No abstract

In this work, we improved the efficiency and accuracy of our proposed cylinder-based registration and model fitting method of point clouds by terrestrial laser scanners for the as-built modeling of piping systems. Our algorithm simultaneously calculated the scanner parameters and cylinder parameters to avoid the propagation of registration errors and modeling errors. Coarse registration is performed by finding the alignment of the cylinder axes based on the random sample consensus approach and using a hash table. The efficiency of the coarse registration is improved by introducing a three-dimensional hash table. The fine registration and modeling is performed by minimizing the fitting errors of the cylinders as a nonlinear function of the positional and geometric parameters of scanners and cylinders. An iteratively reweighted least squares method is applied to the fine registration and modeling, leading to improved robustness. Moreover, for the modeling of pipes that are slightly bent due to gravity, incident angle filtering of scanned points and cylinder subdivision of the pipes to be modeled are introduced. The efficiency and robustness of the improved algorithm were compared with the previous approach using both artificial and real point clouds. The effectiveness of incident angle filtering and cylinder subdivision was confirmed. The proposed algorithm achieved the level of cylindrical modeling precision required for the renovation work of piping systems.

Abstract. Structure-from-Motion (SfM) and Multi-View Stereo (MVS) are widely used methods in three dimensional (3D) model reconstruction for an infrastructure maintenance purpose. However, if a set of images is not captured from well-placed positions, the final dense model can contain low-quality regions. Since MVS requires a much longer processing time than SfM as larger amounts of images are provided, it is impossible for surveyors to wait for the SfM–MVS process to complete and evaluate the geometric quality of a final dense model on-site. This challenge results in response inefficiency and the deterioration of dense models in 3D model reconstruction. If the quality of the final dense model can be predicted immediately after SfM, it will be possible to revalidate the images much earlier and to obtain the dense model with better quality than the existing SfM–MVS process. Therefore, we propose a method for reconstructing a more plausible 3D mesh model that accurately approximates the geometry of the final dense model only from sparse point clouds generated from SfM. This approximated mesh model can be generated using Delaunay triangulation for the sparse point clouds and triangle as well as tetrahedron filtering. The approximated model can be used to predict the geometric quality of the final dense model and for an optimization-based view planning. Some experimental results showed that our method is effective in predicting the quality of the final dense model and finding the potentially degraded regions. Moreover, it was confirmed that the average reconstruction errors of the dense model generated by the optimization-based view planning went below tens of millimeters and falls within an acceptable range for an infrastructure maintenance purpose.

<p><strong>Abstract.</strong> In this paper, we introduce a method for predicting the quality of dense points and selecting low-quality regions on the points generated by the structure from motion (SfM) and multi-view stereo (MVS) pipeline to realize high-quality and efficient as-is model reconstruction, using only results from the former: sparse point clouds and camera poses. The method was shown to estimate the quality of the final dense points as the quality predictor on an approximated model obtained from SfM only, without requiring the time-consuming MVS process. Moreover, the predictors can be used for selection of low-quality regions on the approximated model to estimate the next-best optimum camera poses which could improve quality. Furthermore, the method was applied to the prediction of dense point quality generated from the image sets of a concrete bridge column and construction site, and the prediction was validated in a time much shorter than using MVS. Finally, we discussed the correlation between the predictors and the final dense point quality.</p>

No abstract