SeqSLAM: Visual route-based navigation for sunny summer days and stormy winter nights
The DOI matches its Crossref record; 20 of the 22 references it cites were found in a registry or an index.
Abstract
Learning and then recognizing a route, whether travelled during the day or at night, in clear or inclement weather, and in summer or winter is a challenging task for state of the art algorithms in computer vision and robotics. In this paper, we present a new approach to visual navigation under changing conditions dubbed SeqSLAM. Instead of calculating the single location most likely given a current image, our approach calculates the best candidate matching location within every local navigation sequence. Localization is then achieved by recognizing coherent sequences of these “local best matches”. This approach removes the need for global matching performance by the vision front-end - instead it must only pick the best match within any short sequence of images. The approach is applicable over environment changes that render traditional feature-based techniques ineffective. Using two car-mounted camera datasets we demonstrate the effectiveness of the algorithm and compare it to one of the most successful feature-based SLAM algorithms, FAB-MAP. The perceptual change in the datasets is extreme; repeated traverses through environments during the day and then in the middle of the night, at times separated by months or years and in opposite seasons, and in clear weather and extremely heavy rain. While the feature-based method fails, the sequence-based algorithm is able to match trajectory segments at 100% precision with recall rates of up to 60%.
Overview
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
In the citing paper
“This has led to develop new methods that explicitly account for such variations by matching sequences [173], gathering different visual appearances into a unified representation [48], or using spatial as well as appearance information [106].”
From Past, Present, and Future of Simultaneous Localization and Mapping: Toward the Robust-Perception Age (2016) · cites this paper as [173]
Citations
Cites
Object recognition from local scale-invariant features
Lowe, D. G. (1999) · Proceedings of the Seventh IEEE International Conference on Computer Vision · doi:10.1109/iccv.1999.790410 · Cited by 16,472 · cited 3×
“Many place recognition techniques rely on feature-finding algorithms such as SIFT [9] and SURF [10] which, despite their impressive rotation and scale invariant properties, are inherently unsuitable when dealing with extreme perceptual M.J. Milford and G.F. Wyeth are with the School of Electrical Engineering and Compu…”
MethodSURF: Speeded Up Robust Features
Bay, H., Tuytelaars, T., & Van Gool, L. (2006) · Lecture Notes in Computer Science · doi:10.1007/11744023_32 · Cited by 14,955
MonoSLAM: Real-time single camera SLAM
Davison, A. J., Reid, I. D., Molton, N. D., & Stasse, O. (2007) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2007.1049 · Cited by 3,959 · cited 1×
“isual recognition of places and place sequences in the field of robotic navigation has advanced rapidly in recent years, with several techniques [1-5] now being competitive with laser range-based techniques.”
BackgroundA solution to the simultaneous localization and map building (SLAM) problem
Dissanayake, M. W. M. G., Newman, P., Clark, S., Durrant-Whyte, H. F., & Csorba, M. (2001) · IEEE Transactions on Robotics and Automation · doi:10.1109/70.938381 · Cited by 2,661
Matching Local Self-Similarities across Images and Videos
Shechtman, E., & Irani, M. (2007) · 2007 IEEE Conference on Computer Vision and Pattern Recognition · doi:10.1109/cvpr.2007.383198 · Cited by 1,076 · cited 1×
“One promising step [15] towards reducing the reliance on consistent feature detection uses the self-similarity of objects within a frame or video sequence to robustly recognize objects under varying conditions.”
BackgroundIncremental mapping of large cyclic environments
Gutmann, J.-S., & Konolige, K. (2003) · Proceedings 1999 IEEE International Symposium on Computational Intelligence in Robotics and Automation. CIRA'99 (Cat. No.99EX375) · doi:10.1109/cira.1999.810068 · Cited by 568 · cited 1×
“There have been attempts to increase the robustness of visual techniques to environment change by performing sensor fusion with lasers at the level of landmarks [13] or dense depth maps [14].”
BackgroundFrameSLAM: From Bundle Adjustment to Real-Time Visual Mapping
Konolige, K., & Agrawal, M. (2008) · IEEE Transactions on Robotics · doi:10.1109/tro.2008.2004832 · Cited by 532
RatSLAM: a hippocampal model for simultaneous localization and mapping
Milford, M. J., Wyeth, G. F., & Prasser, D. (2004) · IEEE International Conference on Robotics and Automation, 2004. Proceedings. ICRA '04. 2004 · doi:10.1109/robot.2004.1307183 · Cited by 399 · cited 6×
“The approach taken in this work is partly inspired by the author's previous work in vision-based mapping in RatSLAM [16] [17] [18].”
