CN106503739A - The target in hyperspectral remotely sensed image svm classifier method and system of combined spectral and textural characteristics - Google Patents

The target in hyperspectral remotely sensed image svm classifier method and system of combined spectral and textural characteristics Download PDF

Info

Publication number
CN106503739A
CN106503739A CN201610929442.3A CN201610929442A CN106503739A CN 106503739 A CN106503739 A CN 106503739A CN 201610929442 A CN201610929442 A CN 201610929442A CN 106503739 A CN106503739 A CN 106503739A
Authority
CN
China
Prior art keywords
image
principal component
filtering
spectrum
spectral
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
Application number
CN201610929442.3A
Other languages
Chinese (zh)
Inventor
王毅
张琰
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
China University of Geosciences
Original Assignee
China University of Geosciences
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by China University of Geosciences filed Critical China University of Geosciences
Priority to CN201610929442.3A priority Critical patent/CN106503739A/en
Publication of CN106503739A publication Critical patent/CN106503739A/en
Pending legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • 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

Abstract

The invention discloses the target in hyperspectral remotely sensed image svm classifier method and system of a kind of combined spectral and textural characteristics, the method is comprised the following steps：S1, input original Hyperspectral imaging to be sorted and ground investigation set of data samples；S2, the pixel for extracting respective coordinates position in original Hyperspectral imaging, constitute reference data sample set；S3, training sample set is not randomly selected for various places species；S4, principal component transform is carried out, extract first principal component image；S5, acquisition region segmentation figure；S6, acquisition filtering image；S7, spectral signature information and the texture feature information counted in each cut zone；S8, solution supporting vector machine model；S9, original Hyperspectral imaging is classified, obtain the Hyperspectral imaging that classifies；S10, output category image.The invention provides the New Policy of combined spectral and textural characteristics, can effectively improve Hyperspectral Image Classification precision.

