Augmented reality can improve construction and facility management by visualizing an as-planned model on its corresponding surface for fast, easy, and correct information retrieval. This requires the localization registration of an as-built model in an as-planned model. However, the localization and registration of indoor environments fail, owing to self-similarity in an indoor environment, relatively large as-planned models, and the presence of additional unplanned objects. Therefore, this paper proposes a computer vision-based method to (1) homogenize indoor as-planned and as-built models, (2) reduce the search space of model matching, and (3) localize the structure (e.g., room) for registration of the scanned area in its as-planned model. This method extracts a representative horizontal cross section from the as-built and as-planned point clouds to make these models similar, restricts unnecessary transformation to reduce the search space, and corresponds the line features for the estimation of the registration transformation matrix. The performance of this method, in terms of registration accuracy, is evaluated on as-built point clouds of rooms and a hallway on a building floor. A rotational error of 0.005 rad and a translational error of 0.088 m are observed in the experiments. Hence, the geometric feature described on a representative cross section with transformation restrictions can be a computationally cost-effective solution for indoor localization and registration.
For interactive visualization in AR devices, feature descriptors of point clouds (as-designed model and as-built model) are corresponded and registered. However, point cloud of indoor environment has lots of similar feature descriptors (e.g., indoor scene with similar doors and windows), which leads to many false correspondences and affect registration accuracy. This paper proposes a random sample consensus (RANSAC)-based false correspondence rejection to compute accurate transformation for the registration of such 3D point clouds. Point cloud data is collected from rooms and a hallway of a campus building, and transformation accuracy for the registration of those point clouds is tested. The results show that RANSAC-based false correspondence rejection gives transformation accuracy of 0.017 radians and 0.1924 meters in aligning two point cloud models, and hence the proposed registration approach of a model point cloud with scene point cloud may provide a foundation to accurately implement the AR on a construction jobsite.
Railway transportation plays a vital role in modern society, enabling the safe and efficient movement of people and goods over long distances. To ensure the longevity and safety of a railway infrastructure, the regular maintenance of tracks is crucial. Traditional track inspections, conducted manually to monitor geometric parameters and to identify defects, are time-consuming, labor-intensive, and prone to human error. Current Scan-to-BIM frameworks for railway maintenance also lack standardized methods for extracting geometric parameters that can be easily integrated into Building Information Modeling (BIM). Additionally, the Industry Foundation Classes (IFC) standard, used for BIM data exchange, does not support storing parameter values at specific chainage points along the track, limiting defect localization. A framework is proposed to address these challenges by standardizing the extraction of geometric parameters from point cloud data and ensuring seamless integration with BIM. The framework calculates parameters at station chainage points and generates additional chainage points along the track, associating the data with the corresponding chainage. A case study demonstrates the framework’s ability to enhance defect localization, using the EN 13848-5 European Standard to identify defects at specific chainages. Ultimately, this approach contributes to the more effective lifecycle management of railway tracks.