Extraction de connaissances spatio-temporelles incertaines pour la prédiction de changements en imagerie satellitale

Abstract : The interpretation of remotely sensed images in a spatiotemporal context is becoming a valuable research topic for dynamic phenomena. However, constant growth of the amount of data used in the remote image sensing field makes the manual analysis of satellite images a challenging task. Data mining has recently emerged as a promising research field that led to several interesting discoveries related to remote sensing. This thesis presents a new approach based on data mining to predict spatiotemporal land cover changes in satellite image databases. The proposed approach is divided into three steps: spatiotemporal modeling of satellite images, prediction of land cover changes and result interpretation. The proposed approach integrates three levels of imperfection processing: data related, prediction related and results related imperfection. In order to take into account imperfection related to data, a collaborative segmentation is performed. The goal is to reduce information loss when we attempt to model satellite images. Imperfection related to land cover change prediction is processed by applying a fuzzy decision tree in the prediction process. Decisions describing land cover changes are evaluated through a Case Based Reasoning (CBR) in order to retrieve relevant decisions. Upon completion of these individual processes, relevant decisions are combined through a high decision scheme to obtain more accurate and reliable decisions. The experimentation of the proposed approach is divided into two parts: application and evaluation. Results show good performance of the proposed approach measured in terms of precision accuracy comparatively with existing approaches.
Document type :
Theses
Complete list of metadatas

Cited literature [165 references]  Display  Hide  Download

https://tel.archives-ouvertes.fr/tel-00741990
Contributor : Bibliothèque Télécom Bretagne <>
Submitted on : Monday, October 15, 2012 - 4:10:59 PM
Last modification on : Thursday, October 17, 2019 - 12:35:30 PM
Long-term archiving on : Saturday, December 17, 2016 - 1:22:19 AM

File

2012telb0235_Boulila_Wadi.pdf
Files produced by the author(s)

Identifiers

  • HAL Id : tel-00741990, version 1

Citation

Wadii Boulila. Extraction de connaissances spatio-temporelles incertaines pour la prédiction de changements en imagerie satellitale. Traitement des images [eess.IV]. Télécom Bretagne, Université de Rennes 1, 2012. Français. ⟨tel-00741990⟩

Share

Metrics

Record views

342

Files downloads

1264