Description

The target in hyperspectral remotely sensed image svm classifier method and system of combined spectral and textural characteristics
Technical field
A kind of the present invention relates to technical field of image processing, more particularly to the high-spectrum remote-sensing of combined spectral and textural characteristics Image svm classifier method and system.
Background technology
Compared with multi-spectral remote sensing image, target in hyperspectral remotely sensed image has the spectrum and spatial information for more enriching, these Information can accurately reflect the attribute difference between differently species are other, and realize that atural object is accurately extracted and recognized, be more high-precision The target in hyperspectral remotely sensed image analysis of degree establishes good basis with application.However, Hyperspectral imaging dimension is high, wave band dependency is big, There is the characteristics of image such as the nonlinear characteristic of noise and uniqueness, huge choosing is brought to target in hyperspectral remotely sensed image analysis with process War.Traditional Hyperspectral Remote Sensing Imagery Classification method generally carries out terrain classification merely with pixel spectral signature, and does not consider Contained abundant spatial information, such as spatial structural form, location of pixels and range information etc. in image.These research methoies are obtained The nicety of grading for obtaining has arrived at bottleneck, is difficult to continue to improve.
Compared with Hyperspectral Image Classification method of the tradition based on pixel, in the Spectral Properties for combining Hyperspectral imaging itself Seeking peace, (texture information, spatial structural form, atural object dimension information, atural object profile information, space including image divides spatial information Cloth information etc.) on the basis of, the Hyperspectral Image Classification method of combined spectral and spatial information can further improve image Nicety of grading, obtains comprising the more accurate image classification figure of homogeneous region, meets the needs of cartographic production.Combined spectral and space Information carries out Hyperspectral Remote Sensing Imagery Classification can：(1) the classification noise in spiced salt distribution in classification chart is effectively reduced；(2) take off Show Pixel domain structure and shape facility；(3) identification species difference not spatially in different land use type in the same manner.
Content of the invention
The technical problem to be solved in the present invention is for not considering contained abundant space in image in prior art Information, and the low defect of nicety of grading, there is provided a kind of introducing spectrum statistics with histogram amount is characterizing between differently species are other Textural characteristics difference, and the support vector machine classifier based on complex nucleus is set up, spectral information is organically combined with textural characteristics Get up, improve the target in hyperspectral remotely sensed image svm classifier of the combined spectral and textural characteristics of classification results homogeneity degree and nicety of grading Method and system.
The technical solution adopted for the present invention to solve the technical problems is：
The present invention provides a kind of target in hyperspectral remotely sensed image svm classifier method of combined spectral and textural characteristics, including following Step：
S1, input original Hyperspectral imaging to be sorted, and the image is normalized；Input with to be sorted The corresponding ground investigation set of data samples of Hyperspectral imaging；
S2, the coordinate position for obtaining all samples of ground investigation data sample concentration, it is right in original Hyperspectral imaging to extract The pixel of coordinate position is answered, reference data sample set is constituted；
Comprising multiple atural object classifications in S3, reference data sample set, it is followed successively by each atural object classification and randomly selects a fixed number Training sample set of the reference data sample of amount as supervised classification；
S4, principal component transform is carried out to original Hyperspectral imaging, and extract first principal component image；
S5, the super-pixel segmentation based on entropy rate is carried out to first principal component image, obtain region segmentation figure；
S6, first principal component image is carried out respectively intensity filtering, Gauss-Laplace filtering and Gabor Filtering Processing, Obtain filtering image；
S7, according to filtering image, the spectrum rectangular histogram in statistical regions segmentation figure in each cut zone obtains each point Cut spectral signature information and the texture feature information in region；
S8, the spectral signature information in corresponding for training sample set cut zone and texture feature information are substituted into simultaneously multiple Synkaryon function, solves supporting vector machine model；
S9, the supporting vector machine model according to compound kernel function, classify to original Hyperspectral imaging, obtain classification Hyperspectral imaging；
S10, output category image.
Further, carry out principal component transform to high spectrum image in step S4 of the invention to concretely comprise the following steps：
S41, the covariance matrix for generating data in original spectrum coordinate；
S42, the eigenvalue for obtaining the covariance matrix and characteristic vector；
S43, arrayed feature value, find first, second and subsequent main constituent coordinate axess；
S44, the gray value for using the characteristic vector next life pixel in following formula new in each principal component；
Y=GX
Wherein, the spectral space before X and Y represent conversion respectively and after conversion, according to the mathematical principle of principal component transform, G It is the transposed matrix of the eigenvectors matrix of the covariance matrix in X spaces；
First principal component under S45, acquisition spectral space Y.
Further, concretely comprised the following steps based on the super-pixel segmentation of entropy rate in step S5 of the invention：
S51, initialization region number determination；
S52, calculating data item H (A) are used for obtaining homogeneous cluster cluster, and H (A) is made up of a series of functions：
Wherein, one group side collection of the A for non-directed graph GWherein G=(V, E), vertex set V represent all in image Pixel, E represent the side collection on connection summit；αi=wi/wT, wiRepresent connection i-th summit side weight and,Represent Normaliztion constant, | v | represent summit sum, p in figurei,jRepresent transition probability；
S53, calculated equilibrium item B (A), its formula is：
Wherein, zARepresent the distribution of clustering cluster；
S54, the super-pixel segmentation based on entropy rate is carried out to first principal component image, its computing formula is：
Wherein, λ represents the weight size of object function data item and balance term, λ>0.
Further, be filtered in step S6 of the invention concretely comprises the following steps：
S61, intensity filtering is carried out to first principal component image, its computing formula is as follows：
S62, initialization Gauss-Laplace filtering gaussian kernel standard deviation；Gauss-La Pu is carried out to first principal component image Lars filters, and its computing formula is as follows：
Wherein, σLoGIt is LoG wave filter gaussian kernel standard deviations；
S63, initialization filtering direction and Gabor filter gaussian kernel standard deviation；Gabor is carried out to first principal component image Filtering, its computing formula are as follows：
Wherein, θ represents filtering direction, σGaborIt is Gabor filter gaussian kernel standard deviation, wherein ratio σGabor/ λ is set to 0.5.
Further, in step S7 of the invention statistics spectrum is histogrammic concretely comprises the following steps：
S71, for filtering image in cut zone W, be calculated one group of wave filter { F(α), α=1,2 ..., M }, wherein M is the total number of wave filter；
S72, the wave filter of acquisition and first principal component image are carried out convolutional calculation respectively, obtain a group image { W(1),W(2),...,W(M)}；
S73, the rectangular histogram for calculating sub-image W (α), formula is：
Wherein, t1And t2Represent statistics with histogram scope bound；
S74, according to formula：The spectrum of one group of wave filter to giving is straight Square figure is calculated.
Further, the algebraical sum that kernel function is two gaussian radial basis functions is combined in step S8 of the invention, two Individual kernel function corresponds to Hyperspectral imaging spectral signature respectively and extracts the texture feature information for obtaining, the weight proportion of spectrum core For μ, the weight proportion of textural characteristics core is 1- μ, and the span of μ is [0,1].
Further, the gaussian radial basis function in step S8 of the invention is：
Wherein,The compound kernel function of combined spectral and textural characteristics is represented,WithDifference table Show spectrum kernel function and texture kernel function；X and y represent two pixels in image space, x respectivelyspectAnd yspectRepresent respectively Two spectral signature vectors of spectral space；xtextAnd ytextTexture space two texture feature vectors are represented respectively；μ(0<μ<1) table Give instructions in reply the balance term of spectral information and texture information in synkaryon function.In formula,WithCan be calculated using following formula：
Wherein, σSTKRepresent Gaussian function standard deviation.
Further, original Hyperspectral imaging is classified according to supporting vector machine model in step S9 of the invention Method is specially：
Further, remaining reference data sample in each atural object classification is commented as precision in step S3 of the invention The test sample collection of valency.
The present invention provides the target in hyperspectral remotely sensed image svm classifier system of a kind of combined spectral and textural characteristics, including：
Sample acquisition unit, for being input into original Hyperspectral imaging to be sorted, and is normalized to the image； Input ground investigation set of data samples corresponding with Hyperspectral imaging to be sorted；Obtain ground investigation data sample and concentrate and own The coordinate position of sample, extracts the pixel of respective coordinates position in original Hyperspectral imaging, constitutes reference data sample set；With reference to Data sample is concentrated comprising multiple atural object classifications, is followed successively by each atural object classification and is randomly selected a number of reference data sample Training sample set as supervised classification；
Super-pixel segmentation and texture filtering unit, for carrying out principal component transform to original Hyperspectral imaging, and extract One main constituent image；The super-pixel segmentation based on entropy rate is carried out to first principal component image, obtains region segmentation figure；Lead to first Composition image carries out intensity filtering, Gauss-Laplace filtering and Gabor Filtering Processing respectively, obtains filtering image；
Spectrum histogram statistical unit, for according to filtering image, in statistical regions segmentation figure in each cut zone Spectrum rectangular histogram, obtains spectral signature information and the texture feature information in each cut zone；
