CN111126452A - Ground feature spectral curve expansion method and system based on principal component analysis - Google Patents
Ground feature spectral curve expansion method and system based on principal component analysis Download PDFInfo
- Publication number
- CN111126452A CN111126452A CN201911219327.7A CN201911219327A CN111126452A CN 111126452 A CN111126452 A CN 111126452A CN 201911219327 A CN201911219327 A CN 201911219327A CN 111126452 A CN111126452 A CN 111126452A
- Authority
- CN
- China
- Prior art keywords
- matrix
- principal component
- sample
- spectral
- space
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
- G06F18/2135—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods based on approximation criteria, e.g. principal component analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/58—Extraction of image or video features relating to hyperspectral data
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Multimedia (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Geophysics And Detection Of Objects (AREA)
- Investigating Or Analysing Materials By Optical Means (AREA)
Abstract
The invention discloses a ground feature spectral curve expanding method and system based on principal component analysis, wherein the method comprises the following steps: representing a plurality of spectral curves of a ground object type by a sample matrix, and calculating a mean vector and a covariance matrix of the sample matrix; calculating an eigenvalue and an eigenvector of the covariance matrix; selecting the first p principal component eigenvalues, and forming a new eigenvector matrix by corresponding eigenvectors; generating random samples of normal distribution in an uncorrelated space, and defining standard deviation of each sample set by using characteristic values of variables in a transformation space; and converting the sample sets back to the spectrum correlation space through a new eigenvector matrix to generate an expanded surface feature spectrum curve. The method can quickly generate any number of surface feature (material) spectral curves, and selects the principal component by using principal component analysis, thereby reducing the computational complexity and avoiding the waste of computational resources.
Description
Technical Field
The invention relates to the technical field of remote sensing image simulation, in particular to a ground feature spectral curve expansion method and system based on principal component analysis.
Background
The feature reflection spectrum refers to the rule that the reflectivity of a feature changes with the incident wavelength, and a curve drawn according to the reflection spectrum of the feature is called a feature reflection spectrum curve.
In a real scene, the shapes of the spectral curves of the same ground objects are basically consistent, but the spectral curves have different states (such as the healthy state and the diseased state of vegetation) and are more or less interfered by the external environment or the measuring instrument during the process of obtaining the reflection spectral curves of the ground objects. In the simulation process, if the influence of the interference is ignored, in the scene where the scene shows different laws (namely textures) in reality, the simulation scene does not have corresponding gray level fluctuation, and the textures close to the reality scene cannot be shown. However, the method of obtaining multiple spectral curves through multiple measurements is not only inefficient, but also has very limited detail information. There is a need for a method that can quickly generate a large number of curves from a small number of curves.
Schott et al, j.r. of the rochess institute of technology, usa, have proposed a method that can generate an arbitrary number of curves, but this method has a high dimensionality of the operational data (the dimensionality of the hyperspectral data may even reach several hundred), and the number of curves that need to be generated in the application is large (often several hundred), so that the amount of computation is rapidly increased, and a large amount of operational resources are occupied. Therefore, it is necessary to develop a curve expansion technique with simpler and faster operation.
Principal Component Analysis (PCA) is a commonly used data Analysis algorithm that transforms raw data into a set of representations linearly independent of each dimension through linear transformation for extracting Principal feature components of the data, often for dimensionality reduction of high-dimensional data. In short, the data in the high-dimensional space is projected onto the low-dimensional space to realize the dimension reduction of the data.
Disclosure of Invention
The invention aims to overcome the technical defects and provides a feature spectral curve expansion method based on PCA, which considers the transition region of an image by expanding a curve set, embodies rich detail information of the image, increases the sense of reality of a scene image and provides a convenient and effective way for rapidly and vividly acquiring the image. Even at the initial stage of texture character modeling, a more realistic effect can be obtained.
In order to achieve the aim, the invention discloses a ground feature spectral curve expanding method based on principal component analysis, which comprises the following steps:
representing a plurality of spectral curves of a ground object type by a sample matrix, and calculating a mean vector and a covariance matrix of the sample matrix;
calculating an eigenvalue and an eigenvector of the covariance matrix;
selecting the first p principal component eigenvalues, and forming a new eigenvector matrix by corresponding eigenvectors;
generating random samples of normal distribution in an uncorrelated space, and defining standard deviation of each sample set by using characteristic values of variables in a transformation space;
and converting the sample sets back to the spectrum correlation space through a new eigenvector matrix to generate an expanded surface feature spectrum curve.
