Centre de documentation
scientifique

Learning a discriminative distance metric with label consistency for scene classification / Yuebin Wang in IEEE Transactions on geoscience and remote sensing, vol 55 n° 8 (August 2017)  

Public [article]  inIEEE Transactions on geoscience and remote sensing > vol 55 n° 8 (August 2017) . - pp 4427 - 4440 Titre : Learning a discriminative distance metric with label consistency for scene classification Type de document : Article/Communication Auteurs : Yuebin Wang, Auteur ; Liqiang Zhang, Auteur ; Hao Deng, Auteur ; et al., Auteur Année de publication : 2017 Article en page(s) : pp 4427 - 4440 Note générale : Bibliographie Langues : Anglais (eng) Descripteur : [Vedettes matières IGN] Traitement d'image optique [Termes IGN] algorithme d'apprentissage [Termes IGN] apprentissage dirigé [Termes IGN] image hyperspectrale [Termes IGN] métrique [Termes IGN] précision de la classification Résumé : (Auteur) To achieve high scene classification performance of high spatial resolution remote sensing images (HSR-RSIs), it is important to learn a discriminative space in which the distance metric can precisely measure both similarity and dissimilarity of features and labels between images. While the traditional metric learning methods focus on preserving interclass separability, label consistency (LC) is less involved, and this might degrade scene images classification accuracy. Aiming at considering intraclass compactness in HSR-RSIs, we propose a discriminative distance metric learning method with LC (DDML-LC). The DDML-LC starts from the dense scale invariant feature transformation features extracted from HSR-RSIs, and then uses spatial pyramid maximum pooling with sparse coding to encode the features. In the learning process, the intraclass compactness and interclass separability are enforced while the global and local LC after the feature transformation is constrained, leading to a joint optimization of feature manifold, distance metric, and label distribution. The learned metric space can scale to discriminate out-of-sample HSR-RSIs that do not appear in the metric learning process. Experimental results on three data sets demonstrate the superior performance of the DDML-LC over state-of-the-art techniques in HSR-RSI classification. Numéro de notice : A2017-498 Affiliation des auteurs : non IGN Thématique : IMAGERIE/INFORMATIQUE Nature : Article nature-HAL : ArtAvecCL-RevueIntern DOI : 10.1109/TGRS.2017.2692280 En ligne : http://dx.doi.org/10.1109/TGRS.2017.2692280 Format de la ressource électronique : URL article Permalink : https://documentation.ensg.eu/index.php?lvl=notice_display&id=86440 [article]

Automatic generation of 2.5D terrain models without occluding routes of interest / Hao Deng in Photogrammetric Engineering & Remote Sensing, PERS, vol 78 n° 11 (November 2012)  

Public [article]  inPhotogrammetric Engineering & Remote Sensing, PERS > vol 78 n° 11 (November 2012) . - pp 1175 - 1185 Titre : Automatic generation of 2.5D terrain models without occluding routes of interest Type de document : Article/Communication Auteurs : Hao Deng, Auteur ; L. Zhang, Auteur ; J. Ma, Auteur ; et al., Auteur Année de publication : 2012 Article en page(s) : pp 1175 - 1185 Note générale : Bibliographie Langues : Anglais (eng) Descripteur : [Vedettes matières IGN] Applications photogrammétriques [Termes IGN] données localisées 2,5D [Termes IGN] modèle numérique de terrain [Termes IGN] montagne [Termes IGN] réseau routier [Termes IGN] visibilité Résumé : (Auteur) When a car drives in mountainous regions, the views based on conventional perspective projection often suffer from features of interest being occluded. We propose a method for generating disocclusion views in mountainous regions. The terrain is segmented to build a potential set of occluders; and then an optimized viewpoint is determined, and elevations are rearranged. To obtain a smooth deformed terrain, a smooth displacement function is introduced to deform the level-of-detail terrain models. Compared with previous methods, the merit of this study lies in automatically generating disocclusion views with temporal coherence while keeping the details of the deformed terrain the same as the original terrain. Experiments performed on the 4098 pixel x 4098 pixel mountainous terrain landscape prove that the disocclusion views can achieve 42 to 58 frames/second. Moreover, the shapes of the features of interest on the driving route without occlusion and the spatial configuration of geographical landmarks in its neighborhood can be easily recognized. Numéro de notice : A2012-585 Affiliation des auteurs : non IGN Thématique : IMAGERIE Nature : Article DOI : 10.14358/PERS.78.11.1175 En ligne : https://doi.org/10.14358/PERS.78.11.1175 Format de la ressource électronique : URL article Permalink : https://documentation.ensg.eu/index.php?lvl=notice_display&id=32031 [article]