MethodInfluentialSelf-citationMapping a Suburb With a Single Camera Using a Biologically Inspired SLAM System
Milford, M. J., & Wyeth, G. F. (2008) · IEEE Transactions on Robotics · doi:10.1109/tro.2008.2004520 · Cited by 378 · cited 7×
“The approach taken in this work is partly inspired by the author's previous work in vision-based mapping in RatSLAM [16] [17] [18].”
BackgroundMethodInfluentialSelf-citationDynamic Maps for Long-Term Operation of Mobile Service Robots
Biber, P., & Duckett, T. (2005) · Robotics: Science and Systems I · doi:10.15607/rss.2005.i.003 · Cited by 334 · cited 1×
“The problem of visual place recognition across time has received far less attention, with relatively few attempts to address the problem [6, 7], including high dynamic range approaches [8].”
Background
Cited by
Items 101–200 of 987
Satellite image-based localization via learned embeddings
Kim, D.-K., & Walter, M. R. (2017) · Proceedings - IEEE International Conference on Robotics and Automation/Proceedings · doi:10.1109/icra.2017.7989239 · Cited by 95
BTC: A Binary and Triangle Combined Descriptor for 3-D Place Recognition
Yuan, C., Lin, J., Liu, Z., Wei, H., Hong, X., & Zhang, F. (2024) · IEEE Transactions on Robotics · doi:10.1109/tro.2024.3353076 · Cited by 98
Visual Place Recognition: A Tutorial [Tutorial]
Schubert, S., Neubert, P., Garg, S., Milford, M. J., & Fischer, T. (2023) · IEEE Robotics & Automation Magazine · doi:10.1109/mra.2023.3310859 · Cited by 102
“…[32]), time of day with lighting changes and moving shadows (e.g., StLucia Various Times of the Day [46]) or day versus night (e.g., Alderley [17]), weather that is sunny, cloudy, overcast, rainy, foggy, or snowy (e.g., Oxford RobotCar [7]), seasons like spring, summer, fall and winter or…”
BackgroundMethodInfluentialSelf-citationOmnidirectional CNN for Visual Place Recognition and Navigation
Wang, T.-H. J., Huang, H.-J., Lin, J.-T., Hu, C.-W., Zeng, K.-H., & Sun, M. (2018) · Proceedings - IEEE International Conference on Robotics and Automation/Proceedings · doi:10.1109/icra.2018.8463173 · Cited by 94
Visual place recognition using HMM sequence matching
Hansen, P., & Browning, B. (2014) · IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) · doi:10.1109/iros.2014.6943207 · Cited by 89
“Recently [8], [9] presented sequence SLAM (seqSLAM) that achieves significant performance improvements over FAB-MAP under extreme lighting and atmospheric variations [9].”
BackgroundBridging the appearance gap: Multi-experience localization for long-term visual teach and repeat
Paton, M., MacTavish, K., Warren, M., & Barfoot, T. D. (2016) · IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) · doi:10.1109/iros.2016.7759303 · Cited by 78
“Others provide topological localization through the alignment of image sequences, providing localization across appearance change as drastic as night vs. day [9], [10], [11].”
BackgroundSimultaneous Localization And Mapping: Present, Future, and the Robust-Perception Age.Preprint
Cadena, C., Carlone, L., Carrillo, H., Latif, Y., Scaramuzza, D., Neira, J., Reid, I. D., & Leonard, J. J. (2016) · arXiv (Cornell University) · Cited by 90
Radar Odometry Combining Probabilistic Estimation and Unsupervised Feature Learning
Burnett, K., Yoon, D. J., Schoellig, A. P., & Barfoot, T. (2021) · Robotics: Science and Systems · doi:10.15607/rss.2021.xvii.029 · Cited by 89
The gist of maps - summarizing experience for lifelong localization
Dymczyk, M., Lynen, S., Cieslewski, T., Bosse, M. C., Siegwart, R., & Furgale, P. (2015) · Proceedings - IEEE International Conference on Robotics and Automation/Proceedings · doi:10.1109/icra.2015.7139575 · Cited by 86
Robust Visual Localization in Dynamic Environments Based on Sparse Motion Removal
Cheng, J., Wang, C., & Meng, M. Q. ‐. (2019) · IEEE Transactions on Automation Science and Engineering · doi:10.1109/tase.2019.2940543 · Cited by 76
“V ISUAL localization has been well studied in recent decades as a basic capability [1] in robotics, and many state-of-the-art algorithms [2]–[4] are proposed with rather satisfactory performance.”
