Bags of Binary Words for Fast Place Recognition in Image Sequences

By Dorian Gálvez-López, Juan D. Tardós. IEEE Transactions on robotics, 2012

1,854 citations recorded by the source corpus.

Subfields: Aerospace Engineering, Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/978e7f8557774fa1ebd228e181c4c154449423c9

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

Citations in the corpus

Intensity Scan Context: Coding Intensity and Geometry Relations for Loop Closure Detection (2020, methodology): “In particular, we use DBoW2 [10] which is implemented in ORB SLAM [11].”
Kimera-Multi: Robust, Distributed, Dense Metric-Semantic SLAM for Multi-Robot Systems (2021, methodology): “, bag-of-words vectors [36, 37] and learned full-image descriptors [38]) and local visual features (e.”
Benchmarking 6DOF Outdoor Visual Localization in Changing Conditions (2018, background): “They are often used for place recognition [1, 17, 41, 55, 66, 69] and loop-closure detection [20,25,48].”
Visual-Inertial Navigation: A Concise Review (2019, background): “In particular, whether or not performing loop closures in VINS either via mapping [83, 101, 102] and/or place recognition [103, 104, 105, 106, 107, 108] is one of the key differences between VIO and SLAM.”
Hydra: A Real-time Spatial Perception System for 3D Scene Graph Construction and Optimization (2022, positive): “At the lowest level, our hierarchical descriptors include standard DBoW2 appearance descriptors [19].”
BAD SLAM: Bundle Adjusted Direct RGB-D SLAM (2019, methodology): “We use a standard bag-of-words approach [19] based on binary features [4] to identify the keyframe m most similar to the latest keyframe k.”
LVI-SAM: Tightly-coupled Lidar-Visual-Inertial Odometry via Smoothing and Mapping (2021, methodology): “The image timestamps of loop closure candidates returned by DBoW2 are sent to the LIS for further validation.”
MulRan: Multimodal Range Dataset for Urban Place Recognition (2020, methodology): “To the best of our knowledge, however, there is no common agreement on radar place recognition methods or widely used methods, such as DBoW [12], in the visual domain.”
Past, Present, and Future of Simultaneous Localization and Mapping: Toward the Robust-Perception Age (2016, positive): “e resulting loop closure is topological in nature. For metric relocalization (i.e., estimating the relative pose with respect to the previously built map), feature-based approaches are still the norm [121] but have not been extended to such circumstances. While vision has become the sensor of choice”
Visual SLAM and Structure from Motion in Dynamic Environments (2018, methodology): “…like ORB-SLAM [113] integrate hardware and algorithmic advancement in the past decade by including parallel computing, ORB features [131], statistical model selection [155], loop closures based on bag-of-words place recognition [26, 41], local bundle adjustment [111], and graph optimization [81].”
Computer Vision for Autonomous Vehicles: Problems, Datasets and State-of-the-Art (2017, methodology): “They combine methods from loop detection (Gálvez-López & Tardós (2012)), loop closing (Strasdat et al. (2010, 2011)) and pose graph optimization (Kümmerle et al. (2011)) into one system.”
Scan Context++: Structural Place Recognition Robust to Rotation and Lateral Variations in Urban Environments (2021, methodology): “Using these local descriptors, Bayesian inference [4] or bag-of-words vocabulary tree [7] was applied for place recognition.”
RTAB‐Map as an open‐source lidar and visual simultaneous localization and mapping library for large‐scale and long‐term online operation (2018, methodology): “For ORB ‐ SLAM2 and S ‐ PTAM, when a loop closure is detected using DBoW2 (Gálvez ‐ López & Tardós, 2012), the map is optimized using bundle adjustment.”
PL-SLAM: A Stereo SLAM System Through the Combination of Points and Line Segments (2017, methodology): “1) Visual Place Recognition: Specifically, we have employed the method presented in [21], which was initially developed for BRIEF binary descriptors, and subsequently adapted to ORB keypoints.”
Get Out of My Lab: Large-scale, Real-Time Visual-Inertial Localization (2015, methodology): “In order to match the features against the 3D landmarks, approaches from robotics can be used to first retrieve relevant database images [8, 16, 48].”
ORB-SLAM: A Versatile and Accurate Monocular SLAM System (2015, methodology): “In this study, we build on the main ideas of PTAM, the place recognition work of Gálvez-López and Tardós [5], the scaleaware loop closing of Strasdat et al.”
Kimera: From SLAM to spatial perception with 3D dynamic scene graphs (2021, methodology): “The loop-closure detection (LCD) relies on the DBoW2 library (Gálvez-López and Tardós, 2012) and uses a bag-of-words representation with ORB descriptors to quickly detect putative loop closures.”
A constant-time SLAM back-end in the continuum between global mapping and submapping: application to visual stereo SLAM (2016, methodology): “Data association is aided by the management of an image database based on a binary bag of words (Galvez-Lopez and Tardos 2012).”
ORB-SLAM3: An Accurate Open-Source Library for Visual, Visual–Inertial, and Multimap SLAM (2020, methodology): “Using only the ﬁrst candidate, the raw DBoW2 queries achieve precision and recall in the order of 50-80% [9].”
ORB-SLAM2: An Open-Source SLAM System for Monocular, Stereo, and RGB-D Cameras (2016, positive): “The system has embedded a Place Recognition module based on DBoW2 [16] for relocalization, in case of tracking failure (e.g. an occlusion) or for reinitialization in an already mapped scene, and for loop detection.”