Abstract
Abstract. Horizontal levels are references entities, the base of man-made environments. Their creation is the first step for various applications including the BIM (Building Information Modelling). BIM is an emerging methodology, widely used for new constructions, and increasingly applied to existing buildings (scan-to-BIM). The as-built BIM process is still mainly manual or semi-automatic and therefore is highly time-consuming. The automation of the as-built BIM is a challenging topic among the research community. This study is part of an ongoing research into the scan-to-BIM process regarding the extraction of the principal structure of a building. More specifically, here we present a strategy to automatically detect the building levels from a large point cloud obtained with a terrestrial laser scanner survey. The identification of the horizontal planes is the first indispensable step to produce an as-built BIM model. Our algorithm, developed in C++, is based on plane extraction by means of the RANSAC algorithm followed by the minimization of the quadrate sum of points-plane distance. Moreover, this paper will take an in-depth look at the influence of data resolution in the accuracy of plane extraction and at the necessary accuracy for the construction of a BIM model. A laser scanner survey of a three floors building composed by 36 scan stations has produced a point cloud of about 550 million points. The estimated plane parameters at different data resolution are analysed in terms of distance from the full points cloud resolution.
Highlights
The creation of horizontal levels is the first step in many applications concerning 3D building reconstruction, such as those relative to the Building Information Model (BIM), since they are the main reference of a structure which host the building elements
The as-built BIM model is derived from a point cloud, mainly obtained by a terrestrial laser scanner survey, and the process is called scan-to-BIM
RANSAC (Fischler and Bolles, 1981) is a very simple iterative procedure based on the following three steps: 1.three random points are extracted from the point cloud; 2.the direction numbers n : [a,b,c] of the plane defined by the three points are calculated; 3.the number of points which are less than a certain distance from the plane is calculated
Summary
The creation of horizontal levels is the first step in many applications concerning 3D building reconstruction, such as those relative to the Building Information Model (BIM), since they are the main reference of a structure which host the building elements. The as-built BIM model is derived from a point cloud, mainly obtained by a terrestrial laser scanner survey, and the process is called scan-to-BIM. It deals with a very complex problem: the need to generalize the survey and to generate an implicit model (Surfaces) out of an explicit model (point clouds). While region growing is very fast in image segmentation, before applying to unstructured point cloud it requires a data organization to permit rapid access. RANSAC is widely used to detect and extract planes in laser scanning data due to its simplicity and robustness
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