Image classification unit, for by the spectral signature information in corresponding for training sample set cut zone and textural characteristics Information substitutes into complex nucleus function simultaneously, solves supporting vector machine model；According to the supporting vector machine model of compound kernel function, to original Beginning Hyperspectral imaging is classified, and obtains the Hyperspectral imaging that classifies.
The beneficial effect comprise that：The target in hyperspectral remotely sensed image SVM of the combined spectral and textural characteristics of the present invention Sorting technique, using the Hyperspectral Image Classification method of combined spectral and textural characteristics, overcomes traditional Hyperspectral Image Classification Method does not consider the problem of other characteristic informations of image merely with spectral signature；The present invention takes full advantage of Hyperspectral imaging institute Contain abundant spectral signature and structural texture feature, it is achieved that the sophisticated category of Hyperspectral imaging；Entropy rate is based on by adopting Super-pixel method splitting to first principal component image, overcome the shortcoming of over-segmentation and less divided in partitioning algorithm, So that the present invention has more accurately describes Local textural feature；By characterizing image using local spectrum rectangular histogram Textural characteristics, the present invention have the mistake point problem reduced inside the classification noise and homogeneous region of spiced salt shape, and which is classified As a result more homogeneous, more meet ground surface type actual distribution situation.
Description of the drawings
Below in conjunction with drawings and Examples, the invention will be further described, in accompanying drawing：
Fig. 1 is the stream of the target in hyperspectral remotely sensed image svm classifier method of the combined spectral and textural characteristics of the embodiment of the present invention Cheng Tu；
Fig. 2 (a) is the target in hyperspectral remotely sensed image svm classifier method of the combined spectral and textural characteristics of the embodiment of the present invention Classification results comparison diagram (a)；
Fig. 2 (b) is the target in hyperspectral remotely sensed image svm classifier method of the combined spectral and textural characteristics of the embodiment of the present invention Classification results comparison diagram (b)；
Fig. 2 (c) is the target in hyperspectral remotely sensed image svm classifier method of the combined spectral and textural characteristics of the embodiment of the present invention Classification results comparison diagram (c)；
Fig. 2 (d) is the target in hyperspectral remotely sensed image svm classifier method of the combined spectral and textural characteristics of the embodiment of the present invention Classification results comparison diagram (d).
Specific embodiment
In order that the objects, technical solutions and advantages of the present invention become more apparent, below in conjunction with drawings and Examples, right The present invention is further elaborated.It should be appreciated that specific embodiment described herein is only in order to explain the present invention, not For limiting the present invention.
As shown in figure 1, the target in hyperspectral remotely sensed image svm classifier side of the combined spectral and textural characteristics of the embodiment of the present invention Method, comprises the following steps：
S1, input original Hyperspectral imaging to be sorted, and the image is normalized；Input with to be sorted The corresponding ground investigation set of data samples of Hyperspectral imaging；
S2, the coordinate position for obtaining all samples of ground investigation data sample concentration, it is right in original Hyperspectral imaging to extract The pixel of coordinate position is answered, reference data sample set is constituted；
Comprising multiple atural object classifications in S3, reference data sample set, it is followed successively by each atural object classification and randomly selects a fixed number Training sample set of the reference data sample of amount as supervised classification；Remaining reference data sample in each atural object classification is made Test sample collection for precision evaluation；
S4, principal component transform (PCA) is carried out to original Hyperspectral imaging, and extract first principal component image；
Obtain first principal component image to concretely comprise the following steps：
S41, the covariance matrix for generating data in original spectrum coordinate；
S42, the eigenvalue for obtaining the covariance matrix and characteristic vector；
S43, arrayed feature value, find first, second and subsequent main constituent coordinate axess；
S44, the gray value for using the characteristic vector next life pixel in following formula new in each principal component；
Y=GX
Wherein, the spectral space before X and Y represent conversion respectively and after conversion.According to the mathematical principle of principal component transform, G It is the transposed matrix of the eigenvectors matrix of the covariance matrix in X spaces.
First principal component under S45, acquisition spectral space Y.
S5, the super-pixel segmentation based on entropy rate is carried out to first principal component image, obtain region segmentation figure；
Concretely comprised the following steps based on the super-pixel segmentation of entropy rate：
S51, initialization region number determination；
S52, calculating data item H (A) are used for obtaining homogeneous cluster cluster, and H (A) is made up of a series of functions：
Wherein, one group side collection of the A for non-directed graph GWherein G=(V, E), vertex set V represent all pictures in image Element, E represent the side collection on connection summit；αi=wi/wT, wiRepresent connection i-th summit side weight and,Expression is returned One changes constant, and | v | represents summit sum, p in figurei,jRepresent transition probability；
S53, calculated equilibrium item B (A), its formula is：
Wherein, zARepresent the distribution of clustering cluster；
S54, the super-pixel segmentation based on entropy rate is carried out to first principal component image, its computing formula is：
Wherein, λ represents the weight size of object function data item and balance term, λ>0.
S6, first principal component image is carried out respectively intensity filtering, Gauss-Laplace filtering and Gabor Filtering Processing, Obtain filtering image；
Be filtered concretely comprises the following steps：
S61, intensity filtering is carried out to first principal component image, its computing formula is as follows：
S62, initialization Gauss-Laplace filtering gaussian kernel standard deviation；Gauss-La Pu is carried out to first principal component image Lars filters, and its computing formula is as follows：
Wherein, σLoGIt is LoG wave filter gaussian kernel standard deviations；
S63, initialization filtering direction and Gabor filter gaussian kernel standard deviation；Gabor is carried out to first principal component image Filtering, its computing formula are as follows：
Wherein, θ represents filtering direction, σGaborIt is Gabor filter gaussian kernel standard deviation, wherein ratio σGabor/ λ is set to 0.5.
S7, according to filtering image, the spectrum rectangular histogram in statistical regions segmentation figure in each cut zone obtains each point Cut spectral signature information and the texture feature information in region；
Spectrum is histogrammic concretely comprises the following steps for statistics：
S71, for filtering image in cut zone W, be calculated one group of wave filter { F(α), α=1,2 ..., M }, wherein M is the total number of wave filter；
S72, the wave filter of acquisition and first principal component image are carried out convolutional calculation respectively, obtain a group image W (1),W(2),...,W(M)}；
S73, calculating sub-image W(α)Rectangular histogram, formula is：
Wherein, t1And t2Represent statistics with histogram scope bound；
S74, according to formula：The spectrum Nogata of one group of wave filter to giving Figure is calculated.
S8, the spectral signature information in corresponding for training sample set cut zone and texture feature information are substituted into simultaneously multiple Synkaryon function, solves supporting vector machine model；
Compound kernel function is the algebraical sum of two gaussian radial basis functions, and two kernel functions correspond to Hyperspectral imaging respectively Spectral signature and the texture feature information that obtains is extracted, the weight proportion of spectrum core is μ, the weight proportion of textural characteristics core is The span of 1- μ, μ is [0,1].
The gaussian radial basis function computing formula of combined spectral and textural characteristics is：
Wherein,The compound kernel function of combined spectral and textural characteristics is represented,Respectively Represent spectrum kernel function and texture kernel function；X and y represent two pixels in image space, x respectivelyspectAnd yspectDifference table Show two spectral signature vectors of spectral space；xtextAnd ytextTexture space two texture feature vectors are represented respectively；μ(0<μ<1) Represent the balance term of spectral information and texture information in compound kernel function.In formula,WithCan be calculated using following formula：
Wherein, σSTKRepresent Gaussian function standard deviation.
S9, the supporting vector machine model according to compound kernel function, classify to original Hyperspectral imaging, obtain classification Hyperspectral imaging；
S10, output category image.
In another specific embodiment of the present invention, the method is concretely comprised the following steps：
Step 1, input data.
In the present embodiment, the university of Pavia that a ROSIS-03 optical pickocff to be sorted is obtained is input into The target in hyperspectral remotely sensed image of University of Pavia with to should image handmarking's sample set.
Step 2, obtains reference data sample set：
According to the coordinate position of all samples of ground investigation data, respective coordinates position in EO-1 hyperion raw video is extracted Pixel constitutes reference data sample set；
Step 3, determines training sample and test samples collection：
In Hyperspectral imaging reference data sample set, it is followed successively by each classification interested and randomly selects 250 sample conducts The training sample set of image classification；Using the remaining sample of the corresponding category as image classification test sample collection；
Step 4, carries out principal component transform based on the Hyperspectral imaging to being input into, and extracts first principal component image as rear The continuous input picture for processing；
The first step, generates the covariance matrix of data in original spectrum coordinate；
Second step, obtains eigenvalue and the characteristic vector of the covariance matrix；
3rd step, arrayed feature value find first, second and subsequent main constituent coordinate axess；
4th step, with the new gray value in each principal component of the characteristic vector next life pixel in following formula；
Y=GX
Wherein, the spectral space before X and Y represent conversion respectively and after conversion.According to the mathematical principle of principal component transform, G It is the transposed matrix of the eigenvectors matrix of the covariance matrix in X spaces.
5th step, obtains first principal component under spectral space Y.
Step 5, carries out the super-pixel segmentation based on entropy rate to first principal component image；
The first step, initializes region number determination；
Second step, calculates data item H (A) to obtain homogeneous cluster cluster according to the following formula：
