A meshless approach, called the rigid-plastic reproducing kernel particle method （RKPM）, is presented for three-dimensional （3D） bulk metal forming simulation. The approach is a combination of RKPM with the flow t...A meshless approach, called the rigid-plastic reproducing kernel particle method （RKPM）, is presented for three-dimensional （3D） bulk metal forming simulation. The approach is a combination of RKPM with the flow theory of 3D rigid-plastic mechanics. For the treatments of essential boundary conditions and incompressibility constraint, the boundary singular kernel method and the modified penalty method are utilized, respectively. The arc-tangential friction model is employed to treat the contact conditions. The compression of rectangular blocks, a typical 3D upsetting operation, is analyzed for different friction conditions and the numerical results are compared with those obtained using commercial rigid-plastic FEM （finite element method） software Deform^3D. As results show, when handling 3D plastic deformations, the proposed approach eliminates the need of expensive meshing and remeshing procedures which are unavoidable in conventional FEM and can provide results that are in good agreement with finite element predictions.展开更多
Dimensionality reduction techniques play an important role in data mining. Kernel entropy component analysis( KECA) is a newly developed method for data transformation and dimensionality reduction. This paper conducte...Dimensionality reduction techniques play an important role in data mining. Kernel entropy component analysis( KECA) is a newly developed method for data transformation and dimensionality reduction. This paper conducted a comparative study of KECA with other five dimensionality reduction methods,principal component analysis( PCA),kernel PCA( KPCA),locally linear embedding( LLE),laplacian eigenmaps( LAE) and diffusion maps( DM). Three quality assessment criteria, local continuity meta-criterion( LCMC),trustworthiness and continuity measure(T&C),and mean relative rank error( MRRE) are applied as direct performance indexes to assess those dimensionality reduction methods. Moreover,the clustering accuracy is used as an indirect performance index to evaluate the quality of the representative data gotten by those methods. The comparisons are performed on six datasets and the results are analyzed by Friedman test with the corresponding post-hoc tests. The results indicate that KECA shows an excellent performance in both quality assessment criteria and clustering accuracy assessing.展开更多
A turbulent separation-rcattachment flow in a two-dimensional asymmetrical curved-wall diffuser is studied by a two-dimensional laser doppler velocimeter.The turbulent boundary layer separates on the lower curved wall...A turbulent separation-rcattachment flow in a two-dimensional asymmetrical curved-wall diffuser is studied by a two-dimensional laser doppler velocimeter.The turbulent boundary layer separates on the lower curved wall under strong pressure gradient and then reattaches on a parallel channel.At the inlet of the diffuser,Reynolds number based on the diffuser height is 1.2×10~5 and the velocity is 25.2m/s.The re- sults of experiments are presented and analyzed in new defined streamline-aligned coordinates.The experiment shows that after Transitory Detachment Reynolds shear stress is negative in the near-wall backflow region. Their characteristics are approximately the same as in simple turbulent shear layers near the maximum Reynolds shear stress.A scale is formed using the maximum Reynolds shear stresses.It is found that a Reynolds shear stress similarity exists from separation to reattachment and the Schofield-Perry velocity law ex- ists in the forward shear flow.Both profiles are used in the experimental work that leads to the design of a new eddy-viscosity model.The length scale is taken from that developed by Schofield and Perry.The composite velocity scale is formed by the maximum Reynolds shear stress and the Schofield Perry velocity scale as well as the edge velocity of the boundary layer.The results of these experiments are presented in this paper展开更多
In this paper, an upper bound of fractal dimension of the compact kernel sections for the dissipative non-autonomous Klein-Gordon-Schr<span style="white-space:nowrap;">ö</span>dinger lat...In this paper, an upper bound of fractal dimension of the compact kernel sections for the dissipative non-autonomous Klein-Gordon-Schr<span style="white-space:nowrap;">ö</span>dinger lattice system is obtained, by applying a criterion for estimating fractal dimension of a family of compact subsets of a separable Hilbert space.展开更多
In the need of some real applications, such as text categorization and image classification, the multi-label learning gradually becomes a hot research point in recent years. Much attention has been paid to the researc...In the need of some real applications, such as text categorization and image classification, the multi-label learning gradually becomes a hot research point in recent years. Much attention has been paid to the research of multi-label classification algorithms. Considering the fact that the high dimensionality of the multi-label datasets may cause the curse of dimensionality and wil hamper the classification process, a dimensionality reduction algorithm, named multi-label kernel discriminant analysis （MLKDA）, is proposed to reduce the dimensionality of multi-label datasets. MLKDA, with the kernel trick, processes the multi-label integrally and realizes the nonlinear dimensionality reduction with the idea similar with linear discriminant analysis （LDA）. In the classification process of multi-label data, the extreme learning machine （ELM） is an efficient algorithm in the premise of good accuracy. MLKDA, combined with ELM, shows a good performance in multi-label learning experiments with several datasets. The experiments on both static data and data stream show that MLKDA outperforms multi-label dimensionality reduction via dependence maximization （MDDM） and multi-label linear discriminant analysis （MLDA） in cases of balanced datasets and stronger correlation between tags, and ELM is also a good choice for multi-label classification.展开更多
