Appearance-only SLAM at large scale with FAB-MAP 2.0

By Mark Joseph Cummins, Paul M. Newman. Int. J. Robotics Res., 2011

698 citations recorded by the source corpus.

We describe a new formulation of appearance-only SLAM suitable for very large scale place recognition. The system navigates in the space of appearance, assigning each new observation to either a new or a previously visited location, without reference to metric position. The system is demonstrated performing reliable online appearance mapping and loop-closure detection over a 1000 km trajectory, with mean filter update times of 14 ms. The scalability of the system is achieved by defining a sparse approximation to the FAB-MAP model suitable for implementation using an inverted index. Our formulation of the problem is fully probabilistic and naturally incorporates robustness against perceptual aliasing. We also demonstrate that the approach substantially outperforms the standard term-frequency inverse-document-frequency (tf-idf) ranking measure. The 1000 km data set comprising almost a terabyte of omni-directional and stereo imagery is available for use, and we hope that it will serve as a benchmark for future systems.

Subfields: Aerospace Engineering, Computer Vision and Pattern Recognition, Ocean Engineering

View Paper: https://www.semanticscholar.org/paper/e69fb8be33af7ab7d4d2eda39c4ad8f892ac97ac

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Aerospace Engineering

In Computer Vision and Pattern Recognition

In Ocean Engineering

Citations in the corpus

ORB-SLAM: A Versatile and Accurate Monocular SLAM System (2015, methodology): “While our current implementation makes use of ORB, the techniques proposed are not restricted to these features.”
High-fidelity simulation for evaluating robotic vision performance (2016, methodology): “The most common and efficient approaches for recognizing a place revisited by a robot make use of Bag of Words approach with features, for example, SURF in FAB-MAP [4] and most recently ORB in ORB-SLAM [15] etc.”
Illumination Invariant Imaging : Applications in Robust Vision-based Localisation , Mapping and Classification for Autonomous Vehicles (2014, result): “Visual navigation systems that build upon robust features such as SIFT [2] have produced impressive results in recent years [3]; however, as demonstrated in”
Multiple‐Robot Simultaneous Localization and Mapping: A Review (2016, methodology): “Appearance-based SLAM is an effective method for the loop closure problem (Koenig, Kessler, & Gross, 2008; Cummins & Newman, 2011).”
3-D Mapping With an RGB-D Camera (2014, background): “Recognition of images in large sets of images has been investigated mostly in the context of image retrieval systems [38] but for large scale SLAM as well [39].”
Benchmarking 6DOF Outdoor Visual Localization in Changing Conditions (2018, methodology): “We used the mean field approximation for the new place likelihood (as additional training images were not available for the sampled approach used in [21]) and the fast lookup-table implementation in [27] to perform image retrieval for each of the query locations.”
Bags of Binary Words for Fast Place Recognition in Image Sequences (2012, methodology): “0 [13] needs423ms for extracting SURF,60ms for conversion into bag of words,10ms for retrieving matching candidates against 25K images, and120ms (worst case) for RANSAC geometric verification.”
Vision-based state estimation for autonomous rotorcraft MAVs in complex environments (2013, methodology): “To this end, we are currently investigating appearance-based loop closure detection methods such as those in [13, 14].”
Place Recognition with ConvNet Landmarks: Viewpoint-Robust, Condition-Robust, Training-Free (2015, background): “Other geometry-within-the-image techniques like geometric verification [8, 28] may also improve performance.”
Online global loop closure detection for large-scale multi-session graph-based SLAM (2014, background): “Global loop closure detection approaches, by being independent of the robot’s estimated position [7], can intrinsically solve the problem of determining when a robot comes back to a previous map using a different referential [8].”
Simultaneous Localization and Mapping: A Survey of Current Trends in Autonomous Driving (2017, neutral): “0 [137]) and PIRF-Nav (Position Invariant Robust Feature Navigation) [138] are two”
Scan Context++: Structural Place Recognition Robust to Rotation and Lateral Variations in Urban Environments (2021, methodology): “Describing a place with a set of compact representations has been tackled in depth within the computer vision and robotics community, yielding many state-of-the-art visual place recognition meth-ods [4, 5, 6, 7].”
Large-scale direct SLAM with stereo cameras (2015, methodology): “For very large maps, additional loop-closures can be found by exploiting appearance-based image-retrieval techniques like FAB-MAP [17].”