Adaptive Processing of Laser Scanning Data and Texturing of the Segmentation Outcome
Date
2014-08-06
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Over the past few years, laser scanning systems (airborne, static terrestrial, and mobile terrestrial systems) have been acknowledged as the leading tools for the collection of high density 3D point clouds over physical surfaces for many different applications (3D city modelling, transportation planning, emergency response, cultural heritage documentation, structural health monitoring, and industrial site modelling). However, no interpretation and scene classification is performed during data acquisition. Consequently, the collected data must be processed to extract the required information. To date, a variety of techniques have been developed for the processing of laser scanning data, but the majority of them do not consider the internal characteristics of the laser scanning data and their impact on the quality of the extracted information from these data. In order to resolve these limitations, this dissertation introduces a new framework for the adaptive processing of laser scanning data. This processing procedure is developed while considering the possibility of its application to point clouds with varying internal characteristics (multi-platform/multi-resolution laser scanning point clouds).
In this adaptive processing procedure, a new laser scanning data characterization step is initially proposed to classify the laser scanning point cloud into planar, linear/cylindrical, and rough neighborhoods and to quantify its internal characteristics (i.e., local point density variations and noise level in the data). Then, an adaptive segmentation approach is introduced for the aggregation of laser scanning points into individual planar and linear/cylindrical features while taking into account the internal characteristics of the laser scanning point cloud. Next, a new quality control procedure is presented that can identify different problems which might affect the quality of the segmentation outcome and then propose possible actions for resolving these problems. A sub-surface region-growing algorithm is also introduced to reconstruct the partially occluded segmented surfaces and achieve more complete segmentation results. A new segmentation-based ground/non-ground classification approach is presented as well to discriminate whether the points in a laser scanning point cloud belong to the ground surface or to non-ground objects. Using the proposed processing procedure, however, the segmented and classified surfaces cannot be effectively interpreted due to the lack of descriptive (spectral) information. Therefore, a new region-based texturing procedure is proposed to incorporate the spectral information from overlapping images into the laser scanning-derived surfaces. This texturing procedure is implemented while investigating the visibility of the laser scanning-derived surfaces in the existing images using a newly-developed visibility analysis/occlusion detection technique. The feasibility of the proposed processing and texturing approaches is verified using simulated and real laser scanning datasets and overlapping images. In conclusion, this dissertation provides a new framework for the adaptive processing and texturing of laser scanning data using the overlapping images. The quality and interpretability of the final product of this framework is ensured by taking advantage of the synergic properties that arise from relating laser scanning and imagery data.
Description
Keywords
Remote Sensing, Engineering
Citation
Lari, Z. (2014). Adaptive Processing of Laser Scanning Data and Texturing of the Segmentation Outcome (Doctoral thesis, University of Calgary, Calgary, Canada). Retrieved from https://prism.ucalgary.ca. doi:10.11575/PRISM/24701