BackgroundAddressing appearance change in outdoor robotics with adversarial domain adaptation
Wulfmeier, M., Bewley, A., & Posner, I. (2017) · 2017 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) · doi:10.1109/iros.2017.8205961 · Cited by 72
“Other works in localisation take a multiview approach by either exploiting temporal structure [18] or by accumulating multiple experiences covering different appearances [2] to avoid the issue of modelling appearance change.”
BackgroundScalable Place Recognition Under Appearance Change for Autonomous Driving
Doan, A.-D., Latif, Y., Chin, T.-J., Liu, Y., Do, T.-T., & Reid, I. D. (2019) · IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv.2019.00941 · Cited by 88
“To address appearance change, SeqSLAM [28] proposed to match statistics of the current image sequence to a sequence of images seen in the past, exploiting the temporal relationship.”
Appearance-invariant place recognition by discriminatively training a convolutional neural network
López-Antequera, M., Gomez-Ojeda, R., Petkov, N., & González-Jiménez, J. (2017) · Pattern Recognition Letters · doi:10.1016/j.patrec.2017.04.017 · Cited by 80
Loop closure detection for visual SLAM systems using convolutional neural network
Zhang, X., Su, Y. L., & Zhu, X. (2017) · International Conference on Automation and Computing (ICAC) · doi:10.23919/iconac.2017.8082072 · Cited by 79
“In addition, those image representations describe the local appearance of individual patches, limiting their descriptive power with respect to whole image methods [19].”
BackgroundKidnapped Radar: Topological Radar Localisation using Rotationally-Invariant Metric Learning
Săftescu, Ş., Gadd, M., Martini, D. D., Barnes, D., & Newman, P. M. (2020) · Proceedings - IEEE International Conference on Robotics and Automation/Proceedings · doi:10.1109/icra40945.2020.9196682 · Cited by 81
“A brief history of the community’s progress in this regard includes: probabilistic models around bag-of-words image representations [8], sequence-based approaches [9], and more recently by extracting features from the responses of Convolutional Neural Network (CNN) layers and subsequent use of these features for image…”
MethodMade to measure: Bespoke landmarks for 24-hour, all-weather localisation with a camera
Linegar, C., Churchill, W., & Newman, P. (2016) · 2016 IEEE International Conference on Robotics and Automation (ICRA) · doi:10.1109/icra.2016.7487208 · Cited by 76
“SeqSLAM [1] and FAB-MAP [2] are examples of place recognition systems – they output the image (or place) to which the robot’s live image is most similar, however there is no information about the robot’s 6-DOF pose in the map.”
MethodLynen, S., Bosse, M., Furgale, P., & Siegwart, R. (2014) · 2014 2nd International Conference on 3D Vision · doi:10.1109/3dv.2014.36 · Cited by 73
Photometric Bundle Adjustment for Vision-Based SLAM
Alismail, H., Browning, B., & Lucey, S. (2017) · Lecture notes in computer science · doi:10.1007/978-3-319-54190-7_20 · Cited by 84
“The ability to perform BA without the need for precise correspondences is attractive because it can enable VSLAM applications where corner extraction is unreliable [19], as well as additional modeling capabilities that extend beyond geometric primitives [20,21].”
BackgroundGarg, S., Sünderhauf, N., & Milford, M. J. (2018) · Proceedings - IEEE International Conference on Robotics and Automation/Proceedings · doi:10.1109/icra.2018.8461051 · Cited by 80
Self-citationLiu, Z., Suo, C., Zhou, S., Xu, F., Wei, H., Chen, W., Wang, H., Liang, X., & Liu, Y. (2019) · IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) · doi:10.1109/iros40897.2019.8967875 · Cited by 76
Lazy Data Association For Image Sequences Matching Under Substantial Appearance Changes
Vysotska, O., & Stachniss, C. (2016) · IEEE Robotics and Automation Letters · doi:10.1109/lra.2015.2512936 · Cited by 68
“Additionally, we outperform SeqSLAM and FLANN.”
