Mobile-based 3D reconstruction of building environment

Lei Lei,Ying Zhou,Hanbin Luo

DOI: https://doi.org/10.3311/ccc2018-136

2018-01-01

Abstract:On-time environment perception of construction site is considered as an indispensable step for project management.Real-time tracking and feedback the status of construction facilitate progress monitoring and quality control.Image-based modelling and RGB-D mapping are considered as a non-invasive and low-cost technology which are always used for data collection and reconstruction of as-built building environments.Recently, the arrival of reliable and efficient computational of mobile terminal service has given us an opportunity to develop a mobile-based spatial data reconstruction system.Considering the capacity of processing and real-time performance on a mobile device, Oriented FAST and Rotated BRIEF (ORB) features are extracted.The ORB features are used for subsequent procedures, including tracking, mapping, relocalization and loop closing.In contrast to imagebased off-line modelling, a real time Simultaneous Localization and Mapping (SLAM) algorithm was utilized to estimate the camera trajectory while reconstruction the building environment.Keyframes selection strategy was proposed to reduce the redundant images and generate a robust and trackable sparse point clouds.The keyframes and sparse point clouds are transferred to a computer for generating dense point clouds, grid reconstruction and texture synthesis.Finally, the reconstruction result will be transferred back to mobile and can be displayed directly on a mobile device.As an initial effort, this paper investigated the potential of live reconstruction of indoor building scenes on an android mobile device.Taking the advantages of operable and portable, the system can be used for data acquisition of as-built information by construction workers.