Skip to Main content Skip to Navigation
Theses

Contribution de la texture pour l'analyse d'images à très haute résolution spatiale : application à la détection de changement en milieu périurbain

Antoine Lefebvre 1
1 LETG - Caen - Littoral, Environnement, Télédétection, Géomatique
LETG - Littoral, Environnement, Télédétection, Géomatique UMR 6554
Abstract : Remotely sensed images at very high spatial resolution (VHSR) constitute an important source of information for land cover change detection over small areas. They are particularly interesting for applications in urban and peri-urban areas because they can cope with sudden and irregular changes as well as regular and subtle changes. However the use of VHSR images needs methodological improvements, since usual change detection approaches, based on a pixel analysis, have been developed to process low and medium resolution images and can unfortunately not be applied for VHSR data. On one hand, spectral resolutions of VHSR sensors are often lower than those of other sensors, the spectral resolution decreasing with increasing spatial resolution. On the other hand, the spectral variability of pixels defining classes of land cover increases with the spatial resolution, resulting in a decrease in the separability between different classes of land cover and limitations of conventional classification methods based on pixel values. This thesis presents a series of methodological tools which allow to automatically identify and characterize changes over small areas from VHRS images acquired at different dates and from different sources. The originality of the tools presented here relies on the comparison of the textural properties of the objects of interest, while not using exclusively or primarily spectral information of the image but rather focusing on spatial information. The texture characterization is based on the analysis of the coefficients from a wavelet decomposition. The developed tools include : a preprocessing method dedicated to the vignetting effect correction for old aerial photographs ; an image segmentation approach ; a method for texture orientation estimation ; a classification method ; a change detection method. All these tools have been validated using synthetic examples and then applied in the suburban environment of the city of Rennes using aerial photographs acquired in 1978 and 2001. The classification rates, which ranged from 78 % to 85 %" show the importance of texture for classification of VHSR images. They highlight the possibility of detecting different types of changes and thus to distinguish changes related to agricultural practices and urbanization. The tools developed in this thesis are generic and can be applied for the analysis of any textured pattern. To illustrate this, we have applied some of the proposed tools for detecting and characterizing vineyards and estimating fluid motions from images
Complete list of metadatas

https://tel.archives-ouvertes.fr/tel-00596339
Contributor : Laurence Leroux <>
Submitted on : Friday, May 27, 2011 - 9:50:42 AM
Last modification on : Wednesday, October 14, 2020 - 4:09:28 AM
Long-term archiving on: : Sunday, August 28, 2011 - 2:21:24 AM

Identifiers

  • HAL Id : tel-00596339, version 1

Citation

Antoine Lefebvre. Contribution de la texture pour l'analyse d'images à très haute résolution spatiale : application à la détection de changement en milieu périurbain. Géographie. Université Rennes 2; Université Européenne de Bretagne, 2011. Français. ⟨NNT : 2011REN20010⟩. ⟨tel-00596339⟩

Share

Metrics

Record views

2262

Files downloads

7121