Automated Tool for Visual Progress Monitoring in Construction

  • Author / Creator
    Martens, Oleksii
  • This thesis focuses on determining innovative computer vision algorithms suitable for progress tracking and forming them into automated visual progress tracking framework. The main concept of this approach is reconstruction of an as-built 3D point cloud model of the object of interest based on the spatial information extracted from photographs or videos. The reconstructed as-built model is then compared to the as-planned model, and the progress is reported based on correlation of the as-built and the as-planned models. The perspective framework key components are Structure from Motion, Multi-View Stereo, Coherent Point Drift or Iterative Closest Point, and the Hausdorff distance. At the initial phase, Structure from Motion takes a set of images as an input, estimates camera parameters for each image and produces a sparse point cloud. Next, the obtained data passes to Multi-View Stereo and the dense point cloud is generated. At this stage, the acquired point cloud is the as-built model. The next phase is aligning of the reconstructed as-built model to the corresponding as-planned model. The alignment transformation is calculated with either Coherent Point Drift or Iterative Closest Point. Finally, having the point cloud aligned, the progress is estimated. This phase is performed in three steps. First, the Hausdorff distance is calculated. Second, the as-planned model is color coded with a binary palette, where one color corresponds to the completed parts of the construction object and another corresponds to the parts that are to be built. Third, the ratio of completed points to the number of all points is computed. Finally, the color-coded progress model and percentage of completion are reported to the end user. The perspective cutting edge libraries for Structure from Motion are COLMAP, OpenMVG, VisualSFM, TheiaSFM, and MVE. The chosen libraries for Multi-View Stereo are COLMAP, OpenMVS, CMVS, CMPMVS and MVE. The libraries selected for point cloud alignment are CPD (Coherent Point Drift) and libpointmatcher (Iterative Closest Point). Finally, the Hausdorff distance is computed with Meshlab. The chosen libraries are integrated into frameworks and tested on real case study data obtained from a construction company. The case study experiment indicated successful performance of the following frameworks: COLMAP-COLMAP-CPD-Meshlab, COLMAP-COLMAP-libpointmatcher-Meshlab, and OpenMVG-OpenMVS-CPD-Meshlab. These pipelines demonstrated their capability of performing reliable and fast progress identification. Thus, the specified software combinations are suggested for construction progress tracking. The proposed frameworks improve the cutting-edge computer vision construction project progress monitoring approaches in terms of reconstruction quality and computation time. In fact, the suggested progress monitoring approach allows to reduce progress identification time from a few hours to a few minutes.

  • Subjects / Keywords
  • Graduation date
    Fall 2017
  • Type of Item
  • Degree
    Master of Science
  • DOI
  • License
    This thesis is made available by the University of Alberta Libraries with permission of the copyright owner solely for non-commercial purposes. This thesis, or any portion thereof, may not otherwise be copied or reproduced without the written consent of the copyright owner, except to the extent permitted by Canadian copyright law.