Year-end Sale % OFF 
Abstract
Due to the different spatial properties presented by various ground objects in hyperspectral image (HSI), multiscale-based feature extraction approaches have been developed for HSI classification in recent years. However, the spatial features of different scales are usually acquired at the cost of obscuring the structural information of input image, which severely limits the effectiveness of multiscale strategy. In this article, a convolutional sparse decomposition (CSD) model is introduced to characterize the significant spatial structures of hyperspectral data while removing the irrelevant noise and local textures at the specific scale. Based on the CSD model, a multiscale spectral-spatial feature extraction framework is generated, which consists of the following steps. First, the spectral dimensionality of the original HSI is reduced through a segmented averaging approach. Second, spatial features at different scales are separated from the dimension-reduced data by solving the CSD model with different regularization parameters. Finally, principal component analysis is performed and the obtained multiscale spectral-spatial features are stacked together for classification. Experiments conducted on three widely used hyperspectral datasets demonstrate that the proposed method is robust in capturing effective features of ground objects at different scales and leads to better classification results than several state-of-the-art methods.
Highlights
H YPERSPECTRAL imagery can provide a wealth of information about different land cover objects due to the high resolution of spectral dimension [1]
1) Comparative Methods and Evaluation Indexes: Several representative methods are chosen for comparison to validate the performance of the proposed method during the classification experiments: the traditional pixel-wise support vector machines (SVMs) [6], the sparsity-based classifier JSRC [9], the MASR model [52] which exploits spatial information at multiple scales via an adaptive sparse strategy, the IID feature extraction algorithm [25], the multiscale edge-preserving filtering approach principal component analysis (PCA)–EPFs [35], the segmented PCA and Gaussian pyramid decomposition based multiscale feature fusion method (SPCA-GPs) [36], and a handcrafted feature extraction approach based on multiscale covariance maps (MCMs) for hyperspectral image (HSI) classification using CNNs [53]
For the synthesis-based sparse dictionaries {dA,1, . . . , dA,M } that adaptively learned from the input image, we found in the experiment that the local textures can be better reconstructed with the size of dictionaries set at the range of 5–9, and the proper choice of M is mainly dependent on the texture complexity of the considered image
Summary
H YPERSPECTRAL imagery can provide a wealth of information about different land cover objects due to the high resolution of spectral dimension [1]. As a commonly used technique in image preprocessing, spatial smoothing can exploit the local relationship among neighboring pixels and enhance the main features within a homogenous region [33] In this context, several approaches based on smoothing strategy have been generated for HSI to extract spatial features in different scales. The obtained spatial maps of these methods usually tend to be blurry with foggy artifacts that obscure the important structural information of the input image (see Fig. 1) This severely limits the effectiveness of multiscale strategy on HSI classification, especially for urban scenarios that contain various artificial objects. Based on the fact that the spatial information of an image generally contains a structure component in a specific scale and the corresponding high-frequency textures, we present a robust multiscale feature extraction approach with convolutional sparse decomposition (CSD) for HSI classification. By combining the proposed MCSD with PCA, discriminative multiscale spectralspatial features can be obtained for HSI classification
Sign-in/Register to view highlights & summary Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callMore From: IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
Disclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