BackgroundMethodInfluentialSeqOT: A Spatial–Temporal Transformer Network for Place Recognition Using Sequential LiDAR Data
Ma, J., Chen, X., Xu, J., & Xiong, G. (2022) · IEEE Transactions on Industrial Electronics · doi:10.1109/tie.2022.3229385 · Cited by 69
“More recently, Garg and Milford [9] have proposed SeqNet, which combines the convolution network with a simplified version of SeqSLAM [7] to better use the relations between sequences of descriptors.”
BackgroundMethodInfluentialEvent-Based Visual Place Recognition With Ensembles of Temporal Windows
Fischer, T., & Milford, M. J. (2020) · IEEE Robotics and Automation Letters · doi:10.1109/lra.2020.3025505 · Cited by 88
“review to SeqSLAM [18] and NetVLAD [24], which are used in Section III-B to extract features, as well as prior work on VPR using event cameras.”
InfluentialSelf-citationA Hybrid Compact Neural Architecture for Visual Place Recognition
Chancan, M., Hernandez-Nunez, L., Narendra, A., Barron, A. B., & Milford, M. J. (2020) · IEEE Robotics and Automation Letters · doi:10.1109/lra.2020.2967324 · Cited by 75
“2) FlyNet+SeqSLAM: We incorporate the SeqSLAM algorithm [10] on top of our single-frame FlyNet network, as per previous research described in Sections I and II, see Fig.”
BackgroundMethodInfluentialSelf-citationDASGIL: Domain Adaptation for Semantic and Geometric-Aware Image-Based Localization
Hu, H., Qiao, Z., Cheng, M., Liu, Z., & Wang, H. (2020) · IEEE Transactions on Image Processing · doi:10.1109/tip.2020.3043875 · Cited by 79
“Traditional local feature descriptors are aggregated for image retrieval [35], [36], and have successfully addressed most cases of loop closure detection in real-time Visual SLAM (VSLAM) [37], [38] without huge environmental changes.”
BackgroundDV-LOAM: Direct Visual LiDAR Odometry and Mapping
Wang, W., Liu, J., Wang, C., Luo, B., & Zhang, C. (2021) · Remote Sensing · doi:10.3390/rs13163340 · Cited by 74
“Temporal consistency check: In a SLAM system, it is observed that the occurrence of single loop closure often implies high similarity on the neighbour since the sensor feedback is continuous in time [63].”
BackgroundTraining a Convolutional Neural Network for Appearance-Invariant Place RecognitionPreprint
Gomez-Ojeda, R., López-Antequera, M., Petkov, N., & González-Jiménez, J. (2015) · arXiv (Cornell University) · doi:10.48550/arxiv.1505.07428 · Cited by 70
Robust Place Categorization With Deep Domain Generalization
Mancini, M., Bulò, S. R., Caputo, B., & Ricci, E. (2018) · IEEE Robotics and Automation Letters · doi:10.1109/lra.2018.2809700 · Cited by 65
Towards illumination invariance for visual localization
Ranganathan, A., Matsumoto, S., & Ilstrup, D. (2013) · 2013 IEEE International Conference on Robotics and Automation · doi:10.1109/icra.2013.6631110 · Cited by 62
“The alternate word distribution p(Wj |Wq) is learned in a data-driven manner from a training set of matching descriptors under different lighting conditions.”
BackgroundHigh performance loop closure detection using bag of word pairs
Kejriwal, N., Kumar, S., & Shibata, T. (2015) · Robotics and Autonomous Systems · doi:10.1016/j.robot.2015.12.003 · Cited by 70
“Unlike the approaches that focuses on recognizing a place by a single image, there are methods that focus on recognizing a sequence of places [25] or matching trajectories [26].”
BackgroundFast and effective visual place recognition using binary codes and disparity information
Arroyo, R., Alcantarilla, P. F., Bergasa, L. M., Yebes, J. J., & Bronte, S. (2014) · IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) · doi:10.1109/iros.2014.6942989 · Cited by 65
“Recently, this topic has acquired a great importance in visual navigation, and several interesting works have appeared in the last year considering problems such as visual place recognition at different hours of day [22] or during all the seasons [23].”
