Performance of Global-Appearance Descriptors in Map Building and Localization Using Omnidirectional Vision

Abstract

Map building and localization are two crucial abilities that autonomous robots must develop. Vision sensors have become a widespread option to solve these problems. When using this kind of sensors, the robot must extract the necessary information from the scenes to build a representation of the environment where it has to move and to estimate its position and orientation with robustness. The techniques based on the global appearance of the scenes constitute one of the possible approaches to extract this information. They consist in representing each scene using only one descriptor which gathers global information from the scene. These techniques present some advantages comparing to other classical descriptors, based on the extraction of local features. However, it is important a good configuration of the parameters to reach a compromise between computational cost and accuracy. In this paper we make an exhaustive comparison among some global appearance descriptors to solve the mapping and localization problem. With this aim, we make use of several image sets captured in indoor environments under realistic working conditions. The datasets have been collected using an omnidirectional vision sensor mounted on the robot.

Development of Height Indicators using Omnidirectional Images and Global Appearance Descriptors

“…The altitude estimators we propose go beyond the classical topological notion of connectivity and introduce the concepts of closeness and farness, thus some altitude estimators are proposed that estimate the relative height of a robot except for a scale factor. The results of this paper jointly with the developments presented in [26,34] prove the usefulness of the global appearance descriptors to estimate the position and orientation of the robot in the ground plane and its altitude with respect to this plane, in a straightforward way. This supposes a step ahead towards the definition of global appearance descriptors that permit building models of the environment and localization when the robot moves with 6 DOF.…”

“…Descriptors based on the global appearance of images captured by a catadioptric vision system have proved a good performance both in position and orientation estimation, when the movement of the robot is restricted to the floor plane, as Gaspar et al [36] and Payá et al [26] show. These methods extract the most relevant information from each image and reduce the amount of memory necessary to store the visual information working with the image as a whole, i.e., avoiding the extraction of landmarks or local features.…”

“…Visual global-appearance approaches may be used to create such models, since no metric information can be extracted from global descriptors. They lead to simpler models where the localization process is more straightforward, based mainly on the pairwise comparison between image descriptors [26].…”

Visual Place Recognition for Autonomous Mobile Robots

“…We used our own grid databases [46] (http://www.ti.uni-bielefeld.de/html/people/dfleer/ fisheyedb.html), the Quorum V database [48] (http://arvc.umh.es/db/images/quorumv/), and a map built by our cleaning robot in the CITEC (Cognitive Interaction Technology (Center of Excellence) at Bielefeld University) research apartment (https://www.cit-ec.de/en/central-lab-facilities/ research-apartment-0). We also included a small outdoor database as a preliminary test for outdoor use of the presented methods.…”

“…The Quorum V database [48] was captured in a corridor and five adjacent rooms ( Figure 10) with a camera looking at a hyperbolic mirror. A total of 872 images were captured on a regularly spaced grid with a resolution of 0.4 m. The camera images have a resolution of 402 × 402 pixels.…”