Endmember Bundle Extraction Based on Multiobjective Optimization

Rong Liu,Xiaoxiang Zhu

doi:10.1109/tgrs.2020.3037249

Abstract

A number of endmember extraction methods have been developed to identify pure pixels in hyperspectral images (HSIs). The majority of them use only one spectrum to represent one kind of material, which ignores the spectral variability problem that particularly characterizes a HSI with high spatial resolution. Only a few algorithms have been developed to identify multiple endmembers representing the spectral variability within each class, called endmember bundle extraction (EBE). This article introduces multiobjective particle swarm optimization for the identification of multiple endmember spectra with variability. Unlike existing convex geometry-based EBE methods, which operate on a single geometry of the dataspace, the proposed method divides the observed data into subsets along the spectral dimension and simultaneously operates on multiple dataspaces to obtain candidate endmembers based on multiobjective particle swarm optimization. The candidate endmembers are then refined by spatial post-processing and sequential forward floating selection to produce the final result. Experiments are conducted on both synthetic and real hyperspectral data to demonstrate the effectiveness of the proposed method in comparison with several state-of-the-art methods.

Highlights

W ITH the ability to record abundant spectral information about materials, hyperspectral imagery has been widely used for various applications, including vegetation mapping [1], mineral exploration [2], agricultural assessment [3], and many others [4]–[6]
1) Results of Experiment 1 (Using the Synthetic Data Set): To validate the effectiveness of the velocity and position updating strategies used in MOEBE, the velocity and position updating strategies of the proposed method are replaced with the original ones in competition mechanism-based multiobjective particle swarm optimization (CMOPSO), and the method has been tested on the synthetic image
In experiment 1, the endmembers extracted by endmember bundle extraction (EBE), SSEBE, spectral curve-based endmember extraction (SCEE), and MOEBE were evaluated quantitatively with SAD, rootmean-square error (RMSE), average deviation of the standard deviations (ADS), and average deviation of the mean (ADM) under the condition that the true endmembers with variability and abundances were known

Summary

Introduction

W ITH the ability to record abundant spectral information about materials, hyperspectral imagery has been widely used for various applications, including vegetation mapping [1], mineral exploration [2], agricultural assessment [3], and many others [4]–[6]. To reduce the amount of time and expense involved in field measurement and keep similar atmospheric effects between endmembers and the data to be unmixed, a number of methods have been proposed to automatically extract endmembers directly from the image data. Many of these approaches use a single spectrum to represent one kind of material and only extract a single endmember spectrum for each endmember class. For a scene with large spectral variations, ignoring the endmember variability problem can lead to poor unmixing results

Methods

Results

Conclusion