In formula, one group side collection of the A for non-directed graph GWherein, G=(V, E), vertex set V represent all in image Pixel, E represent the side collection on connection summit, αi=wi/wT, wiRepresent connection i-th summit side weight and,Represent Normaliztion constant, | v | represent summit sum, p in figurei,jRepresent transition probability.
3rd step, according to the following formula calculated equilibrium item B (A)：
In formula, zARepresent the distribution of clustering cluster.
4th step, carries out the super-pixel segmentation based on entropy rate, its computing formula to the first principal component image that PCA is extracted For：
In formula, λ (λ>0) the weight size of object function data item and balance term is represented.
Step 6, carries out texture filtering process to the first principal component image that PCA is extracted；
The first step, carries out intensity filtering to the first principal component image that PCA is extracted, and its computing formula is as follows：
Second step, initialization Gauss-Laplace filtering gaussian kernel standard deviation sigmaLoG=0.5 and σLoG=1；PCA is extracted First principal component image carry out Gauss-Laplace filtering, its computing formula is as follows：
3rd step, initialization filtering direction are θ=0 ° and θ=90 °, Gabor filter gaussian kernel standard deviation sigmaGabor=1； Gabor filtering is carried out to the first principal component image that PCA is extracted, its computing formula is as follows：
Wherein, θ represents filtering direction, σGaborIt is Gabor filter gaussian kernel standard deviation, wherein ratio σGabor/ λ is set to 0.5.
Step 7, counts the spectrum rectangular histogram of each cut zone：
The first step, for cut zone W in filter result image, is calculated one group of wave filter { F(α), α=1, 2 ..., M }, wherein M is the total number of wave filter.
The wave filter of acquisition and first principal component image are carried out convolutional calculation by second step respectively, obtain a group image { W(1),W(2),...,W(M)}.
3rd step, calculates the rectangular histogram of sub-image W (α) according to the following formula：
T in formula1And t2Represent statistics with histogram scope bound.The present invention carries out light to the result of each wave filter 11 sections are divided into during spectrum statistics with histogram.
4th step, the spectrum histogram calculation formula of corresponding one group of wave filter given herein above are as follows：
Spectrum rectangular histogram is normalized characteristic statisticses, and the result of calculation in different size of cut zone can be carried out Relatively.For each pixel, local spectrum rectangular histogram is exactly that the cut zone being located with the pixel is calculated, therefore same Belong to the pixel in a cut zone and there is identical local spectrum statistics with histogram amount, i.e. identical texture eigenvalue.
Step 8, builds compound kernel support vectors according to training sample
Corresponding for training sample set raw spectroscopic data and corresponding texture feature information are substituted into complex nucleus, is solved and is supported Vector.The complex nucleus for being used are the algebraical sum of two gaussian radial basis functions, and two kernel functions correspond to EO-1 hyperion shadow respectively As the texture feature information that spectral signature and extraction are obtained, the weight proportion of spectrum core is that value is μ=0.8.
Step 9, Hyperspectral Image Classification
According to complex nucleus supporting vector machine model, original Hyperspectral imaging is classified, obtain the EO-1 hyperion shadow that classifies Picture；
Step 10, output category result image
2 experimental example figure is described further to the effect of the present invention below in conjunction with the accompanying drawings.
1. emulation experiment condition：
The hardware test platform of this experiment is：Processor is Intel Duo i3, and dominant frequency is 2.4GHz, internal memory 4GB, software Platform is：8.1 operating systems of Windows, Microsoft Visual Studio 2013, Matlab R2012a.The present invention Input picture be university of the Pavia image data collection University of obtained by ROSIS-03 optical pickocffs Pavia.The main earth's surface of image is covered as various construction materials in city.Due to being aerial images, its spatial resolution is 1.3m, wave-length coverage are 0.43～0.86 μm, have 103 wave bands, and its image size is 610 340 pixels.Include nine in image Plant atural object：Asphalt (pitchy highway), Meadows (meadow), Gravel (broken gravel), Trees (forest land), Metal sheets (metallic plate), Bare soil (bare area), Bitumen (pitch roof), Bricks (fragment of brick), Shadows (shade).Image pane Formula is img.
2. emulation content：
Three prior art comparison-of-pair sorting method difference that the present invention is used are as follows, including classical SVM classifier and two kinds The Hyperspectral Image Classification method of conventional joint space and spectral information is as follows
Melgani et al. is in " Classification of hyperspectral remote sensing images with support vector machines.IEEE Transactions on Geoscience and Remote The hyperspectral image classification method proposed in Sensing.2004,42 (8), 1778-1790. ", abbreviation support vector machines point Class method.
Benediktsson et al. is in " Classification of Hyperspectral Data From Urban Areas Based on Extended Morphological Profiles.IEEE Transaction on Geoscience The combined spectral based on expanding morphology hatching proposed in and Remote Sensing.2005,43 (3), 480-491. " With spatial information Hyperspectral Image Classification method, abbreviation EMP sorting techniques.
Mathieu et al. is in " A spatial spectral kernel-based approach for the classification of remote-sensing images.Pattern Recognition.2012,45(1),381- 392. " combined spectral based on space-optical spectrum core proposed in and spatial information Hyperspectral imaging aerial image sorting technique, letter Claim SS-Kernel sorting techniques.
In experimentation, for the classification results that different classifications method is obtained, built according to the true reference data in ground Confusion matrix, and by calculating overall classification accuracy OA, each category classification precision CA, average nicety of grading AA and Kappa systems Number carries out quantitative assessment come the performance to the inventive method.
If classification number is n, it is the matrix of a n × n, wherein MijI-th class and actual measurement number in presentation class data type According to type jth apoplexy due to endogenous wind classified pixels number, then
First evaluation index is overall accuracy (OA), represents that the sample of correct classification accounts for the ratio of all samples, and value is bigger, Illustrate that classifying quality is better.Its computing formula is as follows：
Second evaluation index is classification precision (CA), represents the nicety of grading of each class, and value is bigger, and classifying quality is described Better.Its computing formula is as follows：
3rd evaluation index is mean accuracy (AA), represents the meansigma methodss of each class nicety of grading, is worth bigger, illustrates point Class effect is better.Its computing formula is as follows：
4th evaluation index is Kappa coefficient (Kappa), represents different weights in confusion matrix, is worth bigger, explanation Classifying quality is better.Its computing formula is as follows：
Wherein,It is the sum of all pixels for accuracy evaluation.
Fig. 2 is for the present invention in emulation experiment with prior art to university of high spectrum image Pavia University of The classification results comparison diagram of Pavia.Wherein, Fig. 2 (a) is for directly using SVM methods to university of high-spectrum remote sensing data Pavia The classification results figure that University of Pavia are obtained；Fig. 2 (b) is to high-spectrum remote sensing data Pavia using EMP methods The classification results figure that university University of Pavia are obtained；Fig. 2 (c) is to high-spectrum remote-sensing using SS-Kernel methods The classification results figure that university of data Pavia University of Pavia are obtained；Fig. 2 (d) be the inventive method to EO-1 hyperion The classification results figure that university of remotely-sensed data Pavia University of Pavia are obtained.
3. interpretation
Table 1 is the classification results of each method in accompanying drawing 2 to be evaluated from objective evaluation index.
1. each sorting technique precision evaluation result of table
Consolidated statement 1 and accompanying drawing 2 are as can be seen that there is more classification noise in support vector machines classification results.EMP and SS-Kernel sorting techniques can reduce noise, but be difficult to be completely eliminated the phenomenon of wrong point of homogeneous region, especially meadow and Noise in the two classification homogeneous regions in exposed soil ground, is shown in Fig. 2 (b)-(c).The present invention is equal in terms of visual effect and quantitative analyses Better than first three prior art classification method, preferable classifying quality can be reached in homogeneous region.
Above emulation experiment shows：The inventive method can make full use of the spectral signature of high spectrum image and texture special Levy, preferable classification results can be obtained in image homogeneous region.Also, the inventive method is can solve the problem that in art methods deposits The texture information for ignoring high spectrum image, the low problem of nicety of grading, be a kind of very useful high spectrum image Sorting technique.
The target in hyperspectral remotely sensed image svm classifier system of the combined spectral and textural characteristics of the embodiment of the present invention, including：
Sample acquisition unit, for being input into original Hyperspectral imaging to be sorted, and is normalized to the image； Input ground investigation set of data samples corresponding with Hyperspectral imaging to be sorted；Obtain ground investigation data sample and concentrate and own The coordinate position of sample, extracts the pixel of respective coordinates position in original Hyperspectral imaging, constitutes reference data sample set；With reference to Data sample is concentrated comprising multiple atural object classifications, is followed successively by each atural object classification and is randomly selected a number of reference data sample Training sample set as supervised classification；
Super-pixel segmentation and texture filtering unit, for carrying out principal component transform to original Hyperspectral imaging, and extract One main constituent image；The super-pixel segmentation based on entropy rate is carried out to first principal component image, obtains region segmentation figure；Lead to first Composition image carries out intensity filtering, Gauss-Laplace filtering and Gabor Filtering Processing respectively, obtains filtering image；
Spectrum histogram statistical unit, for according to filtering image, in statistical regions segmentation figure in each cut zone Spectrum rectangular histogram, obtains spectral signature information and the texture feature information in each cut zone；
Image classification unit, for by the spectral signature information in corresponding for training sample set cut zone and textural characteristics Information substitutes into complex nucleus function simultaneously, solves supporting vector machine model；According to the supporting vector machine model of compound kernel function, to original Beginning Hyperspectral imaging is classified, and obtains the Hyperspectral imaging that classifies.
It should be appreciated that for those of ordinary skills, can be improved according to the above description or be converted, And all these modifications and variations should all belong to the protection domain of claims of the present invention.

