are mainly two types of methods, i.e. feature extraction and waveband selection for the dimensionality reduction of EO-1 hyperion.Feature extraction is built upon each light
Wave band is compressed by mathematic(al) manipulation on spectrum wave band, realizes that initial data projects to lower dimensional space, such as principal component from higher dimensional space
It analyzes PCA, independent principal component analysis ICA, linear discriminant analysis LDA, minimal noise and detaches MNF.Feature extracting method shortens
The dimensionality reduction time, but original image is converted, change the physical significance of initial data.In contrast, waveband selection is
The band subset that can represent primary data information (pdi) is selected in the wave band all from high spectrum image, height can not only be substantially reduced
The data dimension of spectrum picture, and information can be used than more completely remaining with.
Band selection method is largely divided into two classes:One kind is the band selection method based on information content, such as maximum entropy method (MEM), most
Good index method OFI, auto-subspace partition ASP, adaptive band selection, these methods are all to select individually to be rich in information content
Wave band and combination, however the group credit union of these high information quantity wave bands, there are bulk redundancy information, classification performance is poor.It is another kind of to be
Waveband selection based on cluster, such as method based on spectral clustering and the inter-class separability factor, based on dichotomy and Dynamic Programming
The method of global optimum's cluster and the method etc. combined based on genetic algorithm and ant colony algorithm.Since these methods are to be based on wave band
Between similitude select the representative wave band of every class to form band subset, and have ignored the correlation of wave band between every class cluster
And the information content of wave band, thus leading to the wave band poor performance selected, image classification accuracy is low.
Invention content
It is an object of the invention to the deficiencies for above-mentioned prior art, propose a kind of based on textural characteristics and affine propagation
The hyperspectral image classification method of clustering algorithm reduces the shadow to selecting wave band to consider band class information amount and correlation
It rings, improves image classification accuracy.
The technical thought of the present invention is to extract wave band texture feature vector according to image texture properties, utilize wave band texture
Feature vector obtains similar matrix;All wave bands of high spectrum image are clustered using reflection propagation algorithm, after choosing cluster
Every group of optimal wave band forms new band subset;Classified to true atural object on image using support vector machine classifier.Its
Implementation includes as follows:
(1) the EO-1 hyperion raw image data that size is m × n × l is read, it is m which, which is converted to l size,
The matrix form of × n, the wave band quantity for obtaining the high-spectral data are l, and wave band size is the 2-D data of m × n;
(2) texture feature extraction is carried out to each band image, and similitude square is calculated according to wave band texture feature vector
Battle array s;
(3) wave band is clustered using similarity matrix as the input of AP algorithms, the wave band for obtaining high spectrum image is poly-
Class result;
(4) the optimal wave band per group cluster result medium wave band subset is calculated:
(4a) calculates the standard deviation sigma of each wave band data;
The multiband matrix that size is m × n is converted to the one-dimensional matrix of t × 1, wherein t=m × n by (4b);
(4c) calculates in every group of band subset the correlation R between wave band two-by-two, and every according to correlation calculations between wave band
The sum of a wave band and the related coefficient absolute value of remaining wave band in group H;
(4e) calculates the wave band coefficient p in every group of band subset according to the sum of wave band standard deviation sigma and wave band related coefficient H;
(4f) chooses wave band generation of the maximum wave band of wave band coefficient as these wave bands in min~max wavelength bands
Table, wherein min=b-2d, max=b+2d, b are cluster centre, and d is the wave band number of every class after cluster;
(5) calibration image Q is read in, training set and test set sample coordinate and corresponding label per class atural object are obtained;
(6) training set of true atural object and test during every group of wave band represents are obtained according to training set and test set sample coordinate
Collect sample;
(7) by the training label input support vector machines training function of training set and corresponding atural object, training pattern is obtained;
(8) test label and training pattern of test set and corresponding atural object are input to the test function of support vector machines
In, test sample type is predicted, the classification results of high spectrum image are obtained.
