CN110992245B - Hyperspectral image dimension reduction method and device - Google Patents
Hyperspectral image dimension reduction method and device Download PDFInfo
- Publication number
- CN110992245B CN110992245B CN201911129254.2A CN201911129254A CN110992245B CN 110992245 B CN110992245 B CN 110992245B CN 201911129254 A CN201911129254 A CN 201911129254A CN 110992245 B CN110992245 B CN 110992245B
- Authority
- CN
- China
- Prior art keywords
- matrix
- neighborhood
- hyperspectral image
- subcube
- dimension reduction
- 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.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 62
- 230000009467 reduction Effects 0.000 title claims abstract description 62
- 239000011159 matrix material Substances 0.000 claims description 243
- 238000010606 normalization Methods 0.000 claims description 32
- 238000000354 decomposition reaction Methods 0.000 claims description 29
- 238000000605 extraction Methods 0.000 claims description 28
- 230000003595 spectral effect Effects 0.000 claims description 15
- 238000004364 calculation method Methods 0.000 claims description 11
- 230000008569 process Effects 0.000 claims description 11
- 230000006641 stabilisation Effects 0.000 claims description 5
- 238000011105 stabilization Methods 0.000 claims description 5
- 230000008707 rearrangement Effects 0.000 claims description 3
- 238000005457 optimization Methods 0.000 abstract description 6
- 238000005516 engineering process Methods 0.000 abstract description 3
- 238000001228 spectrum Methods 0.000 description 5
- 230000006870 function Effects 0.000 description 4
- 238000010586 diagram Methods 0.000 description 3
- 230000009466 transformation Effects 0.000 description 3
- 238000010187 selection method Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 229910052500 inorganic mineral Inorganic materials 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 239000011707 mineral Substances 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000000513 principal component analysis Methods 0.000 description 1
- 238000012847 principal component analysis method Methods 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
- 238000000926 separation method Methods 0.000 description 1
- 238000000638 solvent extraction Methods 0.000 description 1
- 239000000126 substance Substances 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T3/00—Geometric image transformations in the plane of the image
- G06T3/06—Topological mapping of higher dimensional structures onto lower dimensional surfaces
-
- 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
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Image Processing (AREA)
Abstract
The application discloses a hyperspectral image dimension reduction method and device. The hyperspectral image dimension reduction method comprises the following steps: acquiring a neighborhood sub-cube f of a pixel in a hyperspectral image I to be processed using a sliding kernel i The method comprises the steps of carrying out a first treatment on the surface of the For each acquired neighborhood sub-cube f i Performing dimension reduction processing to obtain a neighborhood hyperspectral image of each neighborhood subcube after dimension reductionAccording to the spatial position of each neighborhood subcube in the hyperspectral image, the neighborhood hyperspectral image of each neighborhood subcubeRearranging to obtain hyperspectral image after dimension reductionAccording to the technical scheme, when the dimension of the hyperspectral image is reduced, the linear dynamic system feature model is combined with the sparse feedback optimization technology, so that the wave bands with large information quantity and good class separability in the hyperspectral image are extracted, the classification precision of the hyperspectral image is improved, and meanwhile, the calculated amount of hyperspectral image classification is reduced.
Description
Technical Field
The application relates to the field of dimension reduction of hyperspectral images, in particular to a dimension reduction method and device of hyperspectral images.
Background
The accurate classification of hyperspectral images plays an important role in the application fields of industry, agriculture and aerospace, and can be applied to a plurality of practical application fields such as accurate agriculture, environmental mapping, social security, mineral exploration, biological and chemical detection and the like.
However, the hyperspectral image has high spectral dimension and strong statistical correlation among spectral bands, so that the information redundancy and the calculation complexity are high, and finally the classification precision is low, thereby restricting the application of the hyperspectral remote sensing image. The correlation among spectrum bands can be weakened or eliminated by image dimension reduction, the distinguishability of pixels is improved, the calculated amount is reduced, and the classification precision of remote sensing images is improved, so that the effective dimension reduction method is necessary to reduce the band dimension while keeping enough spectrum information.
The existing hyperspectral image dimension reduction method mainly comprises a feature extraction method based on mathematical transformation and a feature selection method based on band selection. For the feature extraction method based on mathematical transformation, the method mainly comprises the following steps: a principal component analysis method (Principal Component Analysis), a minimum noise separation transformation (Minimum Noise Fraction Rotation), and a nonlinear popular learning method (Nonlinear manifold learning); for the feature selection method based on band selection, for example, a band combination with large information quantity, good category separability or small correlation can be selected.
However, the existing hyperspectral image dimension reduction method ignores the space structure information of the hyperspectral image, and the method is large in calculation amount. Meanwhile, the process of manually setting parameters and selecting thresholds affects the classification accuracy.
How to make up for the above-mentioned shortcomings in the prior art when the dimension of the hyperspectral image is reduced, no related solution exists in the prior art at present.
Disclosure of Invention
In order to solve the technical problems, the application provides a hyperspectral image dimension reduction method, which can improve the classification precision of hyperspectral images, reduce the calculated amount of hyperspectral image classification, and solve the problems that the conventional hyperspectral image dimension reduction needs to be manually participated and space structure information of the hyperspectral images is not fully utilized.