As an improvement of the above method, the plurality of spectral curves of a ground object type represent a sample matrix, and a mean vector and a covariance matrix of the sample matrix are calculated; the method specifically comprises the following steps:
representing the m spectral curves of a known ground object type as a sample matrix X as follows:
wherein n represents the number of bands of each spectral curve;
wherein, mukIs the mean of the kth points on all spectral curves, k being 1, 2.. n; sigmai,jIs the covariance of the ith and jth spectral means of the species, i 1, 2.
As an improvement of the above method, the calculating the eigenvalue and the eigenvector of the covariance matrix specifically includes:
calculating a vector consisting of eigenvalues of a covariance matrix sigmaAnd a matrix of feature vectors
Wherein λ is1≥λ2≥λ3…λnIs not less than 0, and for this type of terrain, lambdaiIs thatThe ith column of feature vectors inThe characteristic value of (2).
As an improvement of the above method, the selecting the first p principal component eigenvalues and forming a new eigenvector matrix by using the corresponding eigenvectors thereof specifically includes:
As an improvement of the above method, random samples of normal distribution are generated in the uncorrelated space, and the standard deviation of each sample set is defined using the eigenvalues of the variables in the transform space; the method specifically comprises the following steps:
a set of gaussian distributed random numbers y is generatedkThe distribution of which satisfies N (0, λ) respectivelyk) Wherein k is 1,2,3 …, n; form the following vector
As an improvement of the above method, the sample sets are converted back to the spectrum correlation space through a new eigenvector matrix to generate an extended surface feature spectrum curve, specifically:
Then its distribution is Nn Since the random variables exhibit the same spectral characteristics as the basis setTo represent an extended spectral curve.
The invention also provides a ground feature spectral curve expanding system based on principal component analysis, which comprises:
the mean value and covariance calculation module is used for representing a plurality of spectral curves of a ground object type to a sample matrix and calculating a mean value vector and a covariance matrix of the sample matrix;
the eigenvalue and eigenvector calculation module is used for calculating the eigenvalue and eigenvector of the covariance matrix;
the new characteristic vector matrix construction module is used for selecting the first p principal component characteristic values and constructing a new characteristic vector matrix by the corresponding characteristic vectors;
a sample set generating module, configured to generate a normally distributed random sample in an uncorrelated space, and define a standard deviation of each sample set by using a characteristic value of a variable in a transformed space;
and the extended surface feature spectral curve generation module is used for converting the sample sets back to the spectral correlation space through the new eigenvector matrix to generate an extended surface feature spectral curve.
The invention has the advantages that:
1. the method can quickly generate any number of surface feature (material) spectral curves, and the PCA is used for selecting the principal component, so that the calculation complexity is reduced, and the waste of calculation resources is avoided;
2. according to the method, the principal components are selected by adopting the PCA algorithm, so that the dimensionality of data in the operation process is reduced, and the calculation complexity can be effectively reduced compared with that of the original method; the hyperspectral data reaches hundreds of wave bands, and excessive computing resources can be avoided by selecting the main components;
3. the method of the invention considers the transition region of the image by expanding the curve set, embodies the rich detail information of the image, increases the sense of reality of the scene image, and provides a convenient and effective way for rapidly and vividly acquiring the image; even at the initial stage of texture character modeling, a more realistic effect can be obtained.
Drawings
FIG. 1 is a flow chart of a feature spectral curve extending method based on principal component analysis according to the present invention.
FIG. 2 is a diagram showing the principal components and accuracy of gold (gold) selection;
FIG. 3(a) is a raw spectral plot of aluminum;
FIG. 3(b) is an expanded set of curves;
FIG. 4(a) is a plot of the raw spectrum of titanium;
FIG. 4(b) is an expanded set of curves;
FIG. 5(a) is a raw spectral curve of stainless steel;
FIG. 5(b) is an expanded set of curves;
FIG. 6(a) is a raw spectral plot of molybdenum;
fig. 6(b) is a set of curves after expansion.
Detailed Description
The technical solution of the present invention will be described in detail below with reference to the accompanying drawings.