The present invention has the following advantages that compared with the conventional method:
First, the present invention clusters wave band due to introducing reflection propagation algorithm, compared to having clustering method, improves
The stability of cluster result.
Second, the present invention after cluster due to introducing optimal band index method, compared to waveband selection after existing cluster
Method can select the band subset of informative and low correlation, improve image classification accuracy.
Description of the drawings
Fig. 1 is the implementation flow chart of the present invention;
Fig. 2 is the wave band coefficient line chart of a group cluster result medium wave band subset;
Fig. 3 is the true atural object distribution map of Pavia universities that the present invention uses;
Fig. 4 is the Pavia universities classification hyperspectral imagery result figure with the method for the present invention.
Fig. 5 is the Pavia universities classification hyperspectral imagery result figure with the hyperspectral image classification method based on MVPCA.
Fig. 6 is the Pavia universities classification hyperspectral imagery result figure with the hyperspectral image classification method based on ABS.
Specific implementation mode
The embodiment of the present invention and effect are described in further detail below in conjunction with the accompanying drawings.
Referring to Fig.1, steps are as follows for realization of the invention:
Step 1, hyperspectral image data is read.
Selected high-spectrum seems the North of Italy Pavia universities region obtained by ROSIS sensors in this example
Airborne Hyperspectral remote sensing images, it is 610 × 340 × 103 to test the image data size that uses, wave-length coverage be 430nm~
860nm, spectral resolution are 4nm~12nm, spatial resolution 1.3m, and it includes the true atural objects for having 9 classes different;
Hyperspectral image data is converted into the band image that 103 sizes are 610 × 340.
Step 2, similarity matrix is calculated according to wave band texture feature vector.
2a) calculate the texture feature vector of wave band:
Common texture characteristic extracting method has Gabor transformation, wavelet transformation, markov random file and gray scale symbiosis square
Battle array etc..This example using gray level co-occurrence matrixes to each band image carry out texture feature extraction, that is, calculate band image at 0 °,
45 °, 90 °, contrast, entropy, correlation and energy this four textural characteristics on 135 ° of four directions;
Similarity matrix s 2b) is calculated according to wave band texture feature vector:
Each element of similarity matrix s is to take the Euclidean distance of negative value to obtain by texture feature vector between wave band, table
Show as follows:
Wherein s (i, j) indicates i-th of wave band feature vector xiWith j-th of wave band feature vector xjSimilitude, formula is such as
Under:
S (i, j)=- | | xi-xj||2
As i=j, s (i, j)=- 1.
Step 3, the wave band cluster result of high spectrum image is obtained.
Wave band is clustered using similarity matrix s as the input of AP algorithms, AP algorithm cores are between sample point
It is alternately transmitted there are two types of message, i.e. Attraction Degree r (i, k) and degree of membership a (i, k), update rule is as follows:
The update rule of Attraction Degree:
r(i,k)←s(i,k)-max{a(i,k')+s(i,k')}(k'≠k)
Wherein, Attraction Degree r (i, k) describe sample object k be suitable as sample object i cluster centre degree, a (i,
K' degree of membership of the i-node to node k') is indicated;
If when i=k,
r(k,k)←s(i,k)-max{s(i,k')}(k≠k');
Degree of membership update rule:
Wherein, degree of membership a (i, k) describes the approval journey that sample object i selects sample object k as its cluster centre
Degree;
If when i=k, being from degree of membership update rule:
It is divided into 10 classes after this example medium wave band cluster.
Step 4, the optimal wave band in every group of band subset is obtained.