In order to achieve the purpose of the application, the application provides a hyperspectral image dimension reduction method, which comprises the following steps:
acquiring a neighborhood sub-cube f of a pixel in a hyperspectral image I to be processed using a sliding kernel i Wherein the size of I is m multiplied by n multiplied by k, m represents the space line number of the hyperspectral image, and n representsShowing the number of spatial columns of the hyperspectral image, k representing the spectral dimension of the hyperspectral image, the sliding kernel having a size w×w, w being an odd number greater than 1, i being a positive integer no greater than min (m-w+1, n-w+1), f i Is w×w×k;
for each acquired neighborhood sub-cube f i Performing dimension reduction processing to obtain a neighborhood hyperspectral image of each neighborhood subcube after dimension reduction
According to the spatial position of each neighborhood subcube in the hyperspectral image, the neighborhood hyperspectral image of each neighborhood subcubeRearranging to obtain the hyperspectral image +.>
Further, the pair obtains each neighborhood subcube f i Performing dimension reduction processing to obtain a neighborhood hyperspectral image of each neighborhood subcube after dimension reductionThe method comprises the following steps:
acquiring each neighborhood subcube f i Matrix M obtained after tiling i Wherein the matrix M i Is k×s, where s=w×w;
for the matrix M i Performing decomposition processing, and establishing a linear dynamic system model (A) of the neighborhood subcube according to the decomposition processing result i ,B i ,C i )；
According to the linear dynamic system model (A i ,B i ,C i ) In said matrix A i Sum matrix B i Acquisition matrix C i Is a sparse matrix of (2)
For the sparse matrixCarrying out normalization processing, and obtaining the neighborhood hyperspectral image ++of each neighborhood subcubes after dimension reduction according to the normalization processing result>
Further, wherein the pair of matrices M i The decomposing treatment is carried out, which comprises the following steps:
decomposing the SDV with singular values to the matrix M by the following formula i Decomposing to obtain a decomposed matrix U i Matrix lambda i Sum matrix V i ：
M i ＝U i Λ i V i T ；
Wherein the matrix U i Is k x s and U i T ×U i =i, the matrix V i And the matrix lambda i Is s x s, V i T ×V i ＝I；
The linear dynamic system model (A) of the neighborhood subcube is built according to the decomposition processing result i ,B i ,C i ) The method comprises the following steps:
by using the matrix U i The matrix lambda i And the matrix V i A linear dynamic system model (A i ,B i ,C i ) The method specifically comprises the following steps:
C i ＝U i
wherein X is i ＝Λ i V i T ；
In calculation B i When first calculateThen utilize formula G i ＝U′ i Λ′ i V′ i Decomposition matrix G i Obtaining U' i Sum Λ'. i Thereby calculating to obtain B i ；
Wherein matrix A i Is s X s, matrix X i Is a system state variable matrix, the size of which is s x s,representing extraction of the matrix X i 1 to s rows and 2 to s columns, < ->Representing extraction of the matrix X i 1 to s rows and 1 to s-1 columns,representing the extraction matrix U' i Column 1, < >>Representing the extraction matrix Λ' i 1 st row 1 st column element of (c).
Further, the linear dynamic system model (A i ,B i ,C i ) In said matrix A i Sum matrix B i Acquisition matrix C i Is a sparse matrix of (2)The method comprises the following steps:
matrix is obtained by using Lyapunov stabilization condition and sparsity assumption condition through the following equationAnd sparse matrix->The optimal solution of the equation is the matrix +.>And the sparse matrix->
||Y i || col →min,
Wherein,,
computing a matrix
The sparse matrixFrom the matrix->Is equal to the sparse matrix->A row corresponding to a column index of not 0.
Further, the pair of sparse matricesThe normalization processing is carried out, which comprises the following steps:
establishing a zero matrixMatrix->Is filled into the matrix +.>The filled position is the row in the matrix +.>Row index of (2), filled +.>Namely +.>Normalization processing results after normalization processing;
obtaining the neighborhood hyperspectral image of each neighborhood subcube after dimension reduction according to the normalization processing resultThe method comprises the following steps:
obtaining the neighborhood hyperspectral image of each neighborhood subcube after dimension reduction through the following formula
The application also provides a hyperspectral image dimension reduction device, which comprises:
a neighborhood subcube acquisition module configured to acquire a neighborhood subcube f of a pixel in the hyperspectral image I to be processed using a sliding kernel i Wherein, the size of I is m multiplied by n multiplied by k,m represents the number of spatial lines of the hyperspectral image, n represents the number of spatial columns of the hyperspectral image, k represents the spectral dimension of the hyperspectral image, the size of the sliding kernel is w×w, w is an odd number greater than 1, i is a positive integer not greater than min (m-w+1, n-w+1), f i Is w×w×k;
a neighborhood hyperspectral image generation module configured to generate, for each acquired neighborhood subcube f i Performing dimension reduction processing to obtain a neighborhood hyperspectral image of each neighborhood subcube after dimension reduction
A neighborhood hyperspectral image rearrangement module configured to rearrange neighborhood hyperspectral images of the respective neighborhood subcubes according to spatial positions of the respective neighborhood subcubes in the hyperspectral imagesRearranging to obtain the hyperspectral image +.>
Further, the neighborhood hyperspectral image generation module includes:
