Location graphs for visual place recognition

Abstract

Abstract-With the growing demand for deployment of robots in real scenarios, robustness in the perception capabilities for navigation lies at the forefront of research interest, as this forms the backbone of robotic autonomy. Existing place recognition approaches traditionally follow the feature-based bag-of-words paradigm in order to cut down on the richness of information in images. As structural information is typically ignored, such methods suffer from perceptual aliasing and reduced recall, due to the ambiguity of observations. In a bid to boost the robustness of appearance-based place recognition, we consider the world as a continuous constellation of visual words, while keeping track of their covisibility in a graph structure. Locations are queried based on their appearance, and modelled by their corresponding cluster of landmarks from the global covisibility graph, which retains important relational information about landmarks. Complexity is reduced by comparing locations by their graphs of visual words in a simplified manner. Test results show increased recall performance and robustness to noisy observations, compared to state-of-the-art methods.

Viewpoint-Tolerant Place Recognition Combining 2D and 3D Information for UAV Navigation

“…Finally, while most of the place recognition systems ignores the underlying structure and geometry between features when comparing features sets, a handful of works have investigated how to incorporate some geometric information in their location models, such as in [21], where locations are represented by both visual landmarks and a distribution of the distances between them in 3D coming from range-finders or stereo cameras. Instead of relying on additional sensors to obtain 3D landmark positions, in [22] landmarks are tracked between successive images using a single camera, recording the binary covisibility between landmarks in a graph-based map of the world. In the general case, the graph matching problem in undirected graphs is an NP-hard problem.…”

Appearance-based landmark selection for efficient long-term visual localization

“…In [9] co-occurrence information is used to infer which types of features are often seen together, as this helps distinguishing places. Furthermore, in [10], [11], and [12] places are described and identified by constellations of visible landmarks or features grouped based on co-observability, therefore incorporating pseudo-geometric information in their representation. Similarly, the works of [13] and [14] rely on landmark co-occurence statistics for prioritizing relevant landmarks or environments for improved place-recognition efficiency.…”

Geometrically Mappable Image Features

“…A few works have addressed the problem of seasonal or day-night variations either by using 3D point clouds [16] or by domain transfer [13]. Others have proposed better or faster matching [22], [23], facilitating image retrieval. In the following sections we describe our method of improved feature learning based on the proposed geometric loss and show its implications for localization related tasks.…”