Claims (9)

1. a kind of target in hyperspectral remotely sensed image svm classifier method of combined spectral and textural characteristics, it is characterised in that including following step Suddenly：
S1, input original Hyperspectral imaging to be sorted, and the image is normalized；Input and bloom to be sorted The corresponding ground investigation set of data samples of spectrum image；
S2, the coordinate position for obtaining all samples of ground investigation data sample concentration, extract and correspondingly sit in original Hyperspectral imaging The pixel of cursor position, constitutes reference data sample set；
Comprising multiple atural object classifications in S3, reference data sample set, be followed successively by each atural object classification randomly select a number of Training sample set of the reference data sample as supervised classification；
S4, principal component transform is carried out to original Hyperspectral imaging, and extract first principal component image；
S5, the super-pixel segmentation based on entropy rate is carried out to first principal component image, obtain region segmentation figure；
S6, first principal component image is carried out respectively intensity filtering, Gauss-Laplace filtering and Gabor Filtering Processing, obtain Filtering image；
S7, according to filtering image, the spectrum rectangular histogram in statistical regions segmentation figure in each cut zone obtains each cut section Spectral signature information and texture feature information in domain；
S8, the spectral signature information in corresponding for training sample set cut zone and texture feature information are substituted into complex nucleus simultaneously Function, solves supporting vector machine model；
S9, the supporting vector machine model according to compound kernel function, classify to original Hyperspectral imaging, obtain the bloom that classifies Spectrum image；
S10, output category image.
2. the target in hyperspectral remotely sensed image svm classifier method of combined spectral according to claim 1 and textural characteristics, its feature It is, principal component transform is carried out to high spectrum image in step S4 and is concretely comprised the following steps：
S41, the covariance matrix for generating data in original spectrum coordinate；
S42, the eigenvalue for obtaining the covariance matrix and characteristic vector；
S43, arrayed feature value, find first, second and subsequent main constituent coordinate axess；
S44, the gray value for using the characteristic vector next life pixel in following formula new in each principal component；
Y=GX
Wherein, the spectral space before X and Y represent conversion respectively and after conversion, according to the mathematical principle of principal component transform, G is that X is empty Between covariance matrix eigenvectors matrix transposed matrix；
First principal component under S45, acquisition spectral space Y.
3. the target in hyperspectral remotely sensed image svm classifier method of combined spectral according to claim 1 and textural characteristics, its feature It is, is concretely comprised the following steps based on the super-pixel segmentation of entropy rate in step S5：
S51, initialization region number determination；
S52, calculating data item H (A) are used for obtaining homogeneous cluster cluster, and H (A) is made up of a series of functions：
H ( A ) = - &Sigma; i &alpha; i &Sigma; j p i , j ( A ) l o g ( p i , j ( A ) )
Wherein, one group side collection of the A for non-directed graph GWherein G=(V, E), vertex set V represent all pixels in image, E Represent the side collection on connection summit；αi=wi/wT, wiRepresent connection i-th summit side weight and,Represent normalization Constant, | v | represent summit sum, p in figurei,jRepresent transition probability；
S53, calculated equilibrium item B (A), its formula is：
B ( A ) = - &Sigma; i pz A ( i ) l o g ( pz A ( i ) ) - N A
Wherein, zARepresent the distribution of clustering cluster；
S54, the super-pixel segmentation based on entropy rate is carried out to first principal component image, its computing formula is：
m a x A H ( A ) + &lambda; B ( A )
Wherein, λ represents the weight size of object function data item and balance term, λ>0.
4. the target in hyperspectral remotely sensed image svm classifier method of combined spectral according to claim 3 and textural characteristics, its feature It is, be filtered in step S6 concretely comprises the following steps：
S61, intensity filtering is carried out to first principal component image, its computing formula is as follows：
&delta; ( x ) = 0 x &NotEqual; 0 &infin; x = 0
S62, initialization Gauss-Laplace filtering gaussian kernel standard deviation；Gauss-Laplace is carried out to first principal component image Filtering, its computing formula are as follows：
L o G ( x , y | &sigma; L o G ) = 1 &pi;&sigma; L o G 4 ( x 2 + y 2 2 &sigma; L o G 2 - 1 ) e - ( x 2 + y 2 ) / 2 &sigma; L o G 2
Wherein, σLoGIt is LoG wave filter gaussian kernel standard deviations；
S63, initialization filtering direction and Gabor filter gaussian kernel standard deviation；Gabor filters are carried out to first principal component image Ripple, its computing formula are as follows：
G a b o r ( x , y | &sigma; G a b o r , &theta; ) = e - &lsqb; ( x c o s &theta; + y s i n &theta; ) 2 + ( - x s i n &theta; + y c o s &theta; ) 2 &rsqb; / 2 &sigma; G a b o r 2 c o s &lsqb; 2 &pi; &lambda; ( x cos &omega; + y s i n &theta; ) &rsqb;
Wherein, θ represents filtering direction, σGaborIt is Gabor filter gaussian kernel standard deviation, wherein ratio σGabor/ λ is set to 0.5.
5. the target in hyperspectral remotely sensed image svm classifier method of combined spectral according to claim 1 and textural characteristics, its feature It is, in step S7, spectrum is histogrammic concretely comprises the following steps for statistics：
S71, for filtering image in cut zone W, be calculated one group of wave filter { F(α), α=1,2 ..., M }, its Middle M is the total number of wave filter；
S72, the wave filter of acquisition and first principal component image are carried out convolutional calculation respectively, obtain a group image { W(1),W(2),...,W(M)}；
S73, calculating sub-image W(α)Rectangular histogram, formula is：
H W ( &alpha; ) ( t 1 , t 2 ) = &Sigma; n &Element; W &Integral; t 1 t 2 &delta; ( t - W ( &alpha; ) ( n ) ) d t
Wherein, t1And t2Represent statistics with histogram scope bound；
S74, according to formula：The spectrum rectangular histogram of the one group of wave filter for giving is entered Row is calculated.
6. the target in hyperspectral remotely sensed image svm classifier method of combined spectral according to claim 1 and textural characteristics, its feature It is, be combined the algebraical sum that kernel function is two gaussian radial basis functions in step S8, two kernel functions correspond to bloom respectively Spectrum image spectral signature and the texture feature information that obtains is extracted, the weight proportion of spectrum core is μ, the weight of textural characteristics core Ratio is 1- μ, and the span of μ is [0,1].
7. the target in hyperspectral remotely sensed image svm classifier method of combined spectral according to claim 6 and textural characteristics, its feature It is, the gaussian radial basis function in step S8 is：
k &sigma; S T K s p e c t - t e x t ( x , y ) = ( 1 - &mu; ) k &sigma; S T K s p e c t ( x s p e x t , y s p e x t ) + &mu;k &sigma; S T K t e x t ( x t e x t , y t e x t )
Wherein,The compound kernel function of combined spectral and textural characteristics is represented,WithLight is represented respectively Spectrum kernel function and texture kernel function；X and y represent two pixels in image space, x respectivelyspectAnd yspectSpectrum is represented respectively Two, space spectral signature vector；xtextAnd ytextTexture space two texture feature vectors are represented respectively；μ(0<μ<1) represent multiple The balance term of spectral information and texture information in synkaryon function, in formula,WithCan be calculated using following formula：
k &sigma; S T K s p e c t ( x s p e t , y s p e c t ) = exp ( - | | x s p e c t - y s p e c t | | 2 2 &sigma; S T K 2 )
k &sigma; S T K t e x t ( x t e x t , y t e x t ) = exp ( - | | x t e x t - y t e x t | | 2 2 &sigma; S T K 2 )
Wherein, σSTKRepresent Gaussian function standard deviation.
8. the target in hyperspectral remotely sensed image svm classifier method of combined spectral according to claim 1 and textural characteristics, its feature It is, using test sample collection of the remaining reference data sample as precision evaluation in each atural object classification in step S3.
9. the target in hyperspectral remotely sensed image svm classifier system of a kind of combined spectral and textural characteristics, it is characterised in that include：
Sample acquisition unit, for being input into original Hyperspectral imaging to be sorted, and is normalized to the image；Input Ground investigation set of data samples corresponding with Hyperspectral imaging to be sorted；Obtain ground investigation data sample and concentrate all samples Coordinate position, extract the pixel of respective coordinates position in original Hyperspectral imaging, constitute reference data sample set；Reference data Comprising multiple atural object classifications in sample set, it is followed successively by each atural object classification and randomly selects a number of reference data sample conduct The training sample set of supervised classification；
Super-pixel segmentation and texture filtering unit, for carrying out principal component transform to original Hyperspectral imaging, and extract the first master Composition image；The super-pixel segmentation based on entropy rate is carried out to first principal component image, obtains region segmentation figure；To first principal component Image carries out intensity filtering, Gauss-Laplace filtering and Gabor Filtering Processing respectively, obtains filtering image；
Spectrum histogram statistical unit, for according to filtering image, the spectrum in statistical regions segmentation figure in each cut zone Rectangular histogram, obtains spectral signature information and the texture feature information in each cut zone；
Image classification unit, for by the spectral signature information in corresponding for training sample set cut zone and texture feature information Substitute into complex nucleus function simultaneously, solve supporting vector machine model；According to the supporting vector machine model of compound kernel function, to original height Spectrum image is classified, and obtains the Hyperspectral imaging that classifies.
CN201610929442.3A 2016-10-31 2016-10-31 The target in hyperspectral remotely sensed image svm classifier method and system of combined spectral and textural characteristics Pending CN106503739A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201610929442.3A CN106503739A (en) 2016-10-31 2016-10-31 The target in hyperspectral remotely sensed image svm classifier method and system of combined spectral and textural characteristics