4a) calculate the standard deviation sigma of each wave band data:
Wherein m, n are the row, column pixel number of band image respectively, and f (x, y) indicates that coordinate is the picture of (x, y) point on wave band
Element value,It is the pixel average of wave band;
The multiband matrix that size is m × n 4b) is converted to the one-dimensional matrix of t × 1, wherein t=m × n;
4c) calculate in every group of band subset the correlation R between wave band two-by-two:
Wherein, fikIndicate k-th of pixel value of i-th of wave band, fjkIndicate k-th of pixel value of j-th of wave band;It is
The pixel average of i wave band,It is the pixel average of j-th of wave band;E represents the total number of pixel in a wave band;
4d) calculate the sum of each wave band and the related coefficient absolute value of remaining wave band in group H:
Wherein, d is the wave band number of every class after cluster, related coefficients of the R between wave band two-by-two;
4e) calculate the wave band coefficient p in every group of band subset:
Wherein, σ indicates that the standard deviation of wave band, H are the sum of the related coefficient absolute value of wave band;
4f) in min~max wavelength bands, the wave band for choosing the maximum wave band of wave band coefficient as these wave bands represents,
Wherein, min=b-2d, max=b+2d, b are cluster centre, and d is the wave band number of every class after cluster;
The wave band coefficient that cluster result is obtained by calculating the wave band coefficient of cluster result medium wave band subset in this example is rolled over
Line chart, such as Fig. 2, wherein Fig. 2 (a) are the wave band coefficient line chart of one group of band subset, and the cluster centre of this group of band subset is
51st band image, and include 11 wave bands, Fig. 2 (b) is wave band of this group of band subset in min~max wavelength bands
Coefficient line chart.
Step 5, training set and test set label are obtained.
5a) read in calibration image Q:
It is 610 × 340 calibration image Q that this example, which uses size, includes the true atural object of 9 classes and corresponding classification mark in image
Label, as shown in Figure 3.
10% 5b) is randomly selected from every class object coordinates and is used as training set sample coordinate, and remaining coordinate is as test set
Sample coordinate;
Corresponding training set label 5c) is obtained according to training set sample coordinate, is corresponded to according to test set sample coordinate
Test set label.
Step 6, the training set and test set sample per class atural object are obtained.
Pixel value 6a) chosen in every group of wave band represents on training set sample coordinate forms training set sample;
Pixel value 6b) chosen in every group of wave band represents on test set sample coordinate forms test set sample;
6c) training set that every class atural object obtains in the representative of all wave bands is merged, obtains the training of such atural object
Collect sample;
6d) test set that every class atural object obtains in the representative of all wave bands is merged, obtains the test of such atural object
Collect sample.
Step 7, training pattern is obtained.
It will be trained in the training label input support vector machines training function svmtrain of training set and corresponding atural object,
Obtain training pattern.
Step 8, the classification results of high spectrum image are obtained.
The test label and training pattern of test set and corresponding atural object are input to the included test letter of support vector machines
In number svmpredict, test sample type is predicted, the classification results of high spectrum image are obtained.
The effect of the present invention can be described further by following experiment.
1. emulation experiment condition:
The hardware test platform of emulation experiment of the present invention is:I3-3240U CPU, dominant frequency 3.4Ghz, memory 4GB;Software is flat
Platform is:7 Ultimates of Windows, 64 bit manipulation system, Matlab R2016a.Data set:Using by ROSIS-03 sensors system
The Airborne Hyperspectral remote sensing images in the North of Italy Pavia universities region that system obtains, wave band size are 610 × 340, wave band
Number is 103.
Similarity matrix is calculated using texture feature vector, and is clustered by AP algorithms, to every group of wave band after cluster
Optimal waveband selection is carried out, result is as shown in table 1 below.
1 cluster result of table and optimal wave band
By calibration image, the training set and test set sample coordinate and corresponding label of the every class atural object of acquisition, and according to
The training set and test set sample of the true atural object of optimal wave band in training set and test set sample coordinate extraction table 1.
Emulation 1, is trained the SVM classifier based on RBF cores using the content of table 1 and the training set of extraction, then will
Test set sample is input in trained svm classifier, is divided Pavia universities high spectrum image using the method for the present invention
Class obtains image classification as a result, as shown in Figure 4.
