Browsing by Author "Steward, Jeremy"

Now showing 1 - 2 of 2
  • Thumbnail Image
    ItemOpen Access
    Explanation for the seam line discontinuity in terrestrial laser scanner point clouds
    (Elsevier, 2019-08) Lichti, Derek D.; Glennie, Craig L.; Al-Durgham, Kaleel; Jahraus, Adam; Steward, Jeremy
    The so‐called seam line discontinuity is a phenomenon that can be observed in point clouds captured with some panoramic terrestrial laser scanners. It is an angular discontinuity that is most apparent where the lower limit of the instrument’s angular field‐of‐view intersects the ground. It appears as step discontinuities at the start (0° horizontal direction) and end (180°) of scanning. To the authors’ best knowledge, its cause and its impact, if any, on point cloud accuracy have not yet been reported. This paper presents the results of a rigorous investigation into several hypothesized causes of this phenomenon: differences between the lower and upper elevation angle scanning limits; the presence of a vertical circle index error; and changes in levelling during scanning. New models for the angular observations have been developed and simulations were performed to independently study the impact of each hypothesized cause and to guide the analyses of real datasets. In order to scrutinize each of the hypothesized causes, experiments were conducted with seven real datasets captured with six different instruments: one hybrid‐architecture scanner and five panoramic scanners, one of which was also operated as a hybrid instrument. This study concludes that the difference between the elevation angle scanning limits is the source of the seam line discontinuity phenomenon. Accuracy assessment experiments over real data captured in an indoor test facility demonstrate that the seam line discontinuity has no metric impact on the point clouds.
  • Thumbnail Image
    ItemOpen Access
    Range Camera Motion Capture: Geometric Parameter Extraction from Human Motion Data in Point Clouds
    (2017) Steward, Jeremy; Lichti, Derek; Shahbazi, Mozhdeh; Yanushkevich, Svetlana
    Motion capture systems presently face many problems with soft-tissue artefacts; and the cost of long set up, examination, and turnover times on actionable data. Soft-tissue artefacts are resultant from sensor choice, whenever affixed markers or sensors are required. Moreover, long set up times are driven from the need for subjects to wear specialized markers, and turnover times largely driven by the complex spatial processing of these targets or markers. To work around these limitations, a markerless system is desired. This thesis describes a set of processes to register, segment, and model a full 3D time-series of motion data. This is realized in the development of the RCMOCAP system, which is a markerless system for motion capture. Coordinate registration, body extraction, limb segmentation, and parameter modeling processes developed for complete 3D motion coverage have been evaluated in the context of real-world motion and effects, captured through the RCMOCAP system.