Header

UZH-Logo

Maintenance Infos

Weakly supervised alignment of multisensor images


Marcos-Gonzalez, Diego; Camps-Valls, Gustau; Tuia, Devis (2015). Weakly supervised alignment of multisensor images. In: IGARSS 2015, Milan (Italy), 26 July 2015 - 31 July 2015, 2588-2591.

Abstract

Manifold alignment has become very popular in recent liter- ature. Aligning data distributions prior to product generation is an appealing strategy, since it allows to provide data spaces that are more similar to each other, regardless of the subse- quent use of the transformed data. We propose a methodology that finds a common representation among data spaces from different sensors using geographic image correspondences, or semantic ties. To cope with the strong deformations between the data spaces considered, we propose to add nonlineari- ties by expanding the input space with Gaussian Radial Basis Function (RBF) features with respect to the centroids of a par- titioning of the data. Such features allow us to cope with non- linear transformations, while keeping a simple and efficient linear formulation. The proposed method is multi-domain and does not require co-registration, rather only a partial de- gree of spatial overlap. We test it on a challenging problem of multisensor classification transferring a model trained on a WorldView 2 image to predict land cover of a 3-bands or- thophoto and show that we can transfer the model with an ac- curacy comparable to the one that would have been obtained by a model trained on the target image with an image-specific ground truth.

Abstract

Manifold alignment has become very popular in recent liter- ature. Aligning data distributions prior to product generation is an appealing strategy, since it allows to provide data spaces that are more similar to each other, regardless of the subse- quent use of the transformed data. We propose a methodology that finds a common representation among data spaces from different sensors using geographic image correspondences, or semantic ties. To cope with the strong deformations between the data spaces considered, we propose to add nonlineari- ties by expanding the input space with Gaussian Radial Basis Function (RBF) features with respect to the centroids of a par- titioning of the data. Such features allow us to cope with non- linear transformations, while keeping a simple and efficient linear formulation. The proposed method is multi-domain and does not require co-registration, rather only a partial de- gree of spatial overlap. We test it on a challenging problem of multisensor classification transferring a model trained on a WorldView 2 image to predict land cover of a 3-bands or- thophoto and show that we can transfer the model with an ac- curacy comparable to the one that would have been obtained by a model trained on the target image with an image-specific ground truth.

Statistics

Citations

1 citation in Web of Science®
2 citations in Scopus®
Google Scholar™

Altmetrics

Downloads

1 download since deposited on 14 Jan 2016
0 downloads since 12 months
Detailed statistics

Additional indexing

Item Type:Conference or Workshop Item (Paper), not refereed, original work
Communities & Collections:07 Faculty of Science > Institute of Geography
Dewey Decimal Classification:910 Geography & travel
Language:English
Event End Date:31 July 2015
Deposited On:14 Jan 2016 08:58
Last Modified:17 Aug 2017 18:08
Publisher:IEEE
ISBN:978-1-4799-7929-5
Free access at:Official URL. An embargo period may apply.
Publisher DOI:https://doi.org/10.1109/IGARSS.2015.7326341
Official URL:http://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=7326341

Download

Preview Icon on Download
Content: Published Version
Language: English
Filetype: PDF - Registered users only
Size: 992kB
View at publisher