Séminaire : MRF and Dempster-Shafer theory for simultaneous shadow/vegetation detection on high resolution aerial color images

Tran Thanh NGO, doctorant au sein de l'équipe MIV,  fera une présentation de ses travaux de recherche le lundi 30 juin 2014 à 10h30 dans la salle C218 au Pôle API d’Illkirch.

Titre : MRF and Dempster-Shafer theory for simultaneous shadow/vegetation detection on high resolution aerial color images (comité de mi-thèse)

Résumé : In recent years, the effects of natural catastrophes and human activities have emphasized the need for developing a broader view of the Earth's surface. According to cartographic experts, shadow and vegetation can provide additional geometric and semantic clues about the state of buildings after natural catastrophes. Current shadow/vegetation detection methods in the literature detect separately shadow regions and vegetation regions. The drawback of these methods is that, for example, a vegetated pixel covered by shadow can be classified as vegetation (by a vegetation detection algorithm), and at the same time as shadow (by a shadow detection algorithm). Thus, these methods can not rovide a sufficiently good segmentation map. In fact, visual inspection also has a similar problem since the pixel information is imprecise and uncertain. In this context, we present a new method for simultaneously detecting shadows and vegetation in remote sensing images, based on Otsu's thresholding method and Dempster-Shafer fusion which aims at combining different shadow indices and vegetation indices in order to increase the information quality and to obtain a more reliable and accurate segmentation result. The notion of mass functions , in the Dempster-Shafer evidence theory is linked to the Gaussian distribution, and the DS fusion, which is carried out pixel by pixel, is incorporated in the Markovian context while obtaining the optimal segmentation with the energy minimization scheme associated with the Markov random field.