a neighborhood subcube matrix generation module configured to obtain the neighborhood subcubes f i Matrix M obtained after tiling i Wherein the matrix M i Is k×s, where s=w×w;
a linear dynamic system model building module configured to build the matrix M i Performing decomposition processing, and establishing a linear dynamic system model (A) of the neighborhood subcube according to the decomposition processing result i ,B i ,C i )；
A sparse matrix acquisition module arranged to acquire a sparse matrix based on the linear dynamic system model (A i ,B i ,C i ) In said matrix A i Sum matrix B i Acquisition matrix C i Is a sparse matrix of (2)
A sparse matrix processing module configured to process the sparse matrixCarrying out normalization processing, and obtaining the neighborhood hyperspectral image ++of each neighborhood subcubes after dimension reduction according to the normalization processing result>
Further, the pair of matrices M i Performing decomposition processing, including:
decomposing the SDV with singular values to the matrix M by the following formula i Decomposing to obtain a decomposed matrix U i Matrix lambda i Sum matrix V i ：
M i ＝U i Λ i V i T ；
Wherein the matrix U i Is k x s and U i T ×U i =i, the matrix V i And the matrix lambda i Is s x s, V i T ×V i ＝I；
The linear dynamic system model (A) of the neighborhood subcube is built according to the decomposition processing result i ,B i ,C i ) Comprising:
by using the matrix U i The matrix lambda i And the matrix V i A linear dynamic system model (A i ,B i ,C i ) The method specifically comprises the following steps:
C i ＝U i
wherein X is i ＝Λ i V i T ；
In calculation B i When first calculateThen utilize formula G i ＝U′ i Λ′ i V′ i Decomposition matrix G i Obtaining U' i Sum Λ'. i Thereby calculating to obtain B i ；
Wherein matrix A i Is s X s, matrix X i Is a system state variable matrix, the size of which is s x s,representing extraction of the matrix X i 1 to s rows and 2 to s columns, < ->Representing extraction of the matrix X i 1 to s rows and 1 to s-1 columns,representing the extraction matrix U' i Column 1, < >>Representing the extraction matrix Λ' i 1 st row 1 st column element of (c).
Further, the sparse matrix acquisition module is specifically configured to:
matrix is obtained by using Lyapunov stabilization condition and sparsity assumption condition through the following equationAnd sparse matrix->The saidThe optimal solution of the equation is the matrix +.>And the sparse matrix->
||Y i || col →min,
Wherein,,
computing a matrix
The sparse matrixFrom the matrix->Is equal to the sparse matrix->A row corresponding to a column index of not 0.
Further, the sparse matrix processing module is specifically configured to:
establishing a zero matrixMatrix->Is filled into the matrix +.>The filled position is the row in the matrix +.>Row index of (2), filled +.>Namely +.>Normalization processing results after normalization processing;
obtaining the neighborhood hyperspectral image of each neighborhood subcube after dimension reduction according to the normalization processing resultComprising the following steps:
obtaining the neighborhood hyperspectral image of each neighborhood subcube after dimension reduction through the following formula
Compared with the prior art, the hyperspectral image dimension reduction method comprises the following steps: acquiring a neighborhood sub-cube f of a pixel in a hyperspectral image I to be processed using a sliding kernel i The method comprises the steps of carrying out a first treatment on the surface of the For each acquired neighborhood sub-cube f i Performing dimension reduction processing to obtain a neighborhood hyperspectral image of each neighborhood subcube after dimension reductionAccording to the spatial position of each neighborhood subcube in the hyperspectral image, the neighborhood hyperspectral image of each neighborhood subcube is +.>Rearranging to obtain the hyperspectral image +.>According to the technical scheme, when the hyperspectral image is subjected to dimension reduction, the linear dynamic system feature model is combined with the sparse feedback optimization technology, so that the wave bands with large information quantity and good class separability in the hyperspectral image are extracted, the classification precision of the hyperspectral image is improved, the calculated amount of hyperspectral image classification is reduced, and the problems that the existing hyperspectral image dimension reduction needs to be manually participated and the space structure information of the hyperspectral image is not fully utilized are solved.
Drawings
The accompanying drawings are included to provide a further understanding of the technical aspects of the present application, and are incorporated in and constitute a part of this specification, illustrate the technical aspects of the present application and together with the examples of the present application, and not constitute a limitation of the technical aspects of the present application. In the drawings:
FIG. 1 is a process flow diagram of a hyperspectral image dimension reduction method in the present application;
FIG. 2 is a schematic structural diagram of a hyperspectral image dimension reduction device in the present application;
the realization, functional characteristics and advantages of the present application will be further described with reference to the embodiments, referring to the attached drawings.
Detailed Description
It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the present application.
For the purposes of making the objects, technical solutions and advantages of the present application more apparent, the present invention will be described in detail hereinafter with reference to the accompanying drawings. It should be noted that, in the case of no conflict, the features in the present application may be arbitrarily combined with each other.
