Morphological analysis of shape semantics from curvature-based signatures
The case study of the columns of Saint-Michel de Cuxa

Anthony Pamart, David Lo Buglio & Livio De Luca (novembre 2015)
Communication publiée dans IEEE proceedings of Digital Heritage 15 – International Congress 2015, Grenada, Spain, novembre 2015.

Abstract: Over the past few years, advancements in the field of 3D digitizing has increased the fidelity of geometric models. So far and despite of this acuity enhancement, a gap remains between the growth of collected data and its uses as vehicle of knowledge. New challenges have emerged to handle massive content of a 3D footprint. Considering those un- interpreted data as starting point for further investigations, the hypothesis is to rely on a “low-level” analysis of geometric features aiming to enrich informative and scientific value of “high-level” semantic studies. This article describes an approach using discrete curvature assets to link morphological identification and semantic characterization. The mean curvature has been parametrized to highlight it use as an eloquent shape description. At this point, a comparative analysis within an architectural collection composed of similar entities would be led according to the remoteness degree compared to an average geometric reference model. The introduced principle explores the construction of curvature-based signatures so as to reassess the conceptual articulations of 31 Romanesque columns from the cloister of the abbey of Saint-Michel de Cuxa.

Retour à la liste


Exemples de travaux

AlIce Blog

Jury des modules I (& III) ? impression des planches
16/01/17 by Denis
Vu qu’il est demandé d’imprimer les planches sur du papier à fort grammage et vu le coût [...]

Jury du 19/01/2017 : modalités
16/01/17 by Denis
Module I : Le jury aura lieu au labo de 9:30 à 15:30, et se base sur la liste de passage en fin de [...]

Planning du jeudi 12/01/2017
11/01/17 by Denis
Module I : Michel Lefèvre fera une permanence d’assistance au code de 10:00 à 13:00 au labo. [...]

Code : bien à vous