Bayesian Spectral Mixture Analysis for Nighttime Light Image

Authors: Guoquan Feng*, University At Buffalo, Le Wang, advisor
Topics: Remote Sensing, Drones
Keywords: unmanned aerial vehicle, nighttime light,streetlight
Session Type: Guided Poster
Day: 4/6/2019
Start / End Time: 8:00 AM / 9:40 AM
Room: Roosevelt 3.5, Marriott, Exhibition Level
Presentation File: No File Uploaded

Nighttime light image is a unique data source in urban remote sensing. Unlike daytime image which is normally obtained from reflected sunlight, nighttime light image records radiation which is produced by human activities. However, most of nighttime light images suffer from low spatial resolution so majority of pixels in urban area are mixed pixels. Given such coarse spatial resolution, subpixel information is essential in mapping finer-scale human activities, such as urbanization, light pollution and so on. However, no spectral unmixing method has been develop for Nighttime light image.The objective of our study is to develop a spectral unmixing method for nighttime light image. Specifically, how to represent endmember variability and how to incorporate endmember variability in spectral unmixing method for nighttime light image.We also use Luojia-1 as reference data to validate our result. Our method has better R-Square and relative good RMSE compared to the traditional linear unmixing. From our study, Endmember variability is a significant problem in NTL image spectral unmixing. When considering endmember variability, Bayesian spectral mixture analysis has comparative performance with the linear spectral unmixing method. Uncertainty of fraction can be derived from Bayesian spectral mixture analysis as a distribution.

Abstract Information

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

To access contact information login