Fig. 1 is a process flow chart of a hyperspectral image dimension reduction method in the present application, as shown in fig. 1, including the following steps:
step 101: acquiring neighborhood sub-pixels in hyperspectral image I to be processed using sliding kernelsCube f i ；
Specifically, in this step, a spatial domain rigid partitioning method is employed, in the hyperspectral image I, sliding in the lateral and longitudinal directions over the entire spatial domain of the image using a sliding kernel, thereby defining a spatial local neighborhood range for each pixel, wherein the hyperspectral image is divided into neighborhood subcubes which retain all spectral information associated with each pixel; wherein, the dimension of I is m multiplied by n multiplied by k, m represents the space line number of the hyperspectral image, n represents the space column number of the hyperspectral image, k represents the spectrum dimension of the hyperspectral image, the dimension of the sliding kernel used is w multiplied by w, w is an odd number larger than 1, and the obtained neighborhood subcubes are represented as f i The value of i is a positive integer not greater than min (m-w+1, n-w+1), f i Is w×w×k;
here, w may be defined as an odd number in the range of 3 to 21.
Step 102: for each acquired neighborhood sub-cube f i Performing dimension reduction processing to obtain a neighborhood hyperspectral image of each neighborhood subcube after dimension reduction
Specifically, this step can be divided into the following sub-steps:
step 1021: acquiring each neighborhood subcube f i Matrix M obtained after tiling i Wherein the matrix M i Is k×s, where s=w×w;
specifically, the neighborhood is sub-cube f i Tiling along the spectral dimension into a matrix, i.e. extracting each pixel in the neighborhood subcubes as a column to form a matrix M i 。
Step 1022: for the matrix M i Performing decomposition processing, and establishing a linear dynamic system model (A) of the neighborhood subcube according to the decomposition processing result i ,B i ,C i )；
Specifically, in this step, the matrix M is first decomposed by the singular value decomposition SDV by equation (1) i Decomposing to obtain a decomposed matrix U i Matrix lambda i Sum matrix V i ：
M i ＝U i Λ i V i T (1)
Wherein matrix U i Is k x s and U i T ×U i =i, matrix V i Sum matrix lambda i Is s x s, V i T ×V i ＝I；
The singular value decomposition is an important matrix decomposition method in linear algebra and matrix theory, and is suitable for the fields of signal processing, statistics and the like.
Then, a linear dynamic system model (A) of the neighborhood subcube is built according to the decomposition processing result i ,B i ,C i ) The method specifically comprises the following steps:
using a matrix U i Matrix lambda i Sum matrix V i Building a linear dynamic system model of the neighborhood subcube (A i ,B i ,C i ) The method specifically comprises the following steps: first, the model parameters A are calculated by the following formula (2) i And C i ，
Matrix A in equation (2) i Is s X s, matrix X i Is a system state variable matrix, the size of which is s x s,representing extraction matrix X i 1 to s rows and 2 to s columns, < ->Representing extraction matrix X i 1 to s rows and 1 to s-1 columns,/->Representing the extraction matrix U' i Is described in the column 1 of the above,/>representing the extraction matrix Λ' i Row 1, column 1 elements of (a);
then the matrix A is utilized by using the following formula (3) i And X i Calculation result obtaining matrix B i ：
It should be noted that: in calculation B i When passing through A i And X i G is calculated i Then for G i Decomposing to obtain U' i 、Λ′ i And V' i Thereby obtaining matrix B i Wherein, the method comprises the steps of, wherein,representing the extraction matrix U' i Column 1, < >>Representing the extraction matrix Λ' i 1 st row 1 st column element of (c).
Step 1023: according to the linear dynamic system model (A i ,B i ,C i ) In said matrix A i Sum matrix B i Acquisition matrix C i Is a sparse matrix of (2)
The step is to automatically find the wave band with large information quantity by using sparse state feedback constraint.
The linear dynamic system model (a i ,B i ,C i ) When the matrix represents the neighborhood subcube hyperspectral image, then the original neighborhood subcube hyperspectral image is described as the output variable y of the system i ＝C i X i Wherein X is i Is a system state variable matrix with the size of s multiplied by s, composed ofThe calculation is carried out in the formula (2); therefore, to obtain a reduced-dimension neighborhood subcube hyperspectral imageThe output matrix C needs to be obtained i Sparse matrix +.>The method comprises the following specific steps:
step one, establishing a Lyapunov stable condition based on a sparse state feedback system;
to original system model (A) i ,B i ,C i ) A closed loop system is formed by adopting state feedback, and a feedback matrix is K i Obtaining a system model (A) i -B i K i ,B i ,C i ) According to the lyapunov stability principle, if a positive symmetric matrix P is present i Satisfying equation (4), then the closed loop system matrix A i -B i K i Is stable;
multiplying both the left and right sides of equation (4) by P i -1 And is arranged to obtain a formula (5),
let matrix P i -1 ＝W i And (2) andmatrix W i Is of size k x k, let Y at the same time i ＝K i W, finishing the formula (5) to obtain the Lyapunov stability condition as shown in the formula (6):
step two, matrix C i Sparsity determination of (2);
||C i || col →min (7)
wherein,,equation (7) represents matrix C i Is a column sparse representation of (2);
step three, adopting convex function optimization to obtain sparsity matrix C i ；
The matrix can be obtained according to formula (8)And sparse matrix->And->There are columns of all 0's in the list, and then calculateFinally, a sparse matrix->For matrix->Middle and->A row corresponding to a column index of not 0.
