Approche non supervisée par processus ponctuels marqués pour l'extraction d'objets à partir d'images aériennes et satellitaires

Saima Ben Hadj 1, 2 Florent Chatelain 3 Xavier Descombes 1, 2 Josiane Zerubia 1
1 ARIANA - Inverse problems in earth monitoring
CRISAM - Inria Sophia Antipolis - Méditerranée , Laboratoire I3S - SIS - Signal, Images et Systèmes
3 GIPSA-SAIGA - SAIGA
GIPSA-DA - Département Automatique, GIPSA-DIS - Département Images et Signal
Abstract : Previously, marked point process models have been successfully applied to solve problems of feature network extraction from aerial and satellite high resolution photographs. The main advantage of these models is that they take into account object geometry. In particular, objects in the image are represented by a set of geometric shapes which are governed by two types of energy: a data energy term that links objects to the processed image, and a regularizing energy term which introduces some prior knowledge about the observed scene. Furthermore, some parameters reflecting the importance of these energies are involved in the definition of this model. Thus, to ensure automatic object extraction, an estimation method of these parameters, based on the Stochastic Expectation-Maximization algorithm was studied and showed promising results on a simple example of a marked point process where the considered objects are disks. We propose to extend this study to extract more general geometric shapes, such as ellipses and rectangles. Hence, we deal with several applications, namely flamingo extraction, tree crown extraction, detection of boats in a seaport, building footprint detection and refugee tent detection. The main originality of this work consists in developing new energy components that allow to model boat alignment and building interactions within an unsupervised framework.
Liste complète des métadonnées

Cited literature [15 references]  Display  Hide  Download

https://hal.inria.fr/hal-00638665
Contributor : Saima Ben Hadj <>
Submitted on : Tuesday, November 8, 2011 - 2:04:14 AM
Last modification on : Tuesday, November 6, 2018 - 1:14:46 AM
Document(s) archivé(s) le : Thursday, February 9, 2012 - 2:21:12 AM

File

SFPT_SBH_FC_XD_JZ.pdf
Publisher files allowed on an open archive

Identifiers

  • HAL Id : hal-00638665, version 1

Citation

Saima Ben Hadj, Florent Chatelain, Xavier Descombes, Josiane Zerubia. Approche non supervisée par processus ponctuels marqués pour l'extraction d'objets à partir d'images aériennes et satellitaires. Revue Française de Photogrammétrie et de Télédétection, Société Française de Photogrammétrie et de Télédétection, 2010, 194, pp.2-15. ⟨hal-00638665⟩

Share

Metrics

Record views

845

Files downloads

502