[RLF15a] Discriminatively trained model mixture object detection in aerial images

Conférence Internationale avec comité de lecture : Image Information Mining 2015, October 2015, pp.-, Romania,

Auteurs: H. Randrianarivo , B. Le Saux , M. Ferecatu

Mots clés: Object detector; aerial images; mixture of statistical models

Résumé: In this work we propose a new method for vehicle detection in very high resolution aerial images. Our model is based on a mixture of filters which capture the visual appearance of the object of interest. Each filter is discriminatively trained in order to model the implicit subcategories in the training dataset. We use an iterative hard-negative mining procedure to focus the detector on difficult samples. We assess our approach on several large datasets and show it tackles efficiently major problems in remote sensing such as orientation change and data size.

Equipe: vertigo

Collaboration: ONERA - DTIM

BibTeX

@inproceedings {
	RLF15a,
	title	=	"{Discriminatively trained model mixture object detection in aerial images}",
	author	=	" H. Randrianarivo and B. Le Saux and M. Ferecatu ",
	booktitle	=	"{Image Information Mining 2015}",
	year	=	2015,
	month	=	"October",
	pages	=	"-",
	address	=	" Romania",
}