Spatially Adaptive Hyperspectral Unmixing

Abstract

Spectral unmixing is a common task in hyperspectral data analysis. In order to sufficiently spectrally unmix the data, three key steps must be accomplished: Estimate the number of endmembers (EMs), identify the EMs, and then unmix the data. Several different statistical and geometrical approaches have been developed for all steps of the unmixing process. However, many of these methods rely on using the full image to estimate the number and extract the EMs from the background data. In this paper, spectral unmixing is accomplished using a spatially adaptive approach. Linear unmixing is performed per pixel with EMs identified at the local level, but global abundance maps are created by clustering the locally determined EMs into common groups. Results show that the unmixing residual error of each pixel's spectrum from real data, estimated from the spatially adaptive methodology, is reduced when compared to a global scale EM estimation and linear unmixing methodology. The component algorithms of the new spatially adaptive approach, which complete the three key unmixing steps, can be interchanged while maintaining spatial information, making this new methodology modular. A final advantage of the spatially adaptive spectral unmixing methodology is the user-defined spatial scale size.

Spectral Unmixing via Data-Guided Sparsity

“…the Spectral Angle Distance [12], [16], [11] and the Root Mean Square Error (RMSE) [12], [9], [56]. SAD is used to evaluate the estimated endmembers.…”

A survey of methods incorporating spatial information in image classification and spectral unmixing

“…A variety of image partitioning approaches have been adopted, including Markov random fields (MRFs) (Rand and Keenan 2001), iterative self-organizing data analysis techniques A (ISODATA) (Shoshany and Svoray 2002), unmixing results (García-Haro, Sommer, and Kemper 2005), local tiles (Canham et al 2011;Goenaga et al 2013), iterative fuzzy clustering (Zare et al 2013), image classification (Zhang et al 2014c), and multi-resolution image segmentation Li et al 2015). The image partitions so produced are commonly grouped to form homogenous fields that represent different land-cover types according to prior knowledge of the scene (Rand and Keenan 2001;Shoshany and Svoray 2002;García-Haro, Sommer, and Kemper 2005;Zhang et al 2014c).…”

“…This image stratification process is also suggested by a comparative study for estimating fractional green vegetation cover (Xiao and Moody 2005). Endmembers in each field are then selected by a manual process (Rand and Keenan 2001;Shoshany and Svoray 2002;García-Haro, Sommer, and Kemper 2005;Zhang et al 2014c;Li et al 2015) or automatic endmember extraction algorithms (Canham et al 2011;Goenaga et al 2013;Zhang et al 2014a). Zare et al (2013) proposed a piecewise convex multiple-model endmember detection (PCOMMEND) algorithm to partition an image into multiple fields constrained by a family of simplices in spectral space.…”

Comparison of hyperion spectral unmixing endmembers to material spectral profiles from Oaxaca, Mexico

“…An automated spatially adaptive spectral unmixing algorithm [6] was applied to the chipped, reflectance Hyperion data to produce EM abundance maps. The partially constrained unmixing approach applies the nonnegative constraint, but not the sum-to-one constraint due to the known low SNR with many Hyperion bands [3].…”