This paper studies the problem applying Radial Basis Function Network(RBFN) which is trained by the Recursive Least Square Algorithm(RLSA) to the recognition of one dimensional images of radar targets. The equivalence...This paper studies the problem applying Radial Basis Function Network(RBFN) which is trained by the Recursive Least Square Algorithm(RLSA) to the recognition of one dimensional images of radar targets. The equivalence between the RBFN and the estimate of Parzen window probabilistic density is proved. It is pointed out that the I/O functions in RBFN hidden units can be generalized to general Parzen window probabilistic kernel function or potential function, too. This paper discusses the effects of the shape parameter a in the RBFN and the forgotten factor A in RLSA on the results of the recognition of three kinds of kernel function such as Gaussian, triangle, double-exponential, at the same time, also discusses the relationship between A and the training time in the RBFN.展开更多
Ear morphological traits such as volume and shape are important features of maize and the quantitative associations among them can help understand kernel yield determination. 150 mature ears each of 4 maize cultivars ...Ear morphological traits such as volume and shape are important features of maize and the quantitative associations among them can help understand kernel yield determination. 150 mature ears each of 4 maize cultivars were collected from field experiments, and ear length(L), diameter(D), area(S) and volume(V) were recorded for individual ears, kernel weight per ear also recorded for a portion of the examined ears. Following principles of dimensional analysis, 8 theoretical equations of 3 sets,which relate ear higher dimensions to its length and diameter, were developed and parameterized and validated with the field observations. The 3 optimized equations showed that the shape of ears in maize can be featured with 3 dimensionless form factors, namely diameter-to-length ratio(c=D/L), areal form factor(b=S/L/D), and volumetric form factor(a=V/L/D/D). Statistically,all of them were significantly different among cultivars, and a's values varied from 0.582 to 0.612, and b's 0.839-0.868, and c's 0.242-0.308. Volumetric form factor and areal form factor could estimate precisely ear volume and area respectively, but diameter-to-length ratio was not suitable to estimate ear diameter by its length. Ear volume explained almost all variation of ear kernel weight and product L*D*D did the same substantially. Dimensional analysis proved to be promising in understanding relationship among morphological traits of ears in maize. Its application in crop researches should improve our knowledge of the physical properties of crop plants.展开更多
Approximate Bayesian Computation (ABC) is a popular sampling method in applications involving intractable likelihood functions. Instead of evaluating the likelihood function, ABC approximates the posterior distributio...Approximate Bayesian Computation (ABC) is a popular sampling method in applications involving intractable likelihood functions. Instead of evaluating the likelihood function, ABC approximates the posterior distribution by a set of accepted samples which are simulated from a generating model. Simulated samples are accepted if the distances between the samples and the observation are smaller than some threshold. The distance is calculated in terms of summary statistics. This paper proposes Local Gradient Kernel Dimension Reduction (LGKDR) to construct low dimensional summary statistics for ABC. The proposed method identifies a sufficient subspace of the original summary statistics by implicitly considering all non-linear transforms therein, and a weighting kernel is used for the concentration of the projections. No strong assumptions are made on the marginal distributions, nor the regression models, permitting usage in a wide range of applications. Experiments are done with simple rejection ABC and sequential Monte Carlo ABC methods. Results are reported as competitive in the former and substantially better in the latter cases in which Monte Carlo errors are compressed as much as possible.展开更多
The precision of the kernel independent component analysis( KICA) algorithm depends on the type and parameter values of kernel function. Therefore,it's of great significance to study the choice method of KICA'...The precision of the kernel independent component analysis( KICA) algorithm depends on the type and parameter values of kernel function. Therefore,it's of great significance to study the choice method of KICA's kernel parameters for improving its feature dimension reduction result. In this paper, a fitness function was established by use of the ideal of Fisher discrimination function firstly. Then the global optimal solution of fitness function was searched by particle swarm optimization( PSO) algorithm and a multi-state information dimension reduction algorithm based on PSO-KICA was established. Finally,the validity of this algorithm to enhance the precision of feature dimension reduction has been proven.展开更多