The method of the present invention can generate any number of spectral curves from a smaller set of curves that contain the desired multi-variable analysis for a given land cover category. The method entails generating an average vector for each land cover category and a covariance matrix containing the variables for each spectral point. The zero center point is obtained by subtracting the mean vector and then converting it to a spectrally uncorrelated space. The creation of a new curve involves generating random samples of a normal distribution in the uncorrelated space, using the eigenvalues of the variables in the transform space to define the standard deviation for each set of samples. These sample sets (vectors) are then converted back into the spectral correlation space where they exhibit the same spectral characteristics as the base set.
As shown in fig. 1, the present invention provides a feature spectral curve extending method based on principal component analysis, including:
according to a set of spectral curves of actual materials (here, the metals aluminum, titanium, stainless steel and molybdenum are taken as examples), the distribution is assumed to conform to a normal distribution N47 (i.e., an average value ofA multidimensional normal distribution with covariance ∑), where the dimensionality of the multidimensional normal distribution is due to the example data being 47 bands per spectral curve.
The known data can be represented as a sample matrix, in this case 3 x 47 in size:
the mean vector is:
zero-averaging each row of the sample matrix, i.e. subtracting the average of each column as:
102, calculating a covariance matrix of a sample matrix;
as shown in the formula, the matrix is a real symmetric matrix, the elements on the main diagonal of the matrix represent the variance of the objects, and the rest elements represent the covariance between the objects.
103, calculating an eigenvalue and an eigenvector of the covariance matrix;
λ here1≥λ2≥λ3…λ47Not less than 0, and for such types of ground objects (material substances), lambdaiIs thatThe eigenvalue of the i-th column eigenvector in (a).
104, selecting a proper amount of main components;
comprehensively considering the calculation complexity and the calculation accuracy, selecting proper (the first k) principal components (eigenvalues), and forming a new eigenvector matrix by using the corresponding k eigenvectors as column vectors
And 105, generating any number of curves according to the selected principal components.
thereinIs an original spectrum curve, and obtains a data set with irrelevant spectrumThese spectrally uncorrelated data distributions satisfy N396 Covariance matrixThe following forms:
covariance matrix of the spectrally uncorrelated dataIs used to generate a distribution satisfying N47 Is used to generate the multi-dimensional random variable. To accomplish this, a set of Gaussian-distributed random numbers y is generatediThe distribution satisfies N (0, λ)i) (where i ═ 1,2,3, …,47), the following vector is formed:
the fitted curve can be back-calculated according to equation (1) above:
to obtainOne distribution is N47 A plurality of random variables, which can be used to represent a spectral curve.
Setting the accuracy of the principal component to be more than 90% aiming at the reflectivity curve of the metal gold, and analyzing the calculation complexity of equation (2): as shown in fig. 2, for the reflectivity of gold (gold), there are only two principal components extracted from the covariance matrix of its sample matrix, but the accuracy of these two principal components is already over 90%, and the computational complexity before and after using the PCA algorithm is shown in table 1. In addition, in practical application, it is common to generate hundreds or even thousands of spectral curves at a time, so that the calculation complexity can be effectively reduced by adopting the method compared with the original method. Moreover, the hyperspectral data reaches hundreds of wave bands, excessive computing resources can be occupied without selecting the main components, and the computing resources are wasted.
Table 1: dimension comparison using operation matrices before and after PCA
In order to verify the method of the invention, a plurality of materials of metal aluminum, titanium, stainless steel and molybdenum are selected as experimental objects, the known spectral curve of each object is 3, and each spectral curve has 47 wave bands. As is clear from the relevant materials, it is not uncommon to generate hundreds of spectral curves in practical applications, and the effect achieved by generating 1000 curves with principal component accuracy of 90% is shown in fig. 3(a), 3(b), 4(a), 4(b), 5(a), 5(b), 6(a) and 6(b), as an example.
Finally, it should be noted that the above embodiments are only used for illustrating the technical solutions of the present invention and are not limited. Although the present invention has been described in detail with reference to the embodiments, it will be understood by those skilled in the art that various changes may be made and equivalents may be substituted without departing from the spirit and scope of the invention as defined in the appended claims.