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201610929442.3A CN106503739A (en) 2016-10-31 2016-10-31 The target in hyperspectral remotely sensed image svm classifier method and system of combined spectral and textural characteristics

Publications (1)

Publication Number Publication Date
CN106503739A true CN106503739A (en) 2017-03-15

Family

ID=58318735

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201610929442.3A Pending CN106503739A (en) 2016-10-31 2016-10-31 The target in hyperspectral remotely sensed image svm classifier method and system of combined spectral and textural characteristics

Country Status (1)

Country Link
CN (1) CN106503739A (en)

Cited By (37)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107066959A (en) * 2017-03-29 2017-08-18 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
CN107092930A (en) * 2017-04-21 2017-08-25 中国科学院遥感与数字地球研究所 It is a kind of by DIGITAL PLANNING map（DLG）Data are used for the method that high-resolution remote sensing image ground mulching is classified
CN107145831A (en) * 2017-04-11 2017-09-08 中国地质大学（武汉） Based on vector probabilistic diffusion and markov random file Hyperspectral Image Classification method
CN107169407A (en) * 2017-03-29 2017-09-15 广东工业大学 Hyperspectral image classification method based on joint bilateral filtering and extreme learning machine
CN107274416A (en) * 2017-06-13 2017-10-20 西北工业大学 High spectrum image conspicuousness object detection method based on spectrum gradient and hierarchical structure
CN107292264A (en) * 2017-06-19 2017-10-24 陈晓龙 A kind of information processing method and electronic equipment
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
CN107832805A (en) * 2017-11-29 2018-03-23 常州大学 It is a kind of that technology of the volumetric position error on the evaluation influence of remote sensing soft nicety of grading is eliminated based on probability positions model
CN107862330A (en) * 2017-10-31 2018-03-30 广东交通职业技术学院 A kind of hyperspectral image classification method of combination Steerable filter and maximum probability
CN108363967A (en) * 2018-01-30 2018-08-03 何德珍 A kind of categorizing system of remote sensing images scene
CN108376232A (en) * 2018-01-04 2018-08-07 北京星衡科技有限公司 A kind of method and apparatus of automatic interpretation for remote sensing image
CN108427913A (en) * 2018-02-05 2018-08-21 中国地质大学（武汉） The Hyperspectral Image Classification method of combined spectral, space and hierarchy information
CN108629297A (en) * 2018-04-19 2018-10-09 北京理工大学 A kind of remote sensing images cloud detection method of optic based on spatial domain natural scene statistics
CN109034213A (en) * 2018-07-06 2018-12-18 华中师范大学 Hyperspectral image classification method and system based on joint entropy principle
CN109101976A (en) * 2018-07-10 2018-12-28 温州大学 A kind of detection method of arc extinguishing grid pieces surface defect
CN109472287A (en) * 2018-10-06 2019-03-15 浙江大学 Three-dimensional fluorescence spectrum feature extracting method based on Two-Dimensional Gabor Wavelets
CN109472199A (en) * 2018-09-29 2019-03-15 深圳大学 A kind of method and device of image co-registration classification
CN109961087A (en) * 2019-02-01 2019-07-02 中国地质科学院矿产资源研究所 Abnormal remote sensing information extraction method and device based on spatial data set analysis
CN109961073A (en) * 2019-03-29 2019-07-02 云南电网有限责任公司电力科学研究院 The acquisition methods and device of a kind of transmission line of electricity and shaft tower information
CN109975794A (en) * 2019-03-29 2019-07-05 江西理工大学 A method of intelligent manufacturing system detection and control are carried out using high light spectrum image-forming ranging model
CN110147735A (en) * 2019-04-24 2019-08-20 湖北大学 A kind of high-spectrum remote sensing data classification method
CN111860153A (en) * 2020-01-09 2020-10-30 九江学院 Scale-adaptive hyperspectral image classification method and system
CN111881933A (en) * 2019-06-29 2020-11-03 浙江大学 Hyperspectral image classification method and system
CN112329887A (en) * 2020-11-26 2021-02-05 深圳大学 Hyperspectral and LiDAR joint classification method based on scale adaptive filtering
CN112488659A (en) * 2020-12-11 2021-03-12 浪潮云信息技术股份公司 Urban event intelligent discovery and automatic order distribution method based on remote sensing
CN112541921A (en) * 2020-11-18 2021-03-23 上海市园林科学规划研究院 Digitized accurate measuring method for urban green land vegetation information
CN112766227A (en) * 2021-02-04 2021-05-07 中国地质调查局武汉地质调查中心 Hyperspectral remote sensing image classification method, device, equipment and storage medium
CN113065518A (en) * 2021-04-25 2021-07-02 中国电子科技集团公司第二十八研究所 Hyperpixel space spectrum multi-kernel hyperspectral image classification method based on LBP (local binary pattern) features
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
CN113343900A (en) * 2021-06-28 2021-09-03 中国电子科技集团公司第二十八研究所 Combined nuclear remote sensing image target detection method based on combination of CNN and superpixel
CN113379720A (en) * 2021-06-29 2021-09-10 云南昆船设计研究院有限公司 Tea cake anti-counterfeiting method based on tea cake image feature code
CN113436090A (en) * 2021-06-16 2021-09-24 中国电子科技集团公司第五十四研究所 Remote sensing image spectrum and texture feature fusion extraction method
CN113489869A (en) * 2021-07-05 2021-10-08 深圳市威视佰科科技有限公司 Clothing material identification method based on hyperspectral camera
CN113902759A (en) * 2021-10-13 2022-01-07 自然资源部国土卫星遥感应用中心 Space-spectrum information combined satellite-borne hyperspectral image segmentation and clustering 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
CN117197137A (en) * 2023-11-07 2023-12-08 山东大学 Tissue sample analysis method and system based on hyperspectral image
CN117315381A (en) * 2023-11-30 2023-12-29 昆明理工大学 Hyperspectral image classification method based on second-order biased random walk