Emulation 2, using it is existing based on the hyperspectral image classification method of MVPCA to Pavia universities high spectrum image into
Row classification, it is as shown in Figure 5 to obtain classification results.
Emulation 3 carries out Pavia universities high spectrum image based on the hyperspectral image classification method of ABS using existing
Classification, it is as shown in Figure 6 to obtain classification results.
Comparison diagram 4, Fig. 5 and Fig. 6 can be seen that the classifying quality of the method for the present invention is better than existing two methods.With overall point
Class precision OA and Kappa coefficient evaluates the classifying quality of above-mentioned three kinds of methods, as a result such as table 2
2 classification results of table are evaluated
OA/% | Kappa | |
MVPCA | 82.18 | 0.76 |
ABS | 85.95 | 0.81 |
The method of the present invention | 89.50 | 0.86 |
The nicety of grading of the method for the present invention is better than existing two methods as can be seen from Table 2.
In conclusion from the point of view of no matter from classification results figure or from the overall classification accuracy and Kapper coefficient values the case where,
The hyperspectral classification method based on textural characteristics and affine propagation clustering algorithm waveband selection of the present invention, with existing two kinds of sides
Method, which is compared, has better classifying quality.
The above content is a further detailed description of the present invention in conjunction with specific preferred embodiments, and it cannot be said that
The specific implementation of the present invention is confined to these explanations.For those of ordinary skill in the art to which the present invention belongs, exist
Under the premise of not departing from present inventive concept, a number of simple deductions or replacements can also be made, all shall be regarded as belonging to the present invention's
Protection domain.
Claims (6)
1. a kind of hyperspectral image classification method based on textural characteristics and affine propagation clustering algorithm, including:
(1) the EO-1 hyperion raw image data that size is m × n × l is read, it is m × n which, which is converted to l size,
Matrix form, obtain the high-spectral data wave band quantity be l, wave band size be m × n 2-D data;
(2) texture feature extraction is carried out to each band image, and similarity matrix s is calculated according to wave band texture feature vector;
(3) wave band is clustered using similarity matrix as the input of AP algorithms, obtains the wave band cluster knot of high spectrum image
Fruit;
(4) the optimal wave band per group cluster result medium wave band subset is calculated:
(4a) calculates the standard deviation sigma of each wave band data;
The multiband matrix that size is m × n is converted to the one-dimensional matrix of t × 1, wherein t=m × n by (4b);
(4c) calculates in every group of band subset the correlation R between wave band two-by-two, and according to each wave band of correlation calculations between wave band
With the sum of related coefficient absolute value of remaining wave band H in group;
(4e) calculates the wave band coefficient p in every group of band subset according to the sum of wave band standard deviation sigma and wave band related coefficient H;
(4f) in min~max wavelength bands, the wave band for choosing the maximum wave band of wave band coefficient as these wave bands represents,
In, min=b-2d, max=b+2d, b are cluster centre, and d is the wave band number of every class after cluster;
(5) calibration image Q is read in, training set and test set sample coordinate and corresponding label per class atural object are obtained;
(6) training set and test set sample of true atural object during every group of wave band represents are obtained according to training set and test set sample coordinate
This;
(7) by the training label input support vector machines training function of training set and corresponding atural object, training pattern is obtained;
(8) test label and training pattern of test set and corresponding atural object are input in the test function of support vector machines,
Test sample type is predicted, the classification results of high spectrum image are obtained.
2. according to the method described in claim 1, the similarity matrix s wherein in step (2), indicates as follows:
Wherein s (i, j) indicates i-th of wave band feature vector xiWith j-th of wave band feature vector xjSimilitude, formula is as follows:
S (i, j)=- | | xi-xj||2
As i=j, s (i, j)=- 1.
