Exploiting Passthrough Information for Multi-view Object Reconstruction with Sparse and Noisy Laser Data - Robotics Institute Carnegie Mellon University

Exploiting Passthrough Information for Multi-view Object Reconstruction with Sparse and Noisy Laser Data

Martin Herrmann and Siddhartha Srinivasa
Tech. Report, CMU-RI-TR-10-07, Robotics Institute, Carnegie Mellon University, February, 2010

Abstract

We describe a probabilistic model for utilizing passthrough information for producing 3D geometry models from a rotating laser scanner. Our method is fast, performs particularly well with relatively sparse data, is robust to noise of the depth data and naturally handles grazing points. We demonstrate our results on the HERB platform where the robot automatically builds a watertight 3D model of an object it is handed.

BibTeX

@techreport{Herrmann-2010-10397,
author = {Martin Herrmann and Siddhartha Srinivasa},
title = {Exploiting Passthrough Information for Multi-view Object Reconstruction with Sparse and Noisy Laser Data},
year = {2010},
month = {February},
institute = {Carnegie Mellon University},
address = {Pittsburgh, PA},
number = {CMU-RI-TR-10-07},
}