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101894256A (en) * 2010-07-02 2010-11-24 西安理工大学 Iris identification method based on odd-symmetric 2D Log-Gabor filter
CN101976361A (en) * 2010-11-23 2011-02-16 中国矿业大学 Multi-kernel support vector machine classification method for remote sensing images
CN103500450A (en) * 2013-09-30 2014-01-08 河海大学 Multi-spectrum remote sensing image change detection method
CN105678281A (en) * 2016-02-04 2016-06-15 中国农业科学院农业资源与农业区划研究所 Plastic film mulching farmland remote sensing monitoring method based on spectrum and texture features

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101894256A (en) * 2010-07-02 2010-11-24 西安理工大学 Iris identification method based on odd-symmetric 2D Log-Gabor filter
CN101976361A (en) * 2010-11-23 2011-02-16 中国矿业大学 Multi-kernel support vector machine classification method for remote sensing images
CN103500450A (en) * 2013-09-30 2014-01-08 河海大学 Multi-spectrum remote sensing image change detection method
CN105678281A (en) * 2016-02-04 2016-06-15 中国农业科学院农业资源与农业区划研究所 Plastic film mulching farmland remote sensing monitoring method based on spectrum and texture features

Non-Patent Citations (8)

* Cited by examiner, † Cited by third party
Title
HONGMAN WANG 等: "Texture Image Segmentation using Spectral Histogram and Skeleton Extracting", 《2009 INTERNATIONAL CONFERENCE ON ELECTRONIC COMPUTER TECHNOLOGY》 *
JIANGYE YUAN 等: "Remote Sensing Image Segmentation by Combining Spectral and Texture Features", 《IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING》 *
MING-YU LIU 等: "Entropy Rate Superpixel Segmentation", 《IEEE CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION》 *
WUHUI DUAN 等: "Superpixel-Based Composite Kernel for Hyperspectral Image Classification", 《2015 IEEE INTERNATIONAL GEOSCIENCE AND REMOTE SENSING SYMPOSIUM》 *
吴昊: "综合纹理特征的高光谱遥感图像分类方法", 《计算机工程与设计》 *
林福宗 等: "《多媒体技术课程设计与学习辅导》", 30 April 2009, 清华大学出版社 *
王亚静 等: "基于熵率超像素和区域合并的岩屑颗粒图像分割", 《计算机工程与设计》 *
陈小娟: "《高校本科专业设置预测模型构建》", 30 April 2015, 广东高等教育出版社 *

