Time filter

Source Type

Sun W.,Ningbo University | Halevy A.,Randolph-Macon College | Benedetto J.J.,University of Maryland University College | Czaja W.,University of Maryland University College | And 5 more authors.
ISPRS Journal of Photogrammetry and Remote Sensing | Year: 2014

The paper proposes an upgraded landmark-Isometric mapping (UL-Isomap) method to solve the two problems of landmark selection and computational complexity in dimensionality reduction using landmark Isometric mapping (LIsomap) for hyperspectral imagery (HSI) classification. First, the vector quantization method is introduced to select proper landmarks for HSI data. The approach considers the variations in local density of pixels in the spectral space. It locates the unique landmarks representing the geometric structures of HSI data. Then, random projections are used to reduce the bands of HSI data. After that, the new method incorporates the Recursive Lanczos Bisection (RLB) algorithm to construct the fast approximate k-nearest neighbor graph. The RLB algorithm accompanied with random projections improves the speed of neighbor searching in UL-Isomap. After constructing the geodesic distance graph between landmarks and all pixels, the method uses a fast randomized low-rank approximate method to speed up the eigenvalue decomposition of the inner-product matrix in multidimensional scaling. Manifold coordinates of landmarks are then computed. Manifold coordinates of non-landmarks are computed through the pseudo inverse transformation of landmark coordinates. Five experiments on two different HSI datasets are run to test the new UL-Isomap method. Experimental results show that UL-Isomap surpasses LIsomap, both in the overall classification accuracy (OCA) and in computational speed, with a speed over 5 times faster. Moreover, the UL-Isomap method, when compared against the Isometric mapping (Isomap) method, obtains only slightly lower OCAs. © 2013 International Society for Photogrammetry and Remote Sensing, Inc. (ISPRS).

Sun W.,Ningbo University | Sun W.,Wuhan University | Liu C.,Tongji University | Liu C.,Key Laboratory of Advanced Engineering Survey of NASMG | And 2 more authors.
Wuhan Daxue Xuebao (Xinxi Kexue Ban)/Geomatics and Information Science of Wuhan University | Year: 2015

Hyperspectral imagery (HSI) classification can be achieved through the combination scheme of nonlinear dimensionality reduction using Laplacian eigenmaps (LE) and classification using the k-nearest neighbor (KNN) classifier. However, both the LE and the KNN classifier omit spatial features of HSI data as the imagery. That seriously restricts the classification result of HSI data. This paper presents the adaptive weighted summation kernel distance (AWSKD) to improve both the LE and the KNN classifier, aiming to promote the classification accuracies of HSI data. The AWSKD considers the spectral and spatial features of HSI data, and adaptively estimate the proper spatial neighborhood size for describing the spatial feature of each pixel. The Indian and PaviaU datasets are utilized to analyze and testify the proposed combination scheme of improved LE (ILE) and improved KNN (IKNN) classifier. Experimental results show that the proposed combination scheme achieves sharply higher classification accuracies than the regular scheme of LE and KNN.

Sun W.,Tongji University | Liu C.,Tongji University | Liu C.,Key Laboratory of Advanced Engineering Survey of NASMG | Shi B.,Shanghai Normal University | Li W.,Tongji University
Wuhan Daxue Xuebao (Xinxi Kexue Ban)/Geomatics and Information Science of Wuhan University | Year: 2012

As a manifold learning method, Isomap has been widely used for making nonlinearly reduction for hyperspectral image. However, during the construction process of the shortest path graph, the boundary points, which are not noise points, have always been omitted for the consideration of the stability of the graph. Therefore, the PLS method is introduced to repair and simulate the manifold coordinates of the lost points in the shortest path graph. And the simulated manifold coordinates have been evaluated from two different aspects to verify our method. The results show that the simulated manifold coordinates agree well with the real one. It will be quite useful for further classification or visualization with low dimensional manifold image.

Sun W.,Tongji University | Liu C.,Tongji University | Liu C.,Key Laboratory of Advanced Engineering Survey of NASMG | Shi B.,Tongji University | And 2 more authors.
Journal of Remote Sensing | Year: 2013

Manifold coordinates from Isometric mapping (Isomap) and Local Tangent Space Alignment (LTSA) preserve the spectral features of ground objects from Hyperspectral Imagery (HSI) through nonlinear dimensionality reduction. However, the theoretical differences result in differing capabilities in preserving spectral features. Thus, a comparison of two coordinates can make the underlying features prominent. Therefore, this paper proposes an innovative method called Difference Maps from Manifold Coordinates (DMMC), which is based on Isomap and LTSA, to extract underlying features. First, spectral interpretations are matched with both coordinates and ensured to preserve the same spectral features. Second, the Isomap and LTSA coordinates are transformed into a uniform system using coordinate normalization and axis-direction adjustment. Finally, the difference maps are obtained through subtraction operations between the weighted manifold maps, and underlying features are extracted using classical image processing approaches. Two case studies are performed to evaluate the proposed method, and the results are compared with those obtained using Isomap and LTSA. The results show that DMMC outperforms Isomap and LTSA in extracting underlying features, such as the underlying shallow water near the river bank and the low spatial-resolution road in the large image scene of a swamp. This method provides a novel scheme for extracting underlying features from HSI data.

Sun W.,Tongji University | Halevy A.,University of Maryland College Park | Benedetto J.J.,University of Maryland College Park | Czaja W.,University of Maryland College Park | And 6 more authors.
IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing | Year: 2014

The problems of neglecting spatial features in hyperspectral imagery (HSI) and the high complexity of Local Tangent Space Alignment (LTSA) still exist in the nonlinear dimensionality reduction with LTSA for classification. Therefore, this paper proposes an innovative ENH-LTSA (Enhanced-Local Tangent Space Alignment) method to solve the two problems. First, random projection is introduced to preliminarily reduce the dimension of HSI data. It aims to improve the speed of neighbor searching and the local tangent space construction. Then, the new method presents the similarity measure via the adaptive weighted summation kernel (AWSK) distance. The AWSK distance considers both spectral and spatial features in HSI data, and attempts to ameliorate the k-nearest neighbors (KNNs) of each pixel. Furthermore, the adaptive spatial window is proposed to automatically estimate the proper window size for the description of spatial features. After that, fast approximate KNNs graph construction via Recursive Lanczos Bisection is incorporated into the new method to reduce the complexity of KNNs searching. When finishing constructing each local tangent space, the new method uses a fast low-rank approximate singular value decomposition to speed up eigenvalue decomposition of the global alignment matrix that is constituted with local manifold coordinates. Five groups of experiments with two different HSI datasets are designed to completely analyze and testify the ENH-LTSA method. Experimental results show that ENH-LTSA outperforms LTSA, both in classification results and in computational speed. © 2013 IEEE.

Discover hidden collaborations