An Analysis of Reconstruction Algorithms Applied to 3D Building Modeling

Abstract

Objectives: The process of reconstructing a city cannot be fully automated. More human intervention is required to achieve high quality mass modeling. Our objectives are to provide a guide for the 3D graphic designers and 3D modelers in developing 3D building models. Methods: We present an up-to-date survey of the reconstruction algorithms which are further classified as point cloud based reconstruction methods and structure-aware shape reconstruction methods. We introduce various existing 3D reconstruction techniques and have compared those techniques based on shape simplification, surface completion and shape modeling methodologies. Findings: We have identified the pipeline process and technical issues encountered in each phase of the reconstruction process from a given image layout. As stated, the inverse procedural reconstruction approaches are best oriented towards the reconstruction of buildings from a given input 2D image. Also, there is a lack of systems that are capable of generating procedural rules from user interaction. Improvements: We propose a new 3D reconstruction methodology and novel framework that combines component based structure capturing and rule based modeling techniques that would best be suitable for 3D building modeling, resembling the real world buildings in future systems.