Automatic Land Cover Reconstruction From Historical Aerial Images: An Evaluation of Features Extraction and Classification Algorithms - Archive ouverte HAL Accéder directement au contenu
Article Dans Une Revue IEEE Transactions on Image Processing Année : 2019

Automatic Land Cover Reconstruction From Historical Aerial Images: An Evaluation of Features Extraction and Classification Algorithms

Résumé

The land cover reconstruction from monochromatic historical aerial images is a challenging task that has recently known an increasing interest from the scientific community with the proliferation of large scale epidemiological studies involving retrospective analysis of spatial pattern. However, the efforts engaged by the computer vision community in remote sensing applications are mostly focused on prospective approaches through the analysis of high resolution multi-spectral data acquired by advanced spatial programs. Hence, four contributions are proposed in this article. They aim at providing a comparison basis for the future development of computer vision algorithms applied to the automation of the land cover reconstruction from monochromatic historical aerial images. Firstly, a new multiscale multi-date dataset composed of 4.9 million non-overlapping annotated patches of the France territory between 1970 and 1990 has been created with the help of Geography experts. This dataset has been named HistAerial (http://eidolon.univ-lyon2.fr/~remi1/HistAerialDataset/). Secondly, an extensive comparison study of state-of-the-art texture features extraction and classification algorithms including deep convolutional neural networks (DCNNs) has been performed. It is presented in the form of an evaluation. Thirdly, a novel low-dimensional local texture filter named Rotated-CorneR Local Binary Pattern (RCRLBP) is presented as a simplification of the Binary Gradient Contours filter through the use of an orthogonal combination representation. Finally, a novel combination of low-dimensional texture descriptors, including the R-CRLBP filter, is introduced as a Light Combination of Local Binary Patterns (LCoLBP). The LCoLBP filter achieved state-of-the-art results on the HistAerial dataset while conserving a relatively low-dimensional feature vector space compared with the DCNN approaches (17 times shorter). HistAerial : http://eidolon.univ-lyon2.fr/~remi1/HistAerialDataset/
Fichier principal
Vignette du fichier
ratajczak2019-tip-preprint.pdf (7.87 Mo) Télécharger le fichier
Origine : Fichiers produits par l'(les) auteur(s)
Loading...

Dates et versions

hal-02003932 , version 1 (07-05-2019)

Identifiants

Citer

Rémi Ratajczak, Carlos F Crispim-Junior, Élodie Faure, Béatrice Fervers, Laure Tougne. Automatic Land Cover Reconstruction From Historical Aerial Images: An Evaluation of Features Extraction and Classification Algorithms. IEEE Transactions on Image Processing, inPress, ⟨10.1109/TIP.2019.2896492⟩. ⟨hal-02003932⟩
1036 Consultations
434 Téléchargements

Altmetric

Partager

Gmail Facebook X LinkedIn More