3. according to the method described in claim 1, the standard deviation sigma of each wave band data is wherein calculated in step (4a), by such as
Lower formula carries out:
Wherein m, n are the row, column pixel number of band image respectively, and f (x, y) indicates the pixel value that coordinate is put for (x, y) on wave band,It is the pixel average of wave band.
4. according to the method described in claim 1, calculating in every group of band subset the phase between wave band two-by-two wherein in step (4c)
Closing property R is calculated by following formula:
Wherein, fikIndicate k-th of pixel value of i-th of wave band, fjkIndicate k-th of pixel value of j-th of wave band;It is i-th
The pixel average of wave band,It is the pixel average of j-th of wave band;E represents the total number of pixel in a wave band.
5. according to the method described in claim 1, it is related to remaining interior wave band of group wherein to calculate each wave band in step (4c)
The sum of absolute coefficient H is calculated by following formula:
Wherein, d is the wave band number of every class after cluster, related coefficients of the R between wave band two-by-two.
6. according to the method described in claim 1, wherein calculating the wave band coefficient p in every group of band subset in step (4e), lead to
Following formula is crossed to calculate:
Wherein, σ indicates that the standard deviation of wave band, H are the sum of the related coefficient absolute value of wave band.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810074087.5A CN108446582A (en) | 2018-01-25 | 2018-01-25 | Hyperspectral image classification method based on textural characteristics and affine propagation clustering algorithm |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810074087.5A CN108446582A (en) | 2018-01-25 | 2018-01-25 | Hyperspectral image classification method based on textural characteristics and affine propagation clustering algorithm |
Publications (1)
Publication Number | Publication Date |
---|---|
CN108446582A true CN108446582A (en) | 2018-08-24 |
Family
ID=63190944
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810074087.5A Pending CN108446582A (en) | 2018-01-25 | 2018-01-25 | Hyperspectral image classification method based on textural characteristics and affine propagation clustering algorithm |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108446582A (en) |
Cited By (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111008970A (en) * | 2019-12-11 | 2020-04-14 | 江南大学 | Hyperspectral image acquisition system-based method for detecting impurities in preserved vegetables |
CN111881933A (en) * | 2019-06-29 | 2020-11-03 | 浙江大学 | Hyperspectral image classification method and system |
CN112232402A (en) * | 2020-10-13 | 2021-01-15 | 燕山大学 | Method for realizing hyperspectral camouflage target recognition based on improved OIF and SVM |
CN112906750A (en) * | 2021-01-25 | 2021-06-04 | 浙江大学 | Hyperspectral image-based material analysis method and system |
CN113205143A (en) * | 2021-05-10 | 2021-08-03 | 郑州轻工业大学 | Multi-scale superpixel hyperspectral remote sensing image classification method based on space-spectrum coupling characteristics |
CN113516025A (en) * | 2021-04-26 | 2021-10-19 | 山东深蓝智谱数字科技有限公司 | Hyperspectral image processing method, device and medium based on model construction |
CN113658069A (en) * | 2021-08-13 | 2021-11-16 | 哈尔滨工业大学 | Hyperspectral microscopic image flat field correction method and system based on common flat field extraction |
CN114863291A (en) * | 2022-04-20 | 2022-08-05 | 重庆市地理信息和遥感应用中心 | Hyperspectral image band selection method based on MCL and spectral difference measurement |
CN115641412A (en) * | 2022-05-13 | 2023-01-24 | 泰瑞数创科技(北京)股份有限公司 | Hyperspectral data-based three-dimensional semantic map generation method |