Step 1024: for the sparse matrixCarrying out normalization processing, and obtaining the neighborhood hyperspectral image ++of each neighborhood subcubes after dimension reduction according to the normalization processing result>
Because the convex function optimization method is an automatic solving process, no parameters are manually set, and the convex function optimization method is obtained through step 1023The dimension of the spectrum is different between pixels after dimension reduction; even if the spectral dimensions between pixels are the same, but the spectral information is from different bands, the comparison calculation between pixels cannot be performed, and the classification processing of the image cannot be realized.
Specifically, the normalization processing method comprises the following steps: solving the matrix obtained in step 1023Let ∈ ->Representing the maximum value of each column sum, matrix +.>The value of the middle column sum is less than +>Wherein sigma is greater than 0 < sigma < 0.5, and a column index other than 0 is recorded; at this time, sparse matrix->For matrix->The rows corresponding to the indexes are formed; establishing a zero matrix +.>Matrix->Is filled into the matrix +.>The filled position is the row in the matrix +.>Row index of (2), filled +.>Namely +.>Normalization processing results after normalization processing; computing neighborhood subcube pixels +.>And (3) completing pixel normalization, and finally realizing automatic dimension reduction of the hyperspectral image.
Step 103: according to the spatial position of each neighborhood subcube in the hyperspectral image, the neighborhood hyperspectral image of each neighborhood subcubeRearranging to obtain the hyperspectral image +.>
Obtaining the dimension reduction result of each neighborhood subcube by utilizing step 102According to the spatial position of each neighborhood subcube in the original whole hyperspectral image I will +.>Rearranging together to finally obtain the hyperspectral remote sensing image after dimension reduction>
Compared with the prior art, the beneficial effect of this application lies in:
(1) According to the method, the space and spectrum information are effectively and fully utilized, an integrated characteristic model based on a linear dynamic system model is established, and dimension reduction of a hyperspectral image is more accurately and effectively achieved;
(2) According to the method, a linear dynamic integrated characteristic model of the hyperspectral image is established, a sparse feedback technology is effectively utilized, the integral relation between space and spectral characteristics in the hyperspectral image is fully considered, spectral information with large information quantity and strong separability is reserved by utilizing convex function optimization and constraint conditions, redundant spectral information is weakened or even deleted, and automatic dimension reduction of the hyperspectral remote sensing image is realized;
(3) The method has more practical application value and demand for improving the classification precision of the hyperspectral image, reducing the calculated amount of the hyperspectral image classification method and improving the application level of the hyperspectral image in the fields of industry, agriculture and aerospace.
Fig. 2 is a schematic structural diagram of a hyperspectral image dimension reduction device in the present application. As shown in fig. 2, the hyperspectral image dimension reduction device includes:
a neighborhood subcube acquisition module configured to acquire a neighborhood subcube f of a pixel in the hyperspectral image I to be processed using a sliding kernel i Wherein the size of I is m multiplied by n multiplied by k, m represents the number of spatial lines of the hyperspectral image, n represents the number of spatial columns of the hyperspectral image, k represents the spectral dimension of the hyperspectral image, the size of the sliding kernel is w multiplied by w, w is an odd number greater than 1, the value of I is a positive integer not greater than min (m-w+1, n-w+1), f i Is w×w×k;
a neighborhood hyperspectral image generation module configured to generate a neighborhood hyperspectral image of the acquired objectEach neighborhood subcube f of (1) i Performing dimension reduction processing to obtain a neighborhood hyperspectral image of each neighborhood subcube after dimension reduction
A neighborhood hyperspectral image rearrangement module configured to rearrange neighborhood hyperspectral images of the respective neighborhood subcubes according to spatial positions of the respective neighborhood subcubes in the hyperspectral imagesRearranging to obtain the hyperspectral image +.>
Specifically, the neighborhood hyperspectral image generation module comprises:
a neighborhood subcube matrix generation module configured to obtain the neighborhood subcubes f i Matrix M obtained after tiling i Wherein the matrix M i Is k×s, where s=w×w;
a linear dynamic system model building module configured to build the matrix M i Performing decomposition processing, and establishing a linear dynamic system model (A) of the neighborhood subcube according to the decomposition processing result i ,B i ,C i )；
A sparse matrix acquisition module arranged to acquire a sparse matrix based on the linear dynamic system model (A i ,B i ,C i ) In said matrix A i Sum matrix B i Acquisition matrix C i Is a sparse matrix of (2)
A sparse matrix processing module configured to process the sparse matrixNormalization processing is carried out according to the Chinese angelicaObtaining the neighborhood hyperspectral image of each neighborhood subcube after dimension reduction by the unification processing result>
Wherein said pair of said matrices M i Performing decomposition processing, including:
decomposing the SDV with singular values to the matrix M by the following formula i Decomposing to obtain a decomposed matrix U i Matrix lambda i Sum matrix V i ：
M i ＝U i Λ i V i T ；
Wherein the matrix U i Is k x s and U i T ×U i =i, the matrix V i And the matrix lambda i Is s x s, V i T ×V i ＝I；
The linear dynamic system model (A) of the neighborhood subcube is built according to the decomposition processing result i ,B i ,C i ) Comprising:
by using the matrix U i The matrix lambda i And the matrix V i A linear dynamic system model (A i ,B i ,C i ) The method specifically comprises the following steps:
C i ＝U i
wherein X is i ＝Λ i V i T ；
In calculation B i When first calculateThen utilize formula G i ＝U′ i Λ′ i V′ i Decomposition matrix G i Obtaining U' i Sum Λ'. i Thereby calculating to obtain B i ；
Wherein matrix A i Is s X s, matrix X i Is a system state variable matrix, the size of which is s x s,representing extraction of the matrix X i 1 to s rows and 2 to s columns, < ->Representing extraction of the matrix X i 1 to s rows and 1 to s-1 columns,representing the extraction matrix U' i Column 1, < >>Representing the extraction matrix Λ' i 1 st row 1 st column element of (c).