BackgroundTextPlace: Visual Place Recognition and Topological Localization Through Reading Scene Texts
Hong, Z., Pétillot, Y., Lane, D. M., Miao, Y., & Wang, S. (2019) · IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv.2019.00295 · Cited by 74
“Milford and Wyeth [17] propose SeqSLAM which uses a sequence of global image descriptor for place recognition.”
MethodInfluentialEffective Visual Place Recognition Using Multi-Sequence Maps
Vysotska, O., & Stachniss, C. (2019) · IEEE Robotics and Automation Letters · doi:10.1109/lra.2019.2897160 · Cited by 60
“6 (bottom) shows that our approach has a comparable performance to the FI with around 80% accuracy and they both outperform FAB-MAP, DBoW2, and SeqSLAM for this dataset.”
ResultInfluentialOnline LiDAR-SLAM for Legged Robots with Robust Registration and Deep-Learned Loop Closure
Ramezani, M., Tinchev, G., Egor, I., & Fallon, M. F. (2020) · Proceedings - IEEE International Conference on Robotics and Automation/Proceedings · doi:10.1109/icra40945.2020.9196769 · Cited by 77
“While robustness to illumination change has been explored, for example [14] and [3], it remains a major challenge to any visual navigation system.”
BackgroundDeep Learning for Visual SLAM in Transportation Robotics: A review
Duan, C., Junginger, S., Huang, J., Jin, K., & Thurow, K. (2019) · Transportation Safety and Environment · doi:10.1093/tse/tdz019 · Cited by 63
Lowry, S., & Milford, M. J. (2016) · IEEE Transactions on Robotics · doi:10.1109/tro.2016.2545711 · Cited by 59
“Personal use is permitted, but republication/redistribution requires IEEE permission.”
BackgroundMethodSelf-citationLong-Term 3D Localization and Pose from Semantic Labellings
Toft, C., Olsson, C., & Kahl, F. (2017) · IEEE International Conference on Computer Vision Workshops · doi:10.1109/iccvw.2017.83 · Cited by 61
“The idea of using multiple images for localization (or rather place recognition) was also pursued in [26] where up to 300 consecutive images were used to perform localization based on a image intensity correlation measure.”
MethodHyperdimensional computing as a framework for systematic aggregation of image descriptors
Neubert, P., & Schubert, S. (2021) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr46437.2021.01666 · Cited by 72
“keypoints [34, 11, 44, 12]), place recognition typically builds upon holistic image descriptors that compute a single descriptor vector for a whole image [1, 62, 57, 38, 43].”
BackgroundDE‐SLAM: SLAM for highly dynamic environment
Xing, Z., Zhu, X., & Dong, D. (2022) · Journal of Field Robotics · doi:10.1002/rob.22062 · Cited by 59
“Traditional loop detection methods used in SLAM system, such as bag of visual words algorithm (Filliat, 2007) and sequence recognition algorithm in SeqSLAM (Milford & Wyeth, 2012), failed to tackle extreme illumination or viewpoint changes over time.”
MethodA systematic literature review on long‐term localization and mapping for mobile robots
Sousa, R. B., Sobreira, H., & Moreira, A. P. (2023) · Journal of Field Robotics · doi:10.1002/rob.22170 · Cited by 70
Delta Descriptors: Change-Based Place Representation for Robust Visual Localization
Garg, S., Harwood, B., Anand, G., & Milford, M. J. (2020) · IEEE Robotics and Automation Letters · doi:10.1109/lra.2020.3005627 · Cited by 63
“local features led to the development of appearance-invariant whole-image description techniques, combined with sequencebased matching [7], [8] in order to deal with extreme appearance variations.”
BackgroundMethodSelf-citationMilford, M. J., Lowry, S. M., Sünderhauf, N., Rowlands, S., Pepperell, E., Upcroft, B., Shen, C., Lin, G., Liu, F., Cadena, C., et al. (2015) · IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW) · doi:10.1109/cvprw.2015.7301395 · Cited by 62
“Cropping was necessary to overcome the poor depth estimation for the road just in front of the car.”