CN115713694A (en) * | 2023-01-06 | 2023-02-24 | 东营国图信息科技有限公司 | Land surveying and mapping information management method |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103440505A (en) * | 2013-09-16 | 2013-12-11 | 重庆邮电大学 | Spatial neighborhood information weighted hyper-spectral remote sensing image classification method |
CN104102839A (en) * | 2013-11-16 | 2014-10-15 | 西安电子科技大学 | Alzheimer's disease cortex auto-classification method based on multi-scale mesh surface form features |
CN105913078A (en) * | 2016-04-07 | 2016-08-31 | 江南大学 | Multi-mode soft measurement method for improving adaptive affine propagation clustering |
CN105989592A (en) * | 2015-02-11 | 2016-10-05 | 中国科学院西安光学精密机械研究所 | Hyperspectral image waveband selection method based on biclustering and neighborhood analysis |
CN107527001A (en) * | 2017-03-29 | 2017-12-29 | 广东交通职业技术学院 | A kind of hyperspectral image classification method based on Steerable filter and linear space correlation information |
-
2018
- 2018-01-25 CN CN201810074087.5A patent/CN108446582A/en active Pending
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103440505A (en) * | 2013-09-16 | 2013-12-11 | 重庆邮电大学 | Spatial neighborhood information weighted hyper-spectral remote sensing image classification method |
CN104102839A (en) * | 2013-11-16 | 2014-10-15 | 西安电子科技大学 | Alzheimer's disease cortex auto-classification method based on multi-scale mesh surface form features |
CN105989592A (en) * | 2015-02-11 | 2016-10-05 | 中国科学院西安光学精密机械研究所 | Hyperspectral image waveband selection method based on biclustering and neighborhood analysis |
CN105913078A (en) * | 2016-04-07 | 2016-08-31 | 江南大学 | Multi-mode soft measurement method for improving adaptive affine propagation clustering |
CN107527001A (en) * | 2017-03-29 | 2017-12-29 | 广东交通职业技术学院 | A kind of hyperspectral image classification method based on Steerable filter and linear space correlation information |
Non-Patent Citations (5)
Title |
---|
刘春红等: "一种新的高光谱遥感图像降维方法", 《中图象图形学报》 * |
张建朋等: "基于仿射聚类的主动SVM多类分类方法", 《计算机应用研究》 * |
施蓓琦等: "应用稀疏非负矩阵分解聚类实现高光谱影像波段的优化选择", 《测绘学报》 * |
王立国等: "基于最大最小距离的高光谱遥感图像波段选择", 《智能系统学报》 * |
苏红军: "高光谱影像光谱——纹理特征提取与多分类器集成技术研究", 《中国优秀博士学位论文全文数据库》 * |
Cited By (17)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111881933A (en) * | 2019-06-29 | 2020-11-03 | 浙江大学 | Hyperspectral image classification method and system |
CN111881933B (en) * | 2019-06-29 | 2024-04-09 | 浙江大学 | Hyperspectral image classification method and system |
CN111008970A (en) * | 2019-12-11 | 2020-04-14 | 江南大学 | Hyperspectral image acquisition system-based method for detecting impurities in preserved vegetables |
CN112232402B (en) * | 2020-10-13 | 2022-07-08 | 燕山大学 | Method for realizing hyperspectral camouflage target recognition based on improved OIF and SVM |
CN112232402A (en) * | 2020-10-13 | 2021-01-15 | 燕山大学 | Method for realizing hyperspectral camouflage target recognition based on improved OIF and SVM |
CN112906750A (en) * | 2021-01-25 | 2021-06-04 | 浙江大学 | Hyperspectral image-based material analysis method and system |
CN113516025A (en) * | 2021-04-26 | 2021-10-19 | 山东深蓝智谱数字科技有限公司 | Hyperspectral image processing method, device and medium based on model construction |
CN113516025B (en) * | 2021-04-26 | 2022-03-08 | 山东深蓝智谱数字科技有限公司 | Hyperspectral image processing method, device and medium based on model construction |
CN113205143A (en) * | 2021-05-10 | 2021-08-03 | 郑州轻工业大学 | Multi-scale superpixel hyperspectral remote sensing image classification method based on space-spectrum coupling characteristics |
CN113658069B (en) * | 2021-08-13 | 2024-04-09 | 哈尔滨工业大学 | Hyperspectral microscopic image flat field correction method and system based on shared flat field extraction |
CN113658069A (en) * | 2021-08-13 | 2021-11-16 | 哈尔滨工业大学 | Hyperspectral microscopic image flat field correction method and system based on common flat field extraction |
CN114863291A (en) * | 2022-04-20 | 2022-08-05 | 重庆市地理信息和遥感应用中心 | Hyperspectral image band selection method based on MCL and spectral difference measurement |