Finding a suitable space is one of the most critical problems for dimensionality reduction. Each space corresponds to a distance metric defined on the sample attributes, and thus finding a suitable space can be conver...Finding a suitable space is one of the most critical problems for dimensionality reduction. Each space corresponds to a distance metric defined on the sample attributes, and thus finding a suitable space can be converted to develop an effective distance metric. Most existing dimensionality reduction methods use a fixed pre-specified distance metric. However, this easy treatment has some limitations in practice due to the fact the pre-specified metric is not going to warranty that the closest samples are the truly similar ones. In this paper, we present an adaptive metric learning method for dimensionality reduction, called AML. The adaptive metric learning model is developed by maximizing the difference of the distances between the data pairs in cannot-links and those in must-links. Different from many existing papers that use the traditional Euclidean distance, we use the more generalized l<sub>2,p</sub>-norm distance to reduce sensitivity to noise and outliers, which incorporates additional flexibility and adaptability due to the selection of appropriate p-values for different data sets. Moreover, considering traditional metric learning methods usually project samples into a linear subspace, which is overstrict. We extend the basic linear method to a more powerful nonlinear kernel case so that well capturing complex nonlinear relationship between data. To solve our objective, we have derived an efficient iterative algorithm. Extensive experiments for dimensionality reduction are provided to demonstrate the superiority of our method over state-of-the-art approaches.展开更多
Space object recognition plays an important role in spatial exploitation and surveillance, followed by two main problems： lacking of data and drastic changes in viewpoints. In this article, firstly, we build a three-...Space object recognition plays an important role in spatial exploitation and surveillance, followed by two main problems： lacking of data and drastic changes in viewpoints. In this article, firstly, we build a three-dimensional （3D） satellites dataset named BUAA Satellite Image Dataset （BUAA-SID 1.0） to supply data for 3D space object research. Then, based on the dataset, we propose to recognize full-viewpoint 3D space objects based on kernel locality preserving projections （KLPP）. To obtain more accurate and separable description of the objects, firstly, we build feature vectors employing moment invariants, Fourier descriptors, region covariance and histogram of oriented gradients. Then, we map the features into kernel space followed by dimensionality reduction using KLPP to obtain the submanifold of the features. At last, k-nearest neighbor （kNN） is used to accomplish the classification. Experimental results show that the proposed approach is more appropriate for space object recognition mainly considering changes of viewpoints. Encouraging recognition rate could be obtained based on images in BUAA-SID 1.0, and the highest recognition result could achieve 95.87%.展开更多
To separate each pattern class more strongly and deal with nonlinear ease, a new nonlinear manifold learning algorithm named supervised kernel uneorrelated diseriminant neighborhood preserving projections （SKUDNPP） ...To separate each pattern class more strongly and deal with nonlinear ease, a new nonlinear manifold learning algorithm named supervised kernel uneorrelated diseriminant neighborhood preserving projections （SKUDNPP） is proposed. The algorithm utilizes supervised weight and kernel technique which makes the algorithm cope with classifying and nonlinear problems competently. The within-class geometric structure is preserved, while maximizing the between-class distance. And the features extracted are statistically uneorrelated by introducing an uneorrelated constraint. Experiment results on millimeter wave （MMW） radar target recognition show that the method can give competitive results in comparison with current papular algorithms.展开更多
Hyperspectral image provides abundant spectral information for remote discrimination of subtle differences in ground covers.However,the increasing spectral dimensions,as well as the information redundancy,make the ana...Hyperspectral image provides abundant spectral information for remote discrimination of subtle differences in ground covers.However,the increasing spectral dimensions,as well as the information redundancy,make the analysis and interpretation of hyperspectral images a challenge.Feature extraction is a very important step for hyperspectral image processing.Feature extraction methods aim at reducing the dimension of data,while preserving as much information as possible.Particularly,nonlinear feature extraction methods (e.g.kernel minimum noise fraction (KMNF) transformation) have been reported to benefit many applications of hyperspectral remote sensing,due to their good preservation of high-order structures of the original data.However,conventional KMNF or its extensions have some limitations on noise fraction estimation during the feature extraction,and this leads to poor performances for post-applications.This paper proposes a novel nonlinear feature extraction method for hyperspectral images.Instead of estimating noise fraction by the nearest neighborhood information (within a sliding window),the proposed method explores the use of image segmentation.The approach benefits both noise fraction estimation and information preservation,and enables a significant improvement for classification.Experimental results on two real hyperspectral images demonstrate the efficiency of the proposed method.Compared to conventional KMNF,the improvements of the method on two hyperspectral image classification are 8 and 11％.This nonlinear feature extraction method can be also applied to other disciplines where high-dimensional data analysis is required.展开更多
基金This work was supported by the National Natural Science Foundation of China （No. 50275094）.
