Building modellingProcessing and data structure

Resum

The knowledge of the real condition of buildings, especially in those cases where there is a lack of a repository of information such as heritage sites, is essential for their conservation and correct maintenance. This issue becomes critical with the global preoccupation with climate change and energy efficiency. BIM (Building Information Models) allows integrating and sharing all the information of buildings along their whole life cycle. However, the compilation of the geometric data and generation of the models can be a time-consuming task and become a bottleneck for the realization of their following applications. Therefore, there is an interest in the development of faster and easier techniques of data acquisition and automation of their processing. LiDAR-based (Light Detection And Ranging) technologies are one of the most used for the acquisition of geometric data, especially with the development of indoor mobile devices in recent years. They are easy and robust enough to use for researchers, technicians, and non-specialist users, allowing simpler and faster acquisitions in indoor environments. The proof of their interest is the continuous increase of LiDAR sensors in the market for the last years, with advanced features. These devices generate a discrete set of measures translated to a set of 3D points in digital standard formats ready to use in the modelling process of the structures. The raw data acquired can be processed, obtaining semantic information to include in the BIM, such as the different structures detected and the relation between them, allowing the use of the information by a wide range of final users. This can be a hard task that requires time and specialist users. For this reason, there is an interest in the automation of the modelling of the acquired raw data for its inclusion into the BIM, identifying the different elements that form the building and incorporating semantic information about their relations and their condition, without increasing time-consumption and tediousness of the task. This Doctoral Thesis proposes different workflows that converge into a methodology for the 3D modelling of buildings from their indoor point clouds. All of them are based on the use of point clouds from mobile devices and begin with the Manhattan World Assumption, first using it as a hypothesis for the modelling and then trying to overcome it. They aim at advancing the automatic 3D modelling of the complete indoor buildings, extracting semantic information, its integration in BIM and its use in posterior thermal or structural analysis. The final use of the BIM is key for the establishment of the format required, which should always be compatible with the tools to be used (such as tools for energy analysis or tools for structural analysis). All the developments in this Doctoral Thesis are based on the philosophy of open-source software. A review of commercial indoor mapping systems, with different scanning technologies and physical configurations, comparing their main features for their practical use, the development of a procedure for an efficient and simplified 3D modelling of the interior of buildings, and a methodology for the semi-automatic 3D modelling of roofs using indoor point clouds, are the main components of this Doctoral Thesis. The algorithms developed for each methodology share the same hypothesis that surfaces have no thickness. The first methodology obtains the 3D model of an indoor point cloud, using LiDAR data, decomposing each room into its planar surfaces. This development applies the Manhattan Wold Assumption, for geometric modelling. The model includes the semantic data of the relations between adjacent surfaces and their belonging to the different rooms. The results are transcribed into a gbXML file for use in later thermal analysis. The second methodology is focused on the 3D modelling of roofs, completing one of the limitations in previous developments: the application of the Manhattan World Assumption, in such a way that more building typologies can be subjected to this methodology. The results are transcribed into a gbXML file too, expanding the possibilities of thermal analysis. The 3D modelling of roofs is complemented with additional work that integrates algorithms for the estimation of the reticular model of the structure in timber cellars, providing in these cases valuable structural information of the real state of the construction. The results obtained from each workflow were promising, in such a way that each workflow has been presented in a scientific article published in international journals with high impact indices and in international conferences. Therefore, this Doctoral Thesis is structured as a compendium of five publications, three of them published in international journals indexed in the Journal Citation Report (JCR), and two presented in International Conferences. The main contribution of the Doctoral Thesis to society is the advance in the development of automatic tools for the knowledge of the real condition of buildings, not only in academic fields but also for its use in industrial applications.