Real-time 3D reconstruction using SLAM for building construction

3D reconstruction in construction site is now receiving remarkable attention in construction management. Currently, researchers mainly focus on off-line reconstruction using image-based algorithms or directly obtain point clouds using laser scanner. Both methods are time-consuming. This research proposes a real-time 3D reconstruction system using simultaneous localization and mapping (SLAM). This system reaches 96.8% accuracy in measuring the outline length of a small cabinet. Moreover, the accuracy, data collecting time and data processing time are evaluated through comparing the results of off-line image-based method Structure from Motion (SfM) and laser scanner-based method. The new system achieves satisfactory accuracy rate and efficient data collecting and processing, however, many details in the appearance of objects were missing and there were many noise points in the 3D reconstruction model.