Cited By (58)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107066959A (en) * 2017-03-29 2017-08-18 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
CN107169407A (en) * 2017-03-29 2017-09-15 广东工业大学 Hyperspectral image classification method based on joint bilateral filtering and extreme learning machine
CN107066959B (en) * 2017-03-29 2020-12-18 广东交通职业技术学院 Hyperspectral image classification method based on guided filtering and linear spatial correlation information
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
CN107527001B (en) * 2017-03-29 2020-12-22 广东交通职业技术学院 Hyperspectral image classification method based on guided filtering and linear spatial correlation information
CN107145831A (en) * 2017-04-11 2017-09-08 中国地质大学（武汉） Based on vector probabilistic diffusion and markov random file Hyperspectral Image Classification method
CN107145831B (en) * 2017-04-11 2020-02-14 中国地质大学（武汉） Hyperspectral image classification method based on vector probability diffusion and Markov random field
CN107092930A (en) * 2017-04-21 2017-08-25 中国科学院遥感与数字地球研究所 It is a kind of by DIGITAL PLANNING map（DLG）Data are used for the method that high-resolution remote sensing image ground mulching is classified
CN107274416A (en) * 2017-06-13 2017-10-20 西北工业大学 High spectrum image conspicuousness object detection method based on spectrum gradient and hierarchical structure
CN107274416B (en) * 2017-06-13 2019-11-01 西北工业大学 High spectrum image conspicuousness object detection method based on spectrum gradient and hierarchical structure
CN107292264A (en) * 2017-06-19 2017-10-24 陈晓龙 A kind of information processing method and electronic equipment
CN107862330A (en) * 2017-10-31 2018-03-30 广东交通职业技术学院 A kind of hyperspectral image classification method of combination Steerable filter and maximum probability
CN107832805B (en) * 2017-11-29 2021-04-16 常州大学 Technology for eliminating influence of spatial position error on remote sensing soft classification precision evaluation based on probability position model
CN107832805A (en) * 2017-11-29 2018-03-23 常州大学 It is a kind of that technology of the volumetric position error on the evaluation influence of remote sensing soft nicety of grading is eliminated based on probability positions model
CN108376232A (en) * 2018-01-04 2018-08-07 北京星衡科技有限公司 A kind of method and apparatus of automatic interpretation for remote sensing image
CN108363967A (en) * 2018-01-30 2018-08-03 何德珍 A kind of categorizing system of remote sensing images scene
CN108427913A (en) * 2018-02-05 2018-08-21 中国地质大学（武汉） The Hyperspectral Image Classification method of combined spectral, space and hierarchy information
CN108427913B (en) * 2018-02-05 2021-12-28 中国地质大学（武汉） Hyperspectral image classification method combining spectral, spatial and hierarchical structure information
CN108629297A (en) * 2018-04-19 2018-10-09 北京理工大学 A kind of remote sensing images cloud detection method of optic based on spatial domain natural scene statistics
CN109034213A (en) * 2018-07-06 2018-12-18 华中师范大学 Hyperspectral image classification method and system based on joint entropy principle
CN109034213B (en) * 2018-07-06 2021-08-03 华中师范大学 Hyperspectral image classification method and system based on correlation entropy principle
CN109101976A (en) * 2018-07-10 2018-12-28 温州大学 A kind of detection method of arc extinguishing grid pieces surface defect
CN109101976B (en) * 2018-07-10 2021-11-30 温州大学 Method for detecting surface defects of arc-extinguishing grid plate
CN109472199B (en) * 2018-09-29 2022-02-22 深圳大学 Image fusion classification method and device
WO2020062360A1 (en) * 2018-09-29 2020-04-02 深圳大学 Image fusion classification method and apparatus
CN109472199A (en) * 2018-09-29 2019-03-15 深圳大学 A kind of method and device of image co-registration classification
CN109472287A (en) * 2018-10-06 2019-03-15 浙江大学 Three-dimensional fluorescence spectrum feature extracting method based on Two-Dimensional Gabor Wavelets
CN109961087A (en) * 2019-02-01 2019-07-02 中国地质科学院矿产资源研究所 Abnormal remote sensing information extraction method and device based on spatial data set analysis
CN109961073A (en) * 2019-03-29 2019-07-02 云南电网有限责任公司电力科学研究院 The acquisition methods and device of a kind of transmission line of electricity and shaft tower information
CN109975794A (en) * 2019-03-29 2019-07-05 江西理工大学 A method of intelligent manufacturing system detection and control are carried out using high light spectrum image-forming ranging model
CN110147735A (en) * 2019-04-24 2019-08-20 湖北大学 A kind of high-spectrum remote sensing data classification method
CN110147735B (en) * 2019-04-24 2023-04-07 湖北大学 Hyperspectral remote sensing data classification method
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
CN111860153B (en) * 2020-01-09 2023-10-13 九江学院 Scale-adaptive hyperspectral image classification method and system
CN111860153A (en) * 2020-01-09 2020-10-30 九江学院 Scale-adaptive hyperspectral image classification method and system
CN112541921B (en) * 2020-11-18 2023-06-20 上海市园林科学规划研究院 Urban green land vegetation information data accurate determination method
CN112541921A (en) * 2020-11-18 2021-03-23 上海市园林科学规划研究院 Digitized accurate measuring method for urban green land vegetation information
CN112329887B (en) * 2020-11-26 2023-12-05 深圳大学 Hyperspectral and LiDAR joint classification method based on scale adaptive filtering
CN112329887A (en) * 2020-11-26 2021-02-05 深圳大学 Hyperspectral and LiDAR joint classification method based on scale adaptive filtering
CN112488659A (en) * 2020-12-11 2021-03-12 浪潮云信息技术股份公司 Urban event intelligent discovery and automatic order distribution method based on remote sensing
CN112766227A (en) * 2021-02-04 2021-05-07 中国地质调查局武汉地质调查中心 Hyperspectral remote sensing image classification method, device, equipment and storage medium
CN112766227B (en) * 2021-02-04 2023-11-03 中国地质调查局武汉地质调查中心 Hyperspectral remote sensing image classification method, device, equipment and storage medium
CN113065518A (en) * 2021-04-25 2021-07-02 中国电子科技集团公司第二十八研究所 Hyperpixel space spectrum multi-kernel hyperspectral image classification method based on LBP (local binary pattern) features
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
CN113436090A (en) * 2021-06-16 2021-09-24 中国电子科技集团公司第五十四研究所 Remote sensing image spectrum and texture feature fusion extraction method
CN113343900A (en) * 2021-06-28 2021-09-03 中国电子科技集团公司第二十八研究所 Combined nuclear remote sensing image target detection method based on combination of CNN and superpixel
CN113379720B (en) * 2021-06-29 2022-08-09 云南昆船设计研究院有限公司 Tea cake anti-counterfeiting method based on tea cake image feature code
CN113379720A (en) * 2021-06-29 2021-09-10 云南昆船设计研究院有限公司 Tea cake anti-counterfeiting method based on tea cake image feature code
CN113489869A (en) * 2021-07-05 2021-10-08 深圳市威视佰科科技有限公司 Clothing material identification method based on hyperspectral camera
CN113902759B (en) * 2021-10-13 2022-04-22 自然资源部国土卫星遥感应用中心 Space-spectrum information combined satellite-borne hyperspectral image segmentation and clustering method
CN113902759A (en) * 2021-10-13 2022-01-07 自然资源部国土卫星遥感应用中心 Space-spectrum information combined satellite-borne hyperspectral image segmentation and clustering 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
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
CN117197137A (en) * 2023-11-07 2023-12-08 山东大学 Tissue sample analysis method and system based on hyperspectral image
CN117197137B (en) * 2023-11-07 2024-02-09 山东大学 Tissue sample analysis method and system based on hyperspectral image
CN117315381A (en) * 2023-11-30 2023-12-29 昆明理工大学 Hyperspectral image classification method based on second-order biased random walk
CN117315381B (en) * 2023-11-30 2024-02-09 昆明理工大学 Hyperspectral image classification method based on second-order biased random walk

Similar Documents

Publication Publication Date Title
CN106503739A (en) The target in hyperspectral remotely sensed image svm classifier method and system of combined spectral and textural characteristics
CN106339674B (en) The Hyperspectral Image Classification method that model is cut with figure is kept based on edge
Xu et al. Regional clustering-based spatial preprocessing for hyperspectral unmixing
CN107451614B (en) Hyperspectral classification method based on fusion of space coordinates and space spectrum features
CN106203523B (en) The hyperspectral image classification method of the semi-supervised algorithm fusion of decision tree is promoted based on gradient
CN106529508B (en) Based on local and non local multiple features semanteme hyperspectral image classification method
CN105069468B (en) Hyperspectral image classification method based on ridge ripple and depth convolutional network
CN107992891B (en) Multispectral remote sensing image change detection method based on spectral vector analysis
CN109492593B (en) Hyperspectral image classification method based on principal component analysis network and space coordinates
CN108241871A (en) Laser point cloud and visual fusion data classification method based on multiple features
CN107292339A (en) The unmanned plane low altitude remote sensing image high score Geomorphological Classification method of feature based fusion
Lv et al. Novel adaptive region spectral-spatial features for land cover classification with high spatial resolution remotely sensed imagery
CN103208011B (en) Based on average drifting and the hyperspectral image space-spectral domain classification method organizing sparse coding
Liu et al. Multimorphological superpixel model for hyperspectral image classification
CN105718942B (en) High spectrum image imbalance classification method based on average drifting and over-sampling
CN107895139A (en) A kind of SAR image target recognition method based on multi-feature fusion
CN104182767B (en) The hyperspectral image classification method that Active Learning and neighborhood information are combined
CN108446582A (en) Hyperspectral image classification method based on textural characteristics and affine propagation clustering algorithm
CN108154094A (en) The non-supervisory band selection method of high spectrum image divided based on subinterval
CN105160351B (en) Semi-supervised hyperspectral classification method based on anchor point sparse graph
CN109034213B (en) Hyperspectral image classification method and system based on correlation entropy principle
CN109697465A (en) Hyperspectral Image Classification method based on the Fisher discriminant analysis of Wavelet Kernel part
CN108427913A (en) The Hyperspectral Image Classification method of combined spectral, space and hierarchy information
CN109829426B (en) Railway construction temporary building monitoring method and system based on high-resolution remote sensing image
Guo et al. Dual graph U-Nets for hyperspectral image classification

Legal Events

Date Code Title Description
C06 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

Application publication date: 20170315

RJ01 Rejection of invention patent application after publication