Information

Abstract

This diploma thesis introduces methods for external sorting and fast k nearest neighbor searching for very large point clouds. Very large point clouds are datasets that can not be processed in main memory. This leads to certain requirements for any used reconstruction technique. The most important ones are out-of-core memory management and sequential data handling algorithms. The paper “Stream- Processing Points” by Renato Pajarola shows a way to design a framework which allows to process a subset of very large point clouds. A subset is defined as a spatially continuous region holding a predefined number of points. This diploma thesis is based on the aforementioned paper and improves on the stream processing pipeline presented therein. The proposed algorithm for searching the k nearest neighbors has a complexity of O(N  log(M)), where N are all points in the point cloud and M are all points in main memory, which is similar to current state of the art algorithms for in-core processed data sets.

Additional Files and Images

Additional images and videos

image: Normals estimated in the Domitilla Catacomb model with approximately 2 billion points. image: Normals estimated in the Domitilla Catacomb model with approximately 2 billion points.

Additional files

thesis: The thesis thesis: The thesis

Weblinks

No further information available.

BibTeX

@mastersthesis{marek-2011-normalest,
  title =      "Normal Estimation of Very Large Point Clouds",
  author =     "Stefan Marek",
  year =       "2011",
  abstract =   "This diploma thesis introduces methods for external sorting
               and fast k nearest neighbor searching for very large point
               clouds. Very large point clouds are datasets that can not be
               processed in main memory. This leads to certain requirements
               for any used reconstruction technique. The most important
               ones are out-of-core memory management and sequential data
               handling algorithms. The paper “Stream- Processing
               Points” by Renato Pajarola shows a way to design a
               framework which allows to process a subset of very large
               point clouds. A subset is defined as a spatially continuous
               region holding a predefined number of points. This diploma
               thesis is based on the aforementioned paper and improves on
               the stream processing pipeline presented therein. The
               proposed algorithm for searching the k nearest neighbors has
               a complexity of O(N  log(M)), where N are all points in the
               point cloud and M are all points in main memory, which is
               similar to current state of the art algorithms for in-core
               processed data sets.",
  month =      feb,
  address =    "Favoritenstrasse 9-11/E193-02, A-1040 Vienna, Austria",
  school =     "Institute of Computer Graphics and Algorithms, Vienna
               University of Technology ",
  keywords =   "Out-of-Core algorithms, Point-based rendering, Normal
               estimation",
  URL =        "https://www.cg.tuwien.ac.at/research/publications/2011/marek-2011-normalest/",
}