Specifically, the sparse matrix acquisition module is specifically configured to:
matrix is obtained by using Lyapunov stabilization condition and sparsity assumption condition through the following equationAnd sparse matrix->The optimal solution of the equation is the matrix +.>And the sparse matrix->
||Y i || col →min,
Wherein,,
computing a matrix
The sparse matrixFrom the matrix->Is equal to the sparse matrix->A row corresponding to a column index of not 0.
Specifically, the sparse matrix processing module is specifically configured to:
establishing a zero matrixMatrix->Is filled into the matrix +.>The filled position is the row in the matrix +.>Row index of (2), filled +.>Namely, isFor the sparse matrix->Normalization processing results after normalization processing;
obtaining the neighborhood hyperspectral image of each neighborhood subcube after dimension reduction according to the normalization processing resultComprising the following steps:
obtaining the neighborhood hyperspectral image of each neighborhood subcube after dimension reduction through the following formula
It should be noted that, in this document, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, method, article, or apparatus that comprises the element.
The foregoing embodiment numbers of the present application are merely for describing, and do not represent advantages or disadvantages of the embodiments.
From the above description of the embodiments, it will be clear to those skilled in the art that the above-described embodiment method may be implemented by means of software plus a necessary general hardware platform, but of course may also be implemented by means of hardware, but in many cases the former is a preferred embodiment. Based on such understanding, the technical solution of the present application may be embodied essentially or in a part contributing to the prior art in the form of a software product stored in a storage medium (such as ROM/RAM, magnetic disk, optical disk), comprising several instructions for causing a terminal device (which may be a mobile phone, a computer, a server, an air conditioner, or a network device, etc.) to perform the method described in the embodiments of the present application.
Although the embodiments of the present invention are described above, the embodiments are only used for facilitating understanding of the present invention, and are not intended to limit the present invention. Any person skilled in the art can make any modification and variation in form and detail without departing from the spirit and scope of the present disclosure, but the scope of the present disclosure is to be determined by the appended claims.
The foregoing is only an optional embodiment of the present application, and is not limited to the patent scope of the present application, and all equivalent structures or equivalent processes using the descriptions and the contents of the present application or directly or indirectly applied to other related technical fields are included in the patent protection scope of the present application.
Claims (6)
1. The hyperspectral image dimension reduction method is characterized by comprising the following steps of:
acquiring a neighborhood sub-cube f of a pixel in a hyperspectral image I to be processed using a sliding kernel i Wherein the size of I is m multiplied by n multiplied by k, m represents the number of spatial lines of the hyperspectral image, n represents the number of spatial columns of the hyperspectral image, k represents the spectral dimension of the hyperspectral image, the size of the sliding kernel is w multiplied by w, w is an odd number greater than 1, the value of I is a positive integer not greater than min (m-w+1, n-w+1), f i Is w×w×k;
for each acquired neighborhood sub-cube f i Performing dimension reduction processing to obtain a neighborhood hyperspectral image of each neighborhood subcube after dimension reduction
According to the spatial position of each neighborhood subcube in the hyperspectral image, the neighborhood hyperspectral image of each neighborhood subcubeRearranging to obtain the hyperspectral image +.>
Wherein the pair of acquired respective neighborhood subcubes f i Performing dimension reduction processing to obtain a neighborhood hyperspectral image of each neighborhood subcube after dimension reductionThe method comprises the following steps:
acquiring each neighborhood subcube f i Matrix M obtained after tiling i Wherein the matrix M i Is k×s, where s=w×w;
for the matrix M i Performing decomposition processing, and establishing a linear dynamic system model (A) of the neighborhood subcube according to the decomposition processing result i ,B i ,C i )；
According to the linear dynamic system model (A i ,B i ,C i ) In said matrix A i Sum matrix B i Acquisition matrix C i Is a sparse matrix of (2)
For the sparse matrixCarrying out normalization processing, and obtaining the neighborhood hyperspectral image ++of each neighborhood subcubes after dimension reduction according to the normalization processing result>
Wherein said pair of said matrices M i The decomposing treatment is carried out, which comprises the following steps:
decomposing the SDV with singular values to the matrix M by the following formula i Decomposing to obtain a decomposed matrix U i Matrix lambda i Sum matrix V i ：
M i ＝U i Λ i V i T ；
Wherein the matrix U i Is k x s and U i T ×U i =i, the matrix V i And the matrix lambda i Is s x s, V i T ×V i ＝I；
The linear dynamic system model (A) of the neighborhood subcube is built according to the decomposition processing result i ,B i ,C i ) The method comprises the following steps:
by using the matrix U i The matrix lambda i And the matrix V i A linear dynamic system model (A i ,B i ,C i ) The method specifically comprises the following steps:
C i ＝U i
wherein X is i ＝Λ i V i T ；
In calculation B i When first calculateThen utilize formula G i ＝U i 'Λ' i V i ' decomposition matrix G i Obtaining U' i Sum Λ'. i Thereby calculating to obtain B i ；
Wherein matrix A i Is s X s, matrix X i Is a system state variable matrix, the size of which is s x s,representing extraction of the matrix X i 1 to s rows and 2 to s columns, < ->Representing extraction of the matrix X i 1 to s rows and 1 to s-1 columns,/->Representing an extraction matrix U i ' column 1>Representing the extraction matrix Λ' i 1 st row 1 st column element of (c).
