Getting pixels and regions to agree with conditional random fields

Devis Tuia; Michele Volpi; Gabriele Moser

doi:10.1109/IGARSS.2016.7729851

Getting pixels and regions to agree with conditional random fields

Devis Tuia, Michele Volpi, Gabriele Moser

Research output: Chapter in Book/Report/Conference proceeding › Conference paper › Academic › peer-review

6 Citations (Scopus)

Abstract

Land cover / land use classification of remotely sensed images is inherently geographical. The use of spatial information, accounting for neighborhood relationship and spatial smoothness of geographical objects, made its proofs in countless occasions and, especially when considering very high resolution images, methods ignoring spatial context do not perform well. In this paper, we propose a hybrid dual-layer conditional random field model that enforces spatial smoothness and consistency between the pixel and region-based maps. We formulate these intuitions as a standard energy minimization problem, and we show that finding a joint solution over both output spaces leads to strong improvements in the numerical and visual senses.

Original language	English
Title of host publication	2016 IEEE International Geoscience and Remote Sensing Symposium, IGARSS 2016 - Proceedings
Place of Publication	Beijing
Publisher	IEEE
Pages	3290-3293
Number of pages	4
ISBN (Electronic)	9781509033324
ISBN (Print)	9781509033317
DOIs	https://doi.org/10.1109/IGARSS.2016.7729851
Publication status	Published - Nov 2016
Externally published	Yes
Event	36th IEEE International Geoscience and Remote Sensing Symposium, IGARSS 2016 - Beijing, China Duration: 10 Jul 2016 → 15 Jul 2016

Publication series

Name	International Geoscience and Remote Sensing Symposium (IGARSS)
Volume	2016-November

Conference

Conference	36th IEEE International Geoscience and Remote Sensing Symposium, IGARSS 2016
Country/Territory	China
City	Beijing
Period	10/07/16 → 15/07/16

Keywords

Conditional random fields
Markov random fields
random forests
structured prediction
urban remote sensing
very high resolution

UN SDGs

This output contributes to the following UN Sustainable Development Goals (SDGs)

Access to Document

10.1109/IGARSS.2016.7729851

Cite this

Tuia, D., Volpi, M., & Moser, G. (2016). Getting pixels and regions to agree with conditional random fields. In 2016 IEEE International Geoscience and Remote Sensing Symposium, IGARSS 2016 - Proceedings (pp. 3290-3293). Article 7729851 (International Geoscience and Remote Sensing Symposium (IGARSS); Vol. 2016-November). IEEE. https://doi.org/10.1109/IGARSS.2016.7729851

@inproceedings{0039e96b74d7443a84aa1193ca0b498a,

title = "Getting pixels and regions to agree with conditional random fields",

abstract = "Land cover / land use classification of remotely sensed images is inherently geographical. The use of spatial information, accounting for neighborhood relationship and spatial smoothness of geographical objects, made its proofs in countless occasions and, especially when considering very high resolution images, methods ignoring spatial context do not perform well. In this paper, we propose a hybrid dual-layer conditional random field model that enforces spatial smoothness and consistency between the pixel and region-based maps. We formulate these intuitions as a standard energy minimization problem, and we show that finding a joint solution over both output spaces leads to strong improvements in the numerical and visual senses.",

keywords = "Conditional random fields, Markov random fields, random forests, structured prediction, urban remote sensing, very high resolution",

author = "Devis Tuia and Michele Volpi and Gabriele Moser",

year = "2016",

month = nov,

doi = "10.1109/IGARSS.2016.7729851",

language = "English",

isbn = "9781509033317",

series = "International Geoscience and Remote Sensing Symposium (IGARSS)",

publisher = "IEEE",

pages = "3290--3293",

booktitle = "2016 IEEE International Geoscience and Remote Sensing Symposium, IGARSS 2016 - Proceedings",

address = "United States",

note = "36th IEEE International Geoscience and Remote Sensing Symposium, IGARSS 2016 ; Conference date: 10-07-2016 Through 15-07-2016",

}

Tuia, D, Volpi, M & Moser, G 2016, Getting pixels and regions to agree with conditional random fields. in 2016 IEEE International Geoscience and Remote Sensing Symposium, IGARSS 2016 - Proceedings., 7729851, International Geoscience and Remote Sensing Symposium (IGARSS), vol. 2016-November, IEEE, Beijing, pp. 3290-3293, 36th IEEE International Geoscience and Remote Sensing Symposium, IGARSS 2016, Beijing, China, 10/07/16. https://doi.org/10.1109/IGARSS.2016.7729851

Getting pixels and regions to agree with conditional random fields. / Tuia, Devis; Volpi, Michele; Moser, Gabriele.
2016 IEEE International Geoscience and Remote Sensing Symposium, IGARSS 2016 - Proceedings. Beijing: IEEE, 2016. p. 3290-3293 7729851 (International Geoscience and Remote Sensing Symposium (IGARSS); Vol. 2016-November).

Research output: Chapter in Book/Report/Conference proceeding › Conference paper › Academic › peer-review

TY - GEN

T1 - Getting pixels and regions to agree with conditional random fields

AU - Tuia, Devis

AU - Volpi, Michele

AU - Moser, Gabriele

PY - 2016/11

Y1 - 2016/11

N2 - Land cover / land use classification of remotely sensed images is inherently geographical. The use of spatial information, accounting for neighborhood relationship and spatial smoothness of geographical objects, made its proofs in countless occasions and, especially when considering very high resolution images, methods ignoring spatial context do not perform well. In this paper, we propose a hybrid dual-layer conditional random field model that enforces spatial smoothness and consistency between the pixel and region-based maps. We formulate these intuitions as a standard energy minimization problem, and we show that finding a joint solution over both output spaces leads to strong improvements in the numerical and visual senses.

AB - Land cover / land use classification of remotely sensed images is inherently geographical. The use of spatial information, accounting for neighborhood relationship and spatial smoothness of geographical objects, made its proofs in countless occasions and, especially when considering very high resolution images, methods ignoring spatial context do not perform well. In this paper, we propose a hybrid dual-layer conditional random field model that enforces spatial smoothness and consistency between the pixel and region-based maps. We formulate these intuitions as a standard energy minimization problem, and we show that finding a joint solution over both output spaces leads to strong improvements in the numerical and visual senses.

KW - Conditional random fields

KW - Markov random fields

KW - random forests

KW - structured prediction

KW - urban remote sensing

KW - very high resolution

U2 - 10.1109/IGARSS.2016.7729851

DO - 10.1109/IGARSS.2016.7729851

M3 - Conference paper

AN - SCOPUS:85007476993

SN - 9781509033317

T3 - International Geoscience and Remote Sensing Symposium (IGARSS)

SP - 3290

EP - 3293

BT - 2016 IEEE International Geoscience and Remote Sensing Symposium, IGARSS 2016 - Proceedings

PB - IEEE

CY - Beijing

T2 - 36th IEEE International Geoscience and Remote Sensing Symposium, IGARSS 2016

Y2 - 10 July 2016 through 15 July 2016

ER -

Getting pixels and regions to agree with conditional random fields

Abstract

Publication series

Conference

Keywords

UN SDGs

Access to Document

Fingerprint

Cite this