This paper explores the applicability of different methods to construct a Building information System Model (BIM) from 2D plans and, more specifically, cadastral plans. The explored procedures use three different approaches to build BIMs: one manual, one semi-automatic based on Rhinoceros/Grasshopper, and one based on Artificial Intelligence (AI). To analyse the performance of the three methods, the case study of a masonry building located in an Italian historic centre is used. The fastest is the AI-based method, although it has some limitations in automatic recognition of some objects. This latter aspect was further explored by analysing a very large dataset of cadastral plans. The results show good recognition of wall elements, but problems arise in the recognition of other elements, such as doors and windows. The manual method, on the other hand, allowed for the construction of a detailed model of the considered structure. However, the method requires strong user interaction and longer time frames than the AI-based method. The semi-automatic based method requires quite a lot of pre-processing but through an algorithm in Grasshopper a fast and detailed modelling of the structure is obtained. The paper also outlines future scenarios for the full exploitation of cadastral data.
Read full abstract