Physics-aware gaussian processes for earth observation

Camps-Valls, Gustau and Svendsen, Daniel H. and Martino, Luca and Muñoz Marí, Jordi and Laparra Pérez-Muelas, Valero and Campos-Taberner, Manuel and Luengo García, David (2017). Physics-aware gaussian processes for earth observation. In: "20th Scandinavian Conference on Image Analysis, SCIA 2017", 12/06/2017 - 14/06/2017, Tromso (Noruega). ISBN 978-3-319-59125-4. pp. 205-217. https://doi.org/10.1007/978-3-319-59129-2 18.

Description

Title: Physics-aware gaussian processes for earth observation
Author/s:
  • Camps-Valls, Gustau
  • Svendsen, Daniel H.
  • Martino, Luca
  • Muñoz Marí, Jordi
  • Laparra Pérez-Muelas, Valero
  • Campos-Taberner, Manuel
  • Luengo García, David
Item Type: Presentation at Congress or Conference (Article)
Event Title: 20th Scandinavian Conference on Image Analysis, SCIA 2017
Event Dates: 12/06/2017 - 14/06/2017
Event Location: Tromso (Noruega)
Title of Book: Image Analysis : 20th Scandinavian Conference, SCIA 2017, Tromsø, Norway, June 12–14, 2017, Proceedings, Part II
Date: 2017
ISBN: 978-3-319-59125-4
Volume: 2
Subjects:
Freetext Keywords: Earth observation; Remote sensing; Vegetation; Kernel methods; Gaussian processes; Inverse modeling; Geosciences; Transfer models
Faculty: E.T.S.I. y Sistemas de Telecomunicación (UPM)
Department: Teoría de la Señal y Comunicaciones
Creative Commons Licenses: Recognition - No derivative works - Non commercial

Full text

[img]
Preview
PDF - Requires a PDF viewer, such as GSview, Xpdf or Adobe Acrobat Reader
Download (3MB) | Preview

Abstract

Earth observation from satellite sensory data pose challenging problems, where machine learning is currently a key player. In recent years, Gaussian Process (GP) regression and other kernel methods have excelled in biophysical parameter estimation tasks from space. GP regression is based on solid Bayesian statistics, and generally yield efficient and accurate parameter estimates. However, GPs are typically used for inverse modeling based on concurrent observations and in situ measurements only. Very often a forward model encoding the well-understood physical relations is available though. In this work, we review three GP models that respect and learn the physics of the underlying processes in the context of inverse modeling. First, we will introduce a Joint GP (JGP) model that combines in situ measurements and simulated data in a single GP model. Second, we present a latent force model (LFM) for GP modeling that encodes ordinary differential equations to blend data-driven modeling and physical models of the system. The LFM performs multi-output regression, adapts to the signal characteristics, is able to cope with missing data in the time series, and provides explicit latent functions that allow system analysis and evaluation. Finally, we present an Automatic Gaussian Process Emulator (AGAPE) that approximates the forward physical model via interpolation, reducing the number of necessary nodes. Empirical evidence of the performance of these models will be presented through illustrative examples of vegetation monitoring and atmospheric modeling.

Funding Projects

TypeCodeAcronymLeaderTitle
Horizon 2020ERC-CoG-2014SEDAL projectUnspecifiedUnspecified
Government of SpainTIN2015-64210-RUnspecifiedUnspecifiedUnspecified

More information

Item ID: 51089
DC Identifier: http://oa.upm.es/51089/
OAI Identifier: oai:oa.upm.es:51089
DOI: 10.1007/978-3-319-59129-2 18
Official URL: http://scia2017.org/
Deposited by: Memoria Investigacion
Deposited on: 31 May 2018 14:41
Last Modified: 01 Jun 2019 22:30
  • Logo InvestigaM (UPM)
  • Logo GEOUP4
  • Logo Open Access
  • Open Access
  • Logo Sherpa/Romeo
    Check whether the anglo-saxon journal in which you have published an article allows you to also publish it under open access.
  • Logo Dulcinea
    Check whether the spanish journal in which you have published an article allows you to also publish it under open access.
  • Logo de Recolecta
  • Logo del Observatorio I+D+i UPM
  • Logo de OpenCourseWare UPM