|
[n° ou bulletin]
est un bulletin de Blog de la RFPT / Société française de photogrammétrie et de télédétection (2021-) ![]()
[n° ou bulletin]
| ![]() |
Dépouillements
![](./images/expand_all.gif)
![](./images/collapse_all.gif)
Feature matching for multi-epoch historical aerial images: A new pipeline feature detection pipeline in open-source MicMac / Lulin Zhang in Blog de la RFPT, sans n° ([17/11/2021])
![]()
[article]
Titre : Feature matching for multi-epoch historical aerial images: A new pipeline feature detection pipeline in open-source MicMac Type de document : Article/Communication Auteurs : Lulin Zhang , Auteur ; Ewelina Rupnik
, Auteur ; Marc Pierrot-Deseilligny
, Auteur
Année de publication : 2021 Note générale : Bibliographie Langues : Anglais (eng) Descripteur : [Vedettes matières IGN] Traitement d'image optique
[Termes IGN] appariement de formes
[Termes IGN] chaîne de traitement
[Termes IGN] corrélation à l'aide de traits caractéristiques
[Termes IGN] détection de changement
[Termes IGN] données multitemporelles
[Termes IGN] image aérienne
[Termes IGN] image ancienne
[Termes IGN] image RVB
[Termes IGN] MicMac
[Termes IGN] modèle numérique de surfaceRésumé : (Auteur) [Introduction] ... We propose a fully automatic approach to computing robust inter-epoch feature correspondences. Our method consists of two steps: a rough co-registration by finding feature correspondences between DSMs (Digital Surface Model) derived within single epochs, and a precise feature matching on original RGB images. Our main contributions include:
- Rough-to-precise matching strategy that helps to drastically reduce ambiguity. In particular, we use the depth information to roughly co-register our epochs. The 3D landscape is globally stable over time and provides sufficient correspondences through time. Once co-registered, we levarage the 3D a priori to narrow down the search space in precise matching.
- Upscaling of the learning based feature matching algorithms to high resolution imagery. To do that, we introduced an image tiling scheme.
In the following we present the methodology and some experiments. If you are interested in using our method, please refer to the source code of MicMac Github 4, as well as 2 jupyter tutorials 5 6. We also provide an introduction video 7, slides 8 and project website 9. The datasets used in our publication 3 will be soon published in an open-access repository.Numéro de notice : A2021-840 Affiliation des auteurs : UGE-LASTIG (2020- ) Thématique : IMAGERIE Nature : Article nature-HAL : ArtSansCL DOI : sans Date de publication en ligne : 17/11/2021 En ligne : https://rfpt-sfpt.github.io/blog/feature%20matching/historical%20images/multi-ep [...] Format de la ressource électronique : URL Article Permalink : https://documentation.ensg.eu/index.php?lvl=notice_display&id=99042
in Blog de la RFPT > sans n° [17/11/2021][article]