Please use this identifier to cite or link to this item:
|Title:||Detecting Ambiguity in Localization Problems using Depth Sensors|
|Authors:||TADDEI PIERLUIGI; SANCHEZ BELENGUER CARLOS; RODRIGUEZ LOPEZ ANTONIO; CERIANI SIMONE; SEQUEIRA Vitor|
|Citation:||2014 International Conference on 3D Vision vol. 2 p. 129-136|
|Publisher:||IEEE Computer Society|
|Type:||Articles in periodicals and books|
|Abstract:||We describe a method to identify ambiguous poses during tracking and localization based on depth sensors. In particular we distinguish between ambiguities related to specific acquisitions (track ambiguities) that hinder a good registration of the current pose with previous acquisitions and ambiguities related to repetitive elements observed in a particular (known) environment visited (map ambiguities). Both types of ambiguities can be exploited to scale tracking and localization problems to very large environments. Detecting track ambiguities permits using accurate point clouds as reference for registration in online tracking. Moreover, it allows triggering a relocalization only if the current observation is suitable for a good registration. Identifying map ambiguities in known environments, instead, can help classification methods to generate compact predictors and to reduce the class space to the interesting areas. We propose a two level classifier that firstly labels an input observation as ambiguous or not, and then provides a prediction of candidate poses from the subset of unambiguous ones. We show that by identifying such poses, real time SLAM systems can avoid spending excessive processing time in the real-time relocalization step. Furthermore, it permits the generation of more compact, highly-discriminative relocalization classifiers. We combine these proposals and use them as a proof of concept. Our preliminary results on real datasets justify the integration of the present proposal in SLAM or Ego-Motion pipelines of such concepts.|
|JRC Directorate:||Nuclear Safety and Security|
Files in This Item:
There are no files associated with this item.
Items in repository are protected by copyright, with all rights reserved, unless otherwise indicated.