文摘A meshless approach, called the rigid-plastic reproducing kernel particle method （RKPM）, is presented for three-dimensional （3D） bulk metal forming simulation. The approach is a combination of RKPM with the flow theory of 3D rigid-plastic mechanics. For the treatments of essential boundary conditions and incompressibility constraint, the boundary singular kernel method and the modified penalty method are utilized, respectively. The arc-tangential friction model is employed to treat the contact conditions. The compression of rectangular blocks, a typical 3D upsetting operation, is analyzed for different friction conditions and the numerical results are compared with those obtained using commercial rigid-plastic FEM （finite element method） software Deform^3D. As results show, when handling 3D plastic deformations, the proposed approach eliminates the need of expensive meshing and remeshing procedures which are unavoidable in conventional FEM and can provide results that are in good agreement with finite element predictions.
基金Climbing Peak Discipline Project of Shanghai Dianji University,China(No.15DFXK02)Hi-Tech Research and Development Programs of China(No.2007AA041600)
文摘Dimensionality reduction techniques play an important role in data mining. Kernel entropy component analysis( KECA) is a newly developed method for data transformation and dimensionality reduction. This paper conducted a comparative study of KECA with other five dimensionality reduction methods,principal component analysis( PCA),kernel PCA( KPCA),locally linear embedding( LLE),laplacian eigenmaps( LAE) and diffusion maps( DM). Three quality assessment criteria, local continuity meta-criterion( LCMC),trustworthiness and continuity measure(T&C),and mean relative rank error( MRRE) are applied as direct performance indexes to assess those dimensionality reduction methods. Moreover,the clustering accuracy is used as an indirect performance index to evaluate the quality of the representative data gotten by those methods. The comparisons are performed on six datasets and the results are analyzed by Friedman test with the corresponding post-hoc tests. The results indicate that KECA shows an excellent performance in both quality assessment criteria and clustering accuracy assessing.
文摘A turbulent separation-rcattachment flow in a two-dimensional asymmetrical curved-wall diffuser is studied by a two-dimensional laser doppler velocimeter.The turbulent boundary layer separates on the lower curved wall under strong pressure gradient and then reattaches on a parallel channel.At the inlet of the diffuser,Reynolds number based on the diffuser height is 1.2×10~5 and the velocity is 25.2m/s.The re- sults of experiments are presented and analyzed in new defined streamline-aligned coordinates.The experiment shows that after Transitory Detachment Reynolds shear stress is negative in the near-wall backflow region. Their characteristics are approximately the same as in simple turbulent shear layers near the maximum Reynolds shear stress.A scale is formed using the maximum Reynolds shear stresses.It is found that a Reynolds shear stress similarity exists from separation to reattachment and the Schofield-Perry velocity law ex- ists in the forward shear flow.Both profiles are used in the experimental work that leads to the design of a new eddy-viscosity model.The length scale is taken from that developed by Schofield and Perry.The composite velocity scale is formed by the maximum Reynolds shear stress and the Schofield Perry velocity scale as well as the edge velocity of the boundary layer.The results of these experiments are presented in this paper
文摘In this paper, an upper bound of fractal dimension of the compact kernel sections for the dissipative non-autonomous Klein-Gordon-Schr<span style="white-space:nowrap;">ö</span>dinger lattice system is obtained, by applying a criterion for estimating fractal dimension of a family of compact subsets of a separable Hilbert space.