Claims (7)
1. A feature spectral curve extending method based on principal component analysis, the method comprising:
representing a plurality of spectral curves of a ground object type by a sample matrix, and calculating a mean vector and a covariance matrix of the sample matrix;
calculating an eigenvalue and an eigenvector of the covariance matrix;
selecting the first p principal component eigenvalues, and forming a new eigenvector matrix by corresponding eigenvectors;
generating random samples of normal distribution in an uncorrelated space, and defining standard deviation of each sample set by using characteristic values of variables in a transformation space;
and converting the sample sets back to the spectrum correlation space through a new eigenvector matrix to generate an expanded surface feature spectrum curve.
2. The principal component analysis-based feature spectral curve extending method according to claim 1, wherein the plurality of spectral curves of a feature type are represented by a sample matrix, and a mean vector and a covariance matrix of the sample matrix are calculated; the method specifically comprises the following steps:
representing the m spectral curves of a known ground object type as a sample matrix X as follows:
wherein n represents the number of bands of each spectral curve;
wherein, mukIs the mean of the kth points on all spectral curves, k being 1, 2.. n; sigmai,jIs the covariance of the ith and jth spectral means of the species, i 1, 2.
3. The principal component analysis-based surface feature spectral curve extending method according to claim 2, wherein the eigenvalues and eigenvectors of the covariance matrix are calculated, specifically:
4. The principal component analysis-based surface feature spectral curve extending method according to claim 3, wherein the selecting the first p principal component eigenvalues and forming a new eigenvector matrix by using their corresponding eigenvectors specifically comprises:
5. The principal component analysis-based surface feature spectral curve extending method according to claim 4, wherein random samples of normal distribution are generated in a non-correlated space, and the standard deviation of each sample set is defined using the eigenvalues of variables in a transformed space; the method specifically comprises the following steps:
a set of gaussian distributed random numbers y is generatedkThe distribution of which satisfies N (0, λ) respectivelyk) Wherein k is 1,2,3 …, n; form the following vector
6. The principal component analysis-based surface feature spectral curve expansion method according to claim 1, wherein the sample sets are converted back to the spectrum correlation space through a new eigenvector matrix to generate an expanded surface feature spectral curve, specifically:
7. A feature spectral curve extension system based on principal component analysis, the system comprising:
the mean value and covariance calculation module is used for representing a plurality of spectral curves of a ground object type to a sample matrix and calculating a mean value vector and a covariance matrix of the sample matrix;
the eigenvalue and eigenvector calculation module is used for calculating the eigenvalue and eigenvector of the covariance matrix;
the new characteristic vector matrix construction module is used for selecting the first p principal component characteristic values and constructing a new characteristic vector matrix by the corresponding characteristic vectors;
a sample set generating module, configured to generate a normally distributed random sample in an uncorrelated space, and define a standard deviation of each sample set by using a characteristic value of a variable in a transformed space;
and the extended surface feature spectral curve generation module is used for converting the sample sets back to the spectral correlation space through the new eigenvector matrix to generate an extended surface feature spectral curve.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201911219327.7A CN111126452B (en) | 2019-12-03 | 2019-12-03 | Feature spectrum curve expansion method and system based on principal component analysis |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201911219327.7A CN111126452B (en) | 2019-12-03 | 2019-12-03 | Feature spectrum curve expansion method and system based on principal component analysis |
Publications (2)
Publication Number | Publication Date |
---|---|
CN111126452A true CN111126452A (en) | 2020-05-08 |
CN111126452B CN111126452B (en) | 2023-05-23 |
Family
ID=70497191
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201911219327.7A Active CN111126452B (en) | 2019-12-03 | 2019-12-03 | Feature spectrum curve expansion method and system based on principal component analysis |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN111126452B (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113033013A (en) * | 2021-04-08 | 2021-06-25 | 北京环境特性研究所 | Infrared smoke screen spectrum transmittance simulation method |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US6879716B1 (en) * | 1999-10-20 | 2005-04-12 | Fuji Photo Film Co., Ltd. | Method and apparatus for compressing multispectral images |
