Influence of the methodology (pixel-based vs object-based) to extract urban vegetation from VHR images in different urban zones

Nathalie Long (1) , Arnaud Bellec (1) , Erwan Bocher (2) , Gwendall Petit (2)

1 LIENSs - LIttoral ENvironnement et Sociétés
2 IRSTV - Institut de Recherche en Sciences et Techniques de la Ville - FR 2488

In a context of sustainable development, the cities have become the main problem for spatial planning directives. Rapid urbanization has a real impact on several natural processes like the meteorological cycles. So to reduce the side-effects of urban development, consistent and accurate data are required to define urban land cover and, especially, the vegetation surface. Indeed, the vegetation presents a particular interest because of its influence on components of the urban environment like temperature or air quality. Images with a very high spatial resolution allow to obtain more details. If object-based approach (OBA) seems to give better results than pixel-based approach (PBA) to classify urban land covers, at a zone scale, this statement can be qualified according to the type of zone. On seven zones of Nantes, France, results of OBA and PBA are compared from 10 m spatial resolution image and impact of spatial resolution of images is also analyzed. Furthermore, this study explores a straightforward spatial analysis method to assess the impact of OBA and PBA. On zones where the land cover is more homogeneous, PBA and 10m image give vegetation surfaces close to surfaces obtained from OBA and 2,5m image. Zones where land covers are mixed, OBA and 2,5m image are essential to detect vegetation accurately.

Mots clés: Remote sensing; Spatial analysis; Image processing and analysis; Vegetation

Domaines: Géographie

Nathalie Long, Arnaud Bellec, Erwan Bocher, Gwendall Petit. Influence of the methodology (pixel-based vs object-based) to extract urban vegetation from VHR images in different urban zones. 5th Geobia Conference, May 2014, Thessaloniki, Greece. ⟨halshs-01093260⟩