Reanimating aerial images in real time for geospatial web applications

Alexandre Devaux (1) , Mathieu Brédif (1)

1 MATIS - Méthodes d'Analyses pour le Traitement d'Images et la Stéréorestitution

Geovisualization has concentrated most work in displaying static data for a long time, progress have brought incredibly detailed visualizations offline but also in web browsers thanks to GPU access. Heavy textured meshes mixed with vectorial data are common representations for users now. Animation, on the other hand, is just at its beginning. As far as we know there are no current applications that integrate imaged-based dynamic visualization of many distinct objects at different scales. We propose a method generating dynamic representations for objects present in the aerial image that change appearance or position over time such as cars, water, boats, vegetation without the need for a user intervention. It offers a compelling experience, emphasizing object class differentiation. This research involves two main tasks. First to be able to extract object layers and direction using aerial images and land covers information in real time, and secondly to animate these objects realistically, combining computer vision and computer graphics domains.

Mots clés: animation; image-based animation; image-based rendering; geovisualization; aerial imagery; gis; web mapping

Domaines: Synthèse d'image et réalité virtuelle [cs.GR]

Alexandre Devaux, Mathieu Brédif. Reanimating aerial images in real time for geospatial web applications. Journées Françaises d’Informatique Graphique, Oct 2017, Rennes, France. ⟨hal-01893899⟩