Hyperspectral Unmixing with Endmember Variability using semi-supervised Partial Membership Latent Dirichlet Allocation

Authors: Sheng Zou*, , Hao Sun, University of Missouri-Columbia, Alina Zare, University of Florida
Topics: Remote Sensing, Spatial Analysis & Modeling, Land Use and Land Cover Change
Keywords: hyperspectral, unmiximg, spectral variability, LDA
Session Type: Paper
Day: 4/14/2018
Start / End Time: 10:00 AM / 11:40 AM
Room: Galvez, , Marriott, 5th Floor
Presentation File: No File Uploaded

Hyperspectral unmixing approaches that account for spectral variability model endmembers (i.e., the pure spectral signatures in hyperspectral image) as a set or distribution. In this work, the semi-supervised Partial Membership Latent Dirichlet Allocation (PM-LDA) model is presented and applied to hyperspectral unmixing while accounting for spectral variability, using imprecise labels from map data, and leveraging spatial neighboring similarity. Experiments on both synthetic and real hyperspectral images demonstrate that the proposed model can better model the underlying endmember distributions in the datasets when compared to other hyperspectral unmixing models. Moreover, the proposed semi-supervised PMLDA approach performs well on datasets with high between-endmember similarity, a very challenging problem in this area.

Abstract Information

This abstract is already part of a session. View the session here.

To access contact information login