Abstract
Most traditional endmember extraction algorithms focus on spectral information, which limits the effectiveness of endmembers. This paper develops a spatial potential energy weighted maximum simplex algorithm (SPEW) for hyperspectral endmember extraction, combining the relevance of hyperspectral spatial context with spectral information to effectively extract endmembers. Specifically, for pixels in a uniform spatial area, SPEW assigns a high weight to pixels with higher spatial potential energy. For pixels scattered in a spatial area, the high weights are assigned to the representative pixels with a smaller spectral angle distance. Then, the optimal endmember collection is determined by the simplex with maximum volume in the space of representative pixels. SPEW not only reduces the complexity of searching for the maximum simplex volume but also improves the performance of endmember extraction. In particular, compared with other newly proposed spatial-spectral hyperspectral endmember extraction methods, SPEW can effectively extract the hidden endmembers in a spatial area without adjusting any parameters. Experiments on synthetic and real data show that the SPEW algorithm has also provides better results than the traditional algorithms.
Highlights
Hyperspectral remote sensing image data has the characteristics of a wide range of spectral bands and high spectral resolution
In order to overcome the problem of ignoring endmembers in the scattered region of pixel categories, we introduce spectral angular distance to search for spectral representative pixels with low spatial potential energy and treat them as candidate endmembers to construct the maximum simplex
Automatic target generation process (ATGP) and simplex growth algorithm (SGA) extracted all types of pure pixels in the same position because these two algorithms are inherently equivalent under the same initial conditions
Summary
Hyperspectral remote sensing image data has the characteristics of a wide range of spectral bands and high spectral resolution. The existence of mixed pixels reduces the accuracy of traditional pixel-level data operation, so the decomposition of mixed pixels is an important task in processing of hyperspectral images. In hyperspectral images (HSIs), the pure pixels that contain only one kind of terrain information are called endmembers, and the fractions of each endmember in the pixels are called abundances [1–3]. Typical endmember extraction algorithms based on convex geometry are widely used in linear spectral unmixing. Pure pixel index (PPI) [4] involves projecting all pixels onto a set of randomly generated unit vector. The mixed pixels will be projected to the middle, and the endmembers will be projected to the endpoints of vector
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 callDisclaimer: 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.
