Advances in Spectral-Spatial Classification of Hyperspectral Images

Abstract

International audienceRecent advances in spectral-spatial classification of hyperspectral images are presented in this paper. Several techniques are investigated for combining both spatial and spectral information. Spatial information is extracted at the object (set of pixels) level rather than at the conventional pixel level. Mathematical morphology is first used to derive the morphological profile of the image, which includes character- istics about the size, orientation, and contrast of the spatial structures present in the image. Then, the morphological neighborhood is defined and used to derive additional features for classification. Classification is performed with support vector machines (SVMs) using the available spectral information and the extracted spatial information. Spatial postprocessing is next investigated to build more homogeneous and spatially consistent thematic maps. To that end, three presegmentation techniques are applied to define regions that are used to regularize the preliminary pixel-wise thematic map. Finally, a multiple-classifier (MC) system is defined to produce relevant markers that are exploited to segment the hyperspectral image with the minimum spanning forest algorithm. Experimental results conducted on three real hyperspectral images with different spatial and spectral resolutions and corresponding to various contexts are presented. They highlight the importance of spectral-spatial strategies for the accurate classification of hyperspectral images and validate the proposed methods

Model-based co-clustering for hyperspectral images

“…More generally, co-clustering approaches provides results on the line of recent advances in spectral-spatial analysis of HSI images. 17 Comparison of the results obtained for co-clustering with K = 4 pixel clusters (see Figure 3) can be made with the ones applying k-means. For this purpose, Figure 5 provides the results obtained for k-means pixel clustering considering four clusters.…”

Constrained Manifold Learning for Hyperspectral Imagery Visualization

“…To take advantage of the spatial information, joint spectral-spatial methods such as Markov random Fields, vectors with stacked spectral-spatial features and morphological profiles were proposed (See [26] for a comprehensive review). In this paper, a composite kernel [27] is applied in constructing the manifold structure of an HSI to incorporate its spectral and spatial information.…”

Integration of Segmentation Techniques for Classification of Hyperspectral Images

“…ith recent advances in remote sensing technology, the spatial resolution of satellite images has become less than one meter [15]. The accurate classification of remote sensing images play a key role in many applications, including crop monitory, forest applications, urban development, mapping and tracking or risk management.…”

“…The accurate classification of remote sensing images play a key role in many applications, including crop monitory, forest applications, urban development, mapping and tracking or risk management. One way for achieving this goal would be to use the spectral and the spatial information sequentially [15]. The goal of considering spatial context in the classification step can be partially achieved by using some specific methods such as morphological filters [15] and Markov random fields [4].…”

“…One way for achieving this goal would be to use the spectral and the spatial information sequentially [15]. The goal of considering spatial context in the classification step can be partially achieved by using some specific methods such as morphological filters [15] and Markov random fields [4]. The abovementioned methods significantly increase the accuracy of the classification by incorporating spatial and spectral information.…”