M-ESTIMATOR SAMPLE CONSENSUS PLANAR EXTRACTION FROM IMAGE-BASED 3D POINT CLOUD FOR BUILDING INFORMATION MODELLING

Abstract

Building Information Models (BIMs) are used as an official standard to manage information
in the construction industry. Creating a BIM model is still a laborious process, especially in existing
buildings, where digital CAD models are often not available. The current process of creation of a BIM model
for existing structures generally involves the generation of a geometric model from a 3D point cloud,
commonly created by a laser scan, which can be time consuming and expensive. However, image-based 3D
modeling techniques are more economical and efficient. In this paper, a 3D point cloud from images and
laser scan were used to detect planes using plane detection algorithms based on the M-estimator Sample
Consensus (MSAC). The accuracy of 3D point clouds of both techniques was compared using the Iterative
Closest Point algorithm. The sample data used in the study was obtained from a laboratory, which contains
3D points from many visible planes, such as walls and floors. The MSAC algorithm was applied to detect
planes in the 3D point clouds from the image-based and laser scanning techniques. The parameters derived
from the plane detection algorithms were subsequently used to create BIM models through the eXtensible
Building Information Modeling (xBIM). The proposed plane detection algorithm shows promising results
with a low mean square error. These results demonstrate that a BIM model can be created from an imagebased 3D point cloud, which is more convenient and easy to use than the point cloud from a laser scanner.