CoChat

Exploiting Deep Matching and SAR Data for the Geo-Localization Accuracy Improvement of Optical Satellite Images

Open access

The DOI matches its Crossref record; 55 of the 55 references it cites were found in a registry or an index.

Abstract

Improving the geo-localization of optical satellite images is an important pre-processing step for many remote sensing tasks like monitoring by image time series or scene analysis after sudden events. These tasks require geo-referenced and precisely co-registered multi-sensor data. Images captured by the high resolution synthetic aperture radar (SAR) satellite TerraSAR-X exhibit an absolute geo-location accuracy within a few decimeters. These images represent therefore a reliable source to improve the geo-location accuracy of optical images, which is in the order of tens of meters. In this paper, a deep learning-based approach for the geo-localization accuracy improvement of optical satellite images through SAR reference data is investigated. Image registration between SAR and optical images requires few, but accurate and reliable matching points. These are derived from a Siamese neural network. The network is trained using TerraSAR-X and PRISM image pairs covering greater urban areas spread over Europe, in order to learn the two-dimensional spatial shifts between optical and SAR image patches. Results confirm that accurate and reliable matching points can be generated with higher matching accuracy and precision with respect to state-of-the-art approaches.

Overview

The claims and statistics of this paper haven’t been checked yet.

Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.

A few free checks a day · papers already verified open instantly · results are public

Integrity signals

Registry

Crossref · 10.3390/rs9060586

Cites

55 of 55 references verified0 citation errors · 0 not checkable

Statistics

Figures

Figure 1 Visual comparison of an optical ( top ) and SAR image ( bottom ) acquired over the same area. Both images have a ground sampling distance of 1.25 m.
Figure 1 Visual comparison of an optical ( top ) and SAR image ( bottom ) acquired over the same area. Both images have a ground sampling distance of 1.25 m.CC-BY · From the PDF
Figure 2 Network architecture ( left ) and a detailed overview of the convolutional layers ( right ). Abbreviations: convolutional neural network (CNN), convolution (Conv), batch normalization (BN) and rectified linear unit (ReLU).
Figure 2 Network architecture ( left ) and a detailed overview of the convolutional layers ( right ). Abbreviations: convolutional neural network (CNN), convolution (Conv), batch normalization (BN) and rectified linear unit (ReLU).CC-BY · From the PDF
Figure 3 Visual comparison between optical ( a ), SAR ( b ) and despeckled SAR patches ( c ).
Figure 3 Visual comparison between optical ( a ), SAR ( b ) and despeckled SAR patches ( c ).CC-BY · From the PDF
Figure 4 Influence of the speckle filter and comparison of different network architectures during training time (all results are generated from the validation set): ( a ) shows the matching accuracy during training. Here, the matching accuracy is measured as the percentage of matching points, where the L 2 distance to the ground truth location is less than or equal to three pixels; ( b ) shows the average L 2 distance between the matching points and the ground truth location during training.
Figure 4 Influence of the speckle filter and comparison of different network architectures during training time (all results are generated from the validation set): ( a ) shows the matching accuracy during training. Here, the matching accuracy is measured as the percentage of matching points, where the L 2 distance to the ground truth location is less than or equal to three pixels; ( b ) shows the average L 2 distance between the matching points and the ground truth location during training.CC-BY · From the PDF
Figure 5 Illustration of influence of the raw score as a threshold: ( a ) the relation between the predicted score and the number of patches; ( b ) relation between the number of patches and the matching accuracy; ( c ) relation between the predicted score and the matching accuracy; and ( d ) relation between the predicted score and the average distance ( L 2 ) between the predicted matching points and the ground truth location. The matching accuracy in Figure 5b is measured as the percentage of matching points, where the L 2 distance to the ground truth location is less than three pixels and in Figure 5c less than 2, 3 and 4 pixels.
Figure 5 Illustration of influence of the raw score as a threshold: ( a ) the relation between the predicted score and the number of patches; ( b ) relation between the number of patches and the matching accuracy; ( c ) relation between the predicted score and the matching accuracy; and ( d ) relation between the predicted score and the average distance ( L 2 ) between the predicted matching points and the ground truth location. The matching accuracy in Figure 5b is measured as the percentage of matching points, where the L 2 distance to the ground truth location is less than three pixels and in Figure 5c less than 2, 3 and 4 pixels.CC-BY · From the PDF
Figure 6 Side by side comparison between ( a ) optical patches (201 × 201 pixels), ( b ) the score maps of NCC, ( c ) MI, and ( d ) our method (51 × 51 pixels), and ( e ) the reference despeckled SAR patches (251 × 251 pixels).
Figure 6 Side by side comparison between ( a ) optical patches (201 × 201 pixels), ( b ) the score maps of NCC, ( c ) MI, and ( d ) our method (51 × 51 pixels), and ( e ) the reference despeckled SAR patches (251 × 251 pixels).CC-BY · From the PDF
(a) Before the geo-localization enhancement of the optical image.
(a) Before the geo-localization enhancement of the optical image.CC-BY · From the PDF
(b) After the geo-localization enhancement of the optical image.
(b) After the geo-localization enhancement of the optical image.CC-BY · From the PDF

Verification

Claims the paper makes

The claims and statistics of this paper haven’t been checked yet.

Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.

A few free checks a day · papers already verified open instantly · results are public

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

The self-check needs the paper’s full text.

Citations

Cites

Show all 55 references

Cited by

See all 131

Metadata

Authors
Merkle, N., Luo, W., Auer, S., Müller, R., & Urtasun, R.
Affiliations
  • Deutsches Zentrum für Luft- und Raumfahrt e. V. (DLR) (DE)
  • University of Toronto (CA)
Corresponding author
Merkle, Nina
Venue
Remote Sensing
Publisher
MDPI AG
Published
2017-06-10
Volume
9
Issue
6
Pages
586
ISSN
2072-4292 (ISSN-L)
OpenAlex
W2608922915
Cited by
120
References
55
Citations by year
2026: 6 · 2025: 6 · 2024: 17 · 2023: 15 · 2022: 18 · 2021: 21 · 2020: 16 · 2019: 10 · 2018: 11
FWCI
8.51
Type
journal article
Language
en
Open access
Locations
  • Remote Sensing (publishedVersion, cc-by)
  • elib (German Aerospace Center) (acceptedVersion)
  • TSpace (University of Toronto) (submittedVersion, cc-by)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
Topics
  • Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Satellite Image Processing and Photogrammetry
  • Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Advanced Image and Video Retrieval Techniques
  • Physical Sciences › Engineering › Media Technology › Advanced Image Fusion Techniques
Keywords
geolocation, optical satellite imagery, synthetic aperture radar, TerraSAR-X, PRISM, image registration, Siamese network, urban environment

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.