Localization and Matching Using the Planar Trifocal Tensor With Bearing-Only Data

Abstract

This paper addresses the robot and landmark localization problem from bearing-only data in three views, simultaneously to the robust association of this data. The localization algorithm is based on the 1D trifocal tensor, which relates linearly the observed data and the robot localization parameters. The aim of this work is to bring this useful geometric construction from computer vision closer to robotic applications. One contribution is the evaluation of two linear approaches of estimating the 1D tensor: the commonly used approach that needs seven bearing-only correspondences and another one which uses only five correspondences plus two calibration constraints. The results in this paper show that the inclusion of these constraints provides a simpler and faster solution and better estimation of robot and landmark locations in the presence of noise. Moreover, a new method that makes use of scene planes and requires only four correspondences is presented. This proposal improves the performance of the two previously mentioned methods in typical man-made scenarios with dominant planes, while it gives similar results in other cases. The three methods are evaluated with simulation tests as well as with experiments that perform automatic real data matching in conventional and omnidirectional images. The results show sufficient accuracy and stability to be used in robotic tasks such as navigation, global localization or initialization of SLAM algorithms.

A novel 1D trifocal tensor-based control for differential-drive robots

“…On the other hand, the second part of the table shows how the precision is preserved even if the initial orientation is fixed to φ 1 = 0 in the controller for all the cases. We can assert We can compute the trifocal tensor from omnidirectional cameras [9], so, we can assume that we have no restriction in the field of view, and consequently, the large rotation that the robot performs in the outer curve motion case can be carried out. Another option to keep the target in the field of view is to perform a initial rotation in order to reach the condition t x1 = t x2 = 0, and then, to execute the rectilinear motion to the target.…”

“…According to the existence conditions of sliding modes, the bounded controller (16) is able to locally stabilize the system (9). Its attraction region is bigger as long as the control gains M and N are higher.…”

“…This tensor completely describes the relative geometry of three views and it is independent of the observed scene [7]. The effectiveness of applying the trifocal tensor to recover location information has been proved in [8] and [9]. The first work uses conventional cameras and artificial landmarks on a plane while the second one uses both conventional and omnidirectional cameras.…”

Feature-based map merging with dynamic consensus on information increments

“…The robots usually start their operation at unknown poses and, before merging their maps, they must agree on a common reference frame. This common frame needs to be computed at least once, and usually only requires the robots to know the relative pose of its nearby teammates, see e.g., [24]- [26] where different methods for computing robot-to-robot measurements are presented. There exist several distributed algorithms that combine these measurements to produce the common frame, e.g., [27]- [30] and references therein.…”

Scale space for central catadioptric systems: Towards a generic camera feature extractor

“…SIFT [15] has become the most used feature extraction approach. It has also been used directly in omnidirectional images [10], although it is not designed to work on them. This SIFT approach has inspired different works trying to replicate its good results on different imagery systems, in particular on wide-angle cameras.…”