CN102609703A (en) * | 2012-03-05 | 2012-07-25 | 中国科学院对地观测与数字地球科学中心 | Method and device for detecting target ground object in hyperspectral image |
CN102938072A (en) * | 2012-10-20 | 2013-02-20 | 复旦大学 | Dimension reducing and sorting method of hyperspectral imagery based on blocking low rank tensor analysis |
CN107122799A (en) * | 2017-04-25 | 2017-09-01 | 西安电子科技大学 | Hyperspectral image classification method based on expanding morphology and Steerable filter |
CN108896499A (en) * | 2018-05-09 | 2018-11-27 | 西安建筑科技大学 | In conjunction with principal component analysis and the polynomial spectral reflectance recovery method of regularization |
CN109508647A (en) * | 2018-10-22 | 2019-03-22 | 北京理工大学 | A kind of spectra database extended method based on generation confrontation network |
-
2019
- 2019-12-03 CN CN201911219327.7A patent/CN111126452B/en active Active
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US6879716B1 (en) * | 1999-10-20 | 2005-04-12 | Fuji Photo Film Co., Ltd. | Method and apparatus for compressing multispectral images |
CN102609703A (en) * | 2012-03-05 | 2012-07-25 | 中国科学院对地观测与数字地球科学中心 | Method and device for detecting target ground object in hyperspectral image |
CN102938072A (en) * | 2012-10-20 | 2013-02-20 | 复旦大学 | Dimension reducing and sorting method of hyperspectral imagery based on blocking low rank tensor analysis |
CN107122799A (en) * | 2017-04-25 | 2017-09-01 | 西安电子科技大学 | Hyperspectral image classification method based on expanding morphology and Steerable filter |
CN108896499A (en) * | 2018-05-09 | 2018-11-27 | 西安建筑科技大学 | In conjunction with principal component analysis and the polynomial spectral reflectance recovery method of regularization |
CN109508647A (en) * | 2018-10-22 | 2019-03-22 | 北京理工大学 | A kind of spectra database extended method based on generation confrontation network |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113033013A (en) * | 2021-04-08 | 2021-06-25 | 北京环境特性研究所 | Infrared smoke screen spectrum transmittance simulation method |
CN113033013B (en) * | 2021-04-08 | 2023-04-14 | 北京环境特性研究所 | Infrared smoke screen spectrum transmittance simulation method |
Also Published As
Publication number | Publication date |
---|---|
CN111126452B (en) | 2023-05-23 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110210313B (en) | Hyperspectral remote sensing image classification method based on multi-scale PCA-3D-CNN (principal component analysis-three dimensional-CNN) space spectrum combination | |
JP7315748B2 (en) | Data classifier training method, data classifier training device, program and training method | |
US20190164047A1 (en) | Object recognition using a convolutional neural network trained by principal component analysis and repeated spectral clustering | |
EP2880596B1 (en) | System and method for reduced incremental spectral clustering | |
CN109190511B (en) | Hyperspectral classification method based on local and structural constraint low-rank representation | |
CN110866439B (en) | Hyperspectral image joint classification method based on multi-feature learning and super-pixel kernel sparse representation | |
CN109727295B (en) | Electromagnetic image extraction method, electromagnetic image extraction device, computer equipment and storage medium | |
Glumov et al. | Detection of objects on the image using a sliding window mode | |
CN110687514B (en) | Nonlinear discrimination learning true and false target one-dimensional range profile feature extraction method | |
CN112784907A (en) | Hyperspectral image classification method based on spatial spectral feature and BP neural network | |
JPH04299469A (en) | Recognizing method for substance in image and application on substance tracking in series of images | |
CN111126452B (en) | Feature spectrum curve expansion method and system based on principal component analysis | |
CN112819769B (en) | Nonlinear hyperspectral image anomaly detection algorithm based on kernel function and joint dictionary | |
Fursov et al. | Thematic classification with support subspaces in hyperspectral images | |
CN112784747A (en) | Multi-scale eigen decomposition method for hyperspectral remote sensing image | |
CN111105363A (en) | Rapid unmixing method for noisy hyperspectral image | |
Chung et al. | Recursive K-distribution parameter estimation | |
CN113486869B (en) | Method, device and medium for lithology identification based on unsupervised feature extraction | |
CN111985501B (en) | Hyperspectral image feature extraction method based on self-adaptive high-order tensor decomposition | |
CN114295600A (en) | Improved Raman spectrum multivariate data analysis and imaging method | |
Zahedi et al. | 3-D color histogram equalization by principal component analysis | |
CN109145132B (en) | Method and device for extracting hash code from image and image retrieval method and device | |
KR100397467B1 (en) | Simulation method for correlating probability variables and computer-readable Medium having stored the method | |
JP4049288B2 (en) | Canonical discriminant analyzer, feature extractor and recording medium | |
Priyadharshini et al. | Characterization of Environmental Covariates of Coimbatore District Using Principal Component Analysis |
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 |