2. The method according to claim 1, characterized in that the method according to the linear dynamic system model (a i ,B i ,C i ) In said matrix A i Sum matrix B i Acquisition matrix C i Is a sparse matrix of (2)Comprising the following steps:
matrix is obtained by using Lyapunov stabilization condition and sparsity assumption condition through the following equationAnd sparse matrix->The optimal solution of the equation is the matrix +.>And the sparse matrix->
||Y i || col →min,
s.t.A i W i +W i A i T -B i Y i -Y i T B i T ＜0
Wherein W is i ＞0,W i ＝W i T ；
Computing a matrix
The sparse matrixFrom the matrix->Is equal to the sparse matrix->A row corresponding to a column index of not 0.
3. The method of claim 2, wherein the pair of sparse matricesThe normalization processing is carried out, which comprises the following steps:
establishing a zero matrixMatrix->Is filled into the matrix +.>The filled position is the row in the matrix +.>Row index of (2), filled +.>Namely +.>Normalization processing results after normalization processing;
obtaining the neighborhood hyperspectral image of each neighborhood subcube after dimension reduction according to the normalization processing resultThe method comprises the following steps:
obtaining the neighborhood hyperspectral image of each neighborhood subcube after dimension reduction through the following formula
4. A hyperspectral image dimension reduction device, comprising:
a neighborhood subcube acquisition module configured to acquire a neighborhood subcube f of a pixel in the hyperspectral image I to be processed using a sliding kernel i Wherein the dimension of I is m multiplied by n multiplied by k, m represents the number of spatial lines of the hyperspectral image, and n represents the heightThe number of spatial columns of the spectral image, k represents the spectral dimension of the hyperspectral image, the size of the sliding kernel is w×w, w is an odd number greater than 1, i is a positive integer not greater than min (m-w+1, n-w+1), f i Is w×w×k;
a neighborhood hyperspectral image generation module configured to generate, for each acquired neighborhood subcube f i Performing dimension reduction processing to obtain a neighborhood hyperspectral image of each neighborhood subcube after dimension reduction
A neighborhood hyperspectral image rearrangement module configured to rearrange neighborhood hyperspectral images of the respective neighborhood subcubes according to spatial positions of the respective neighborhood subcubes in the hyperspectral imagesRearranging to obtain the hyperspectral image +.>
The neighborhood hyperspectral image generation module comprises:
a neighborhood subcube matrix generation module configured to obtain the neighborhood subcubes f i Matrix M obtained after tiling i Wherein the matrix M i Is k×s, where s=w×w;
a linear dynamic system model building module configured to build the matrix M i Performing decomposition processing, and establishing a linear dynamic system model (A) of the neighborhood subcube according to the decomposition processing result i ,B i ,C i )；
A sparse matrix acquisition module arranged to acquire a sparse matrix based on the linear dynamic system model (A i ,B i ,C i ) In said matrix A i Sum matrix B i Acquisition matrix C i Is a sparse matrix of (2)
A sparse matrix processing module configured to process the sparse matrixCarrying out normalization processing, and obtaining the neighborhood hyperspectral image ++of each neighborhood subcubes after dimension reduction according to the normalization processing result>
Wherein said pair of said matrices M i Performing decomposition processing, including:
decomposing the SDV with singular values to the matrix M by the following formula i Decomposing to obtain a decomposed matrix U i Matrix lambda i Sum matrix V i ：
M i ＝U i Λ i V i T ；
Wherein the matrix U i Is k x s and U i T ×U i =i, the matrix V i And the matrix lambda i Is s x s, V i T ×V i ＝I；
The linear dynamic system model (A) of the neighborhood subcube is built according to the decomposition processing result i ,B i ,C i ) Comprising:
by using the matrix U i The matrix lambda i And the matrix V i A linear dynamic system model (A i ,B i ,C i ) The method specifically comprises the following steps:
C i ＝U i
wherein X is i ＝Λ i V i T ；
In calculation B i When first calculateThen utilize formula G i ＝U i 'Λ' i V i ' decomposition matrix G i Obtaining U' i Sum Λ'. i Thereby calculating to obtain B i ；
Wherein matrix A i Is s X s, matrix X i Is a system state variable matrix, the size of which is s x s,representing extraction of the matrix X i 1 to s rows and 2 to s columns, < ->Representing extraction of the matrix X i 1 to s rows and 1 to s-1 columns,/->Representing the extraction matrix U' i ' column 1>Representing the extraction matrix Λ' i 1 st row 1 st column element of (c).