CN114863291B (en) * | 2022-04-20 | 2023-08-08 | 重庆市地理信息和遥感应用中心 | Hyperspectral image band selection method based on MCL and spectrum difference measurement |
CN115641412A (en) * | 2022-05-13 | 2023-01-24 | 泰瑞数创科技(北京)股份有限公司 | Hyperspectral data-based three-dimensional semantic map generation method |
CN115641412B (en) * | 2022-05-13 | 2023-08-04 | 泰瑞数创科技(北京)股份有限公司 | Three-dimensional semantic map generation method based on hyperspectral data |
CN115713694A (en) * | 2023-01-06 | 2023-02-24 | 东营国图信息科技有限公司 | Land surveying and mapping information management method |
CN115713694B (en) * | 2023-01-06 | 2023-05-12 | 东营国图信息科技有限公司 | Land mapping information management method |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108446582A (en) | Hyperspectral image classification method based on textural characteristics and affine propagation clustering algorithm | |
Tirandaz et al. | PolSAR image segmentation based on feature extraction and data compression using weighted neighborhood filter bank and hidden Markov random field-expectation maximization | |
CN106815601B (en) | Hyperspectral image classification method based on recurrent neural network | |
CN103971123B (en) | Hyperspectral image classification method based on linear regression Fisher discrimination dictionary learning (LRFDDL) | |
Al Bashish et al. | A framework for detection and classification of plant leaf and stem diseases | |
CN107358260B (en) | Multispectral image classification method based on surface wave CNN | |
CN106503739A (en) | The target in hyperspectral remotely sensed image svm classifier method and system of combined spectral and textural characteristics | |
Mughees et al. | Multiple deep-belief-network-based spectral-spatial classification of hyperspectral images | |
CN107563442B (en) | Hyperspectral image classification method based on sparse low-rank regular graph tensor embedding | |
Shen et al. | Shape recognition by bag of skeleton-associated contour parts | |
CN103208011B (en) | Based on average drifting and the hyperspectral image space-spectral domain classification method organizing sparse coding | |
CN108427913B (en) | Hyperspectral image classification method combining spectral, spatial and hierarchical structure information | |
CN110472682A (en) | A kind of Hyperspectral Remote Sensing Imagery Classification method for taking space and local feature into account | |
CN105718942B (en) | High spectrum image imbalance classification method based on average drifting and over-sampling | |
CN105913092B (en) | Figure canonical hyperspectral image band selection method based on sub-space learning | |
CN107895139A (en) | A kind of SAR image target recognition method based on multi-feature fusion | |
CN110533077A (en) | Form adaptive convolution deep neural network method for classification hyperspectral imagery | |
CN105184314B (en) | Wrapper formula EO-1 hyperion band selection methods based on pixel cluster | |
CN108154094A (en) | The non-supervisory band selection method of high spectrum image divided based on subinterval | |
CN110276746B (en) | Robust remote sensing image change detection method | |
CN104809471B (en) | A kind of high spectrum image residual error integrated classification method based on spatial spectral information | |
CN109697465A (en) | Hyperspectral Image Classification method based on the Fisher discriminant analysis of Wavelet Kernel part | |
CN109359525A (en) | The Classification of Polarimetric SAR Image method of differentiation spectral clustering based on sparse low-rank | |
CN109034213B (en) | Hyperspectral image classification method and system based on correlation entropy principle | |
CN107392140A (en) | Polarization SAR terrain classification method with incremental learning is decomposed based on low-rank sparse |
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 | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20180824 |