基金supported by the National Natural Science Foundation of China(5110505261173163)the Liaoning Provincial Natural Science Foundation of China(201102037)
文摘In the need of some real applications, such as text categorization and image classification, the multi-label learning gradually becomes a hot research point in recent years. Much attention has been paid to the research of multi-label classification algorithms. Considering the fact that the high dimensionality of the multi-label datasets may cause the curse of dimensionality and wil hamper the classification process, a dimensionality reduction algorithm, named multi-label kernel discriminant analysis （MLKDA）, is proposed to reduce the dimensionality of multi-label datasets. MLKDA, with the kernel trick, processes the multi-label integrally and realizes the nonlinear dimensionality reduction with the idea similar with linear discriminant analysis （LDA）. In the classification process of multi-label data, the extreme learning machine （ELM） is an efficient algorithm in the premise of good accuracy. MLKDA, combined with ELM, shows a good performance in multi-label learning experiments with several datasets. The experiments on both static data and data stream show that MLKDA outperforms multi-label dimensionality reduction via dependence maximization （MDDM） and multi-label linear discriminant analysis （MLDA） in cases of balanced datasets and stronger correlation between tags, and ELM is also a good choice for multi-label classification.
基金Supported by the National Natural Science Foundationthe Doctoral Foundation of the State Education Commission of China
文摘This paper studies the problem applying Radial Basis Function Network(RBFN) which is trained by the Recursive Least Square Algorithm(RLSA) to the recognition of one dimensional images of radar targets. The equivalence between the RBFN and the estimate of Parzen window probabilistic density is proved. It is pointed out that the I/O functions in RBFN hidden units can be generalized to general Parzen window probabilistic kernel function or potential function, too. This paper discusses the effects of the shape parameter a in the RBFN and the forgotten factor A in RLSA on the results of the recognition of three kinds of kernel function such as Gaussian, triangle, double-exponential, at the same time, also discusses the relationship between A and the training time in the RBFN.
基金Supported by the National Natural Science Foundation of China(31271658)National Key Research and Development Program of China(2016YFD0300306)
文摘Ear morphological traits such as volume and shape are important features of maize and the quantitative associations among them can help understand kernel yield determination. 150 mature ears each of 4 maize cultivars were collected from field experiments, and ear length(L), diameter(D), area(S) and volume(V) were recorded for individual ears, kernel weight per ear also recorded for a portion of the examined ears. Following principles of dimensional analysis, 8 theoretical equations of 3 sets,which relate ear higher dimensions to its length and diameter, were developed and parameterized and validated with the field observations. The 3 optimized equations showed that the shape of ears in maize can be featured with 3 dimensionless form factors, namely diameter-to-length ratio(c=D/L), areal form factor(b=S/L/D), and volumetric form factor(a=V/L/D/D). Statistically,all of them were significantly different among cultivars, and a's values varied from 0.582 to 0.612, and b's 0.839-0.868, and c's 0.242-0.308. Volumetric form factor and areal form factor could estimate precisely ear volume and area respectively, but diameter-to-length ratio was not suitable to estimate ear diameter by its length. Ear volume explained almost all variation of ear kernel weight and product L*D*D did the same substantially. Dimensional analysis proved to be promising in understanding relationship among morphological traits of ears in maize. Its application in crop researches should improve our knowledge of the physical properties of crop plants.
文摘Approximate Bayesian Computation (ABC) is a popular sampling method in applications involving intractable likelihood functions. Instead of evaluating the likelihood function, ABC approximates the posterior distribution by a set of accepted samples which are simulated from a generating model. Simulated samples are accepted if the distances between the samples and the observation are smaller than some threshold. The distance is calculated in terms of summary statistics. This paper proposes Local Gradient Kernel Dimension Reduction (LGKDR) to construct low dimensional summary statistics for ABC. The proposed method identifies a sufficient subspace of the original summary statistics by implicitly considering all non-linear transforms therein, and a weighting kernel is used for the concentration of the projections. No strong assumptions are made on the marginal distributions, nor the regression models, permitting usage in a wide range of applications. Experiments are done with simple rejection ABC and sequential Monte Carlo ABC methods. Results are reported as competitive in the former and substantially better in the latter cases in which Monte Carlo errors are compressed as much as possible.
文摘The precision of the kernel independent component analysis( KICA) algorithm depends on the type and parameter values of kernel function. Therefore,it's of great significance to study the choice method of KICA's kernel parameters for improving its feature dimension reduction result. In this paper, a fitness function was established by use of the ideal of Fisher discrimination function firstly. Then the global optimal solution of fitness function was searched by particle swarm optimization( PSO) algorithm and a multi-state information dimension reduction algorithm based on PSO-KICA was established. Finally,the validity of this algorithm to enhance the precision of feature dimension reduction has been proven.