5. The apparatus of claim 4, wherein the sparse matrix acquisition module is specifically configured to:
matrix is obtained by using Lyapunov stabilization condition and sparsity assumption condition through the following equationAnd sparse matrix->The optimal solution of the equation is the matrix +.>And the sparse matrix->
||Y i || col →min,
s.t.A i W i +W i A i T -B i Y i -Y i T B i T ＜0
Wherein W is i ＞0,W i ＝W i T ；
Computing a matrix
The sparse matrixFrom the matrix->Is equal to the sparse matrix->A row corresponding to a column index of not 0.
6. The apparatus of claim 5, wherein the sparse matrix processing module is specifically configured to:
establishing a zero matrixMatrix->Is filled into the matrix +.>The filled position is the row in the matrix +.>Row index of (2), filled +.>Namely +.>Normalization processing results after normalization processing;
obtaining the neighborhood hyperspectral image of each neighborhood subcube after dimension reduction according to the normalization processing resultComprising the following steps:
obtaining the neighborhood hyperspectral image of each neighborhood subcube after dimension reduction through the following formula
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911129254.2A CN110992245B (en) | 2019-11-18 | 2019-11-18 | Hyperspectral image dimension reduction method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911129254.2A CN110992245B (en) | 2019-11-18 | 2019-11-18 | Hyperspectral image dimension reduction method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110992245A CN110992245A (en) | 2020-04-10 |
| CN110992245B true CN110992245B (en) | 2023-07-25 |
Family
ID=70085048
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911129254.2A Active CN110992245B (en) | 2019-11-18 | 2019-11-18 | Hyperspectral image dimension reduction method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110992245B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114037899B (en) * | 2021-12-01 | 2024-09-06 | 福州大学 | Hyperspectral remote sensing image classification-oriented radial accumulation position coding system based on VIT |
Citations (3)
| 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 |
| CN108182449A (en) * | 2017-12-25 | 2018-06-19 | 南京信息工程大学 | A kind of hyperspectral image classification method |
| CN110263777A (en) * | 2019-06-26 | 2019-09-20 | 中国人民解放军火箭军工程大学 | Object detection method and system based on the locality preserving projections algorithm that empty spectrum combines |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8660360B1 (en) * | 2012-08-03 | 2014-02-25 | Raytheon Company | System and method for reduced incremental spectral clustering |
-
2019
- 2019-11-18 CN CN201911129254.2A patent/CN110992245B/en active Active
Patent Citations (3)
| 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 |
| CN108182449A (en) * | 2017-12-25 | 2018-06-19 | 南京信息工程大学 | A kind of hyperspectral image classification method |
| CN110263777A (en) * | 2019-06-26 | 2019-09-20 | 中国人民解放军火箭军工程大学 | Object detection method and system based on the locality preserving projections algorithm that empty spectrum combines |
Non-Patent Citations (1)
| Title |
|---|
| Zhangyang Wang 等.Deep Dual-Domain Based Fast Restoration of JPEG-Compressed Images.《arxiv.org/pdf/1601.04149.pdf》.2016,全文. * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110992245A (en) | 2020-04-10 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| He et al. | Sparsity-regularized robust non-negative matrix factorization for hyperspectral unmixing | |
| Iordache et al. | Collaborative sparse regression for hyperspectral unmixing | |
| Bonnefoy et al. | Dynamic screening: Accelerating first-order algorithms for the lasso and group-lasso | |
| US9317929B2 (en) | Decomposition apparatus and method for refining composition of mixed pixels in remote sensing images | |
| CN104008538B (en) | Based on single image super-resolution method | |
| US10268931B2 (en) | Spatiotemporal method for anomaly detection in dictionary learning and sparse signal recognition | |
| CN104952050B (en) | High spectrum image adaptive de mixing method based on region segmentation | |
| Xu et al. | Fast tensor nuclear norm for structured low-rank visual inpainting | |
| Chen et al. | Convolutional neural network based dem super resolution | |
| US8139650B2 (en) | Fast noise reduction in digital images and video | |
| Greenewald et al. | Robust kronecker product PCA for spatio-temporal covariance estimation | |
| Yuan et al. | Image quality assessment: A sparse learning way | |
| CN107491793B (en) | Polarized SAR image classification method based on sparse scattering complete convolution | |
| CN102682441A (en) | Hyperspectral image super-resolution reconstruction method based on subpixel mapping | |
| CN109002794B (en) | Nonlinear non-negative matrix factorization face recognition construction method, system and storage medium | |
| CN105184302B (en) | A kind of high optical spectrum image end member extraction method | |
| CN109657081A (en) | Distributed approach, system and the medium of EO-1 hyperion satellite remote sensing date | |
| CN113902622B (en) | Spectrum super-resolution method based on depth priori joint attention | |
| CN107316309A (en) | High spectrum image conspicuousness object detection method based on matrix decomposition | |
| Gkillas et al. | Connections between deep equilibrium and sparse representation models with application to hyperspectral image denoising | |
| CN107680081B (en) | Hyperspectral image unmixing method based on convolutional neural network | |
| Takeyama et al. | A constrained convex optimization approach to hyperspectral image restoration with hybrid spatio-spectral regularization | |
| CN110992245B (en) | Hyperspectral image dimension reduction method and device | |
| Cardoso et al. | Monte Carlo guided diffusion for Bayesian linear inverse problems | |
| Cohen | Environmental multiway data mining |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |