Model-driven and data-driven approaches using LIDAR data : analysis and comparison
Résumé
Following the development of 3D data acquisition techniques like digital photogrammetry, remote sensing and airborne laser
scanning, the automatic building modelling is still a challenging task. Indeed, it takes a growing place in many research fields like 3D
city modelling, cartographic analysis, urban planning, visualization and Geographic Information Systems (GIS) database
construction. Generally, two types of approaches characterize the automatic building reconstruction. The first one is a model-driven
approach. This parametric modelling approach consists of searching the most appropriate model among basic building models
contained in a models library. The second one is a data-driven approach, also called non-parametric modelling approach. This
technique attempts to model a primitive or a complex building by using series of more or less complex operations. It allows the
generation of a model without belonging to a specific library. After an extensive state of the art, this paper confronts methods
belonging to each type of automatic building construction approach. Based on a concrete experiment, the essential points
characterizing each approach, including their concept and the obtained model characteristics are analysed. It is of great interest to
study these issues, since not many of such investigations have been made before and have delivered the useful formulas.
Consequently, improvements are introduced and have led to develop a new method for each approach. After comparison of the
accuracy of these two methods and the characteristics of the resulting models, some solutions and recommendations are proposed.
scanning, the automatic building modelling is still a challenging task. Indeed, it takes a growing place in many research fields like 3D
city modelling, cartographic analysis, urban planning, visualization and Geographic Information Systems (GIS) database
construction. Generally, two types of approaches characterize the automatic building reconstruction. The first one is a model-driven
approach. This parametric modelling approach consists of searching the most appropriate model among basic building models
contained in a models library. The second one is a data-driven approach, also called non-parametric modelling approach. This
technique attempts to model a primitive or a complex building by using series of more or less complex operations. It allows the
generation of a model without belonging to a specific library. After an extensive state of the art, this paper confronts methods
belonging to each type of automatic building construction approach. Based on a concrete experiment, the essential points
characterizing each approach, including their concept and the obtained model characteristics are analysed. It is of great interest to
study these issues, since not many of such investigations have been made before and have delivered the useful formulas.
Consequently, improvements are introduced and have led to develop a new method for each approach. After comparison of the
accuracy of these two methods and the characteristics of the resulting models, some solutions and recommendations are proposed.
Origine :
Fichiers produits par l'(les) auteur(s)
Loading...