The visual compass: Performance and limitations of an appearance‐based method

Abstract

In this article we present an algorithm to estimate the orientation of a robot relative to an orientation specified at the beginning of the process. This is done by computing the rotation of the robot between successive panoramic images, grabbed on the robot while it moves, using a subsymbolic method to match the images. The context of the work is simultaneous localization and mapping ͑SLAM͒ in unstructured and unmodified environments. As such, very few assumptions are made about the environment and the robot's displacement. The algorithm's performance depends on the value of a number of parameters being determined to provide overall good performance of the system. The performance is evaluated in different situations ͑trajectories and environments͒ with the same parameters and the results show that the method performs adequately for its intended use. In particular, the error is shown to be drifting slowly, in fact much slower than unprocessed inertial sensors, thus only requiring infrequent realignment, for example, when relocalizing in a topological map. Limitations of the proposed methods are also shown and discussed. Figure 23. CARPARK: error between the magnetic and visual orientations.

An appearance-based visual compass for mobile robots

“…As the visual compass described in [10,11] only compares the latest two images, and therefore small errors in estimation add up over time, making their results difficult to compare.…”

“…More closely related to our work, an appearance-based visual compass is developed in [10,11]. The robot compares each captured image with the previous one, and computes the Manhattan distance between the two images.…”

Illumination Tolerance for Visual Navigation with the Holistic Min-Warping Method

“…• Several visual compass methods have been developed-most of them in the DID framework-which extract the azimuthal orientation difference from two panoramic images [25,26,[35][36][37][38][43][44][45][46][47]. Two panoramic images are rotated relative to each other in azimuthal direction and a pixel-wise distance is computed.…”

“…Vertical and horizontal differences are simply stacked in a joint vector and compared by the difference measure. An example closely related to min-warping is the visual compass [25,44]. Here entire images-in this case sets of all horizontal and vertical differences-are compared by the difference measure for varying relative azimuthal orientation between the two images.…”

Combined visual odometry and visual compass for off-road mobile robots localization

“…20 A slightly different approach is the template matching method. [22][23][24] It avoids the problem of finding and tracking features, and instead it looks at the change in the appearance of the world (images). For that purpose, it takes a template or patch from an image and tries to match it in the previous image.…”