文摘Finding a suitable space is one of the most critical problems for dimensionality reduction. Each space corresponds to a distance metric defined on the sample attributes, and thus finding a suitable space can be converted to develop an effective distance metric. Most existing dimensionality reduction methods use a fixed pre-specified distance metric. However, this easy treatment has some limitations in practice due to the fact the pre-specified metric is not going to warranty that the closest samples are the truly similar ones. In this paper, we present an adaptive metric learning method for dimensionality reduction, called AML. The adaptive metric learning model is developed by maximizing the difference of the distances between the data pairs in cannot-links and those in must-links. Different from many existing papers that use the traditional Euclidean distance, we use the more generalized l<sub>2,p</sub>-norm distance to reduce sensitivity to noise and outliers, which incorporates additional flexibility and adaptability due to the selection of appropriate p-values for different data sets. Moreover, considering traditional metric learning methods usually project samples into a linear subspace, which is overstrict. We extend the basic linear method to a more powerful nonlinear kernel case so that well capturing complex nonlinear relationship between data. To solve our objective, we have derived an efficient iterative algorithm. Extensive experiments for dimensionality reduction are provided to demonstrate the superiority of our method over state-of-the-art approaches.
基金National Natural Science Foundation of China (60776793,60802043)National Basic Research Program of China (2010CB327900)
文摘Space object recognition plays an important role in spatial exploitation and surveillance, followed by two main problems： lacking of data and drastic changes in viewpoints. In this article, firstly, we build a three-dimensional （3D） satellites dataset named BUAA Satellite Image Dataset （BUAA-SID 1.0） to supply data for 3D space object research. Then, based on the dataset, we propose to recognize full-viewpoint 3D space objects based on kernel locality preserving projections （KLPP）. To obtain more accurate and separable description of the objects, firstly, we build feature vectors employing moment invariants, Fourier descriptors, region covariance and histogram of oriented gradients. Then, we map the features into kernel space followed by dimensionality reduction using KLPP to obtain the submanifold of the features. At last, k-nearest neighbor （kNN） is used to accomplish the classification. Experimental results show that the proposed approach is more appropriate for space object recognition mainly considering changes of viewpoints. Encouraging recognition rate could be obtained based on images in BUAA-SID 1.0, and the highest recognition result could achieve 95.87%.
基金Natural Science Foundation of Jiangsu Higher Education Institutions of China (No. 11KJB510020)National Natural Science Foundation of China (No. 61171077)College Industrialization Project of Jiangsu Province,China (No. JH09-24)
文摘To separate each pattern class more strongly and deal with nonlinear ease, a new nonlinear manifold learning algorithm named supervised kernel uneorrelated diseriminant neighborhood preserving projections （SKUDNPP） is proposed. The algorithm utilizes supervised weight and kernel technique which makes the algorithm cope with classifying and nonlinear problems competently. The within-class geometric structure is preserved, while maximizing the between-class distance. And the features extracted are statistically uneorrelated by introducing an uneorrelated constraint. Experiment results on millimeter wave （MMW） radar target recognition show that the method can give competitive results in comparison with current papular algorithms.
基金the National Natural Science Foundation of China [Grant Number 41722108],(Grant Number 91638201)%FWO project:data fusion for image analysis in remote sensing(Grant Number G037115N)
文摘Hyperspectral image provides abundant spectral information for remote discrimination of subtle differences in ground covers.However,the increasing spectral dimensions,as well as the information redundancy,make the analysis and interpretation of hyperspectral images a challenge.Feature extraction is a very important step for hyperspectral image processing.Feature extraction methods aim at reducing the dimension of data,while preserving as much information as possible.Particularly,nonlinear feature extraction methods (e.g.kernel minimum noise fraction (KMNF) transformation) have been reported to benefit many applications of hyperspectral remote sensing,due to their good preservation of high-order structures of the original data.However,conventional KMNF or its extensions have some limitations on noise fraction estimation during the feature extraction,and this leads to poor performances for post-applications.This paper proposes a novel nonlinear feature extraction method for hyperspectral images.Instead of estimating noise fraction by the nearest neighborhood information (within a sliding window),the proposed method explores the use of image segmentation.The approach benefits both noise fraction estimation and information preservation,and enables a significant improvement for classification.Experimental results on two real hyperspectral images demonstrate the efficiency of the proposed method.Compared to conventional KMNF,the improvements of the method on two hyperspectral image classification are 8 and 11％.This nonlinear feature extraction method can be also applied to other disciplines where high-dimensional data analysis is required.