BackgroundMethodSelf-citationLearning Sequential Descriptors for Sequence-Based Visual Place Recognition
Mereu, R., Trivigno, G., Berton, G., Masone, C., & Caputo, B. (2022) · IEEE Robotics and Automation Letters · doi:10.1109/lra.2022.3194310 · Cited by 67
A Neurologically Inspired Sequence Processing Model for Mobile Robot Place Recognition
Neubert, P., Schubert, S., & Protzel, P. (2019) · IEEE Robotics and Automation Letters · doi:10.1109/lra.2019.2927096 · Cited by 54
“[2] proposed to search for linear segments of”
BackgroundMethodInfluentialHBST: A Hamming Distance Embedding Binary Search Tree for Feature-Based Visual Place Recognition
Schlegel, D., & Grisetti, G. (2018) · IEEE Robotics and Automation Letters · doi:10.1109/lra.2018.2856542 · Cited by 60
Semantic Mapping for View-Invariant Relocalization
Li, J., Meger, D., & Dudek, G. (2019) · Proceedings - IEEE International Conference on Robotics and Automation/Proceedings · doi:10.1109/icra.2019.8793624 · Cited by 59
“Consequently a variety of methods have been proposed for condition-invariance, including matching image sequences [15]–[17], training location-specific image detectors [18]– [20], predicting environmental changes in images [21]– [23], and transforming images to become more illumination invariant [24]–[26].”
MethodCondition-Invariant Multi-View Place RecognitionPreprint
Fácil, J. M., Olid, D., Montesano, L., & Civera, J. (2019) · arXiv (Cornell University) · doi:10.48550/arxiv.1902.09516 · Cited by 54
“C V ] 25 F e b 2019 [15] and Alderley [11].”
BackgroundMethodInfluentialNeuromorphic vision: From sensors to event‐based algorithms
Annamalai, L., Chakraborty, A., & Thakur, C. S. (2019) · Wiley Interdisciplinary Reviews Data Mining and Knowledge Discovery · doi:10.1002/widm.1310 · Cited by 59
“They have built a place recognition module on top of the conventional SeqSLAM algorithm (Milford & Wyeth, 2012).”
MethodModest-vocabulary loop-closure detection with incremental bag of tracked words
Tsintotas, K. A., Bampis, L., & Gasteratos, A. C. (2021) · Robotics and Autonomous Systems · doi:10.1016/j.robot.2021.103782 · Cited by 53
MultiRes-NetVLAD: Augmenting Place Recognition Training With Low-Resolution Imagery
Khaliq, A., Milford, M. J., & Garg, S. (2022) · IEEE Robotics and Automation Letters · doi:10.1109/lra.2022.3147257 · Cited by 57
“Existing VPR research has delivered a variety of robust techniques such as global image descriptor learning [10], [11], local feature matching [12], GANs-based image translation [13], and use of additional information in the form of semantics [14], [15], [16], depth [17], odometry [18], point clouds [19], [20], and se…”
BackgroundSelf-citationPoint cloud descriptors for place recognition using sparse visual information
Cieslewski, T., Stumm, E. S., Gawel, A., Bosse, M., Lynen, S., & Siegwart, R. (2016) · Proceedings - IEEE International Conference on Robotics and Automation/Proceedings · doi:10.1109/icra.2016.7487687 · Cited by 57
“While for instance SeqSLAM [20] presents better results for stronger appearance changes, it is very sensitive to viewpoint changes [31].”
BackgroundMethodAGRI-SLAM: a real-time stereo visual SLAM for agricultural environment
Islam, R., Habibullah, H., & Hossain, T. (2023) · Autonomous Robots · doi:10.1007/s10514-023-10110-y · Cited by 55
“In addition to that, insufficient correspondences reduce the place recognition performance of the visual SLAM system while it plays a vital role in the loop detection and loop closure of the system (Milford & Wyeth, 2012; Islam & Habibullah, 2022).”
BackgroundGeneral Place Recognition Survey: Toward Real-World Autonomy
Yin, P., Jiao, J., Zhao, S., Xu, L., Huang, G. P., Choset, H., Scherer, S., & Han, J. (2025) · IEEE Transactions on Robotics · doi:10.1109/tro.2025.3550771 · Cited by 73
“Patch-NetVLAD [27] extends NetVLAD by designing a multiscale patch feature fusion mechanism, focusing on local details.”
InfluentialData-Efficient Large Scale Place Recognition with Graded Similarity Supervision
Leyva-Vallina, M., Strisciuglio, N., & Petkov, N. (2023) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr52729.2023.02249 · Cited by 75
“[26] Michael J Milford and Gordon F Wyeth.”
Li, Z., Shang, T., Xu, P., & Deng, Z. (2025) · Artificial Intelligence Review · doi:10.1007/s10462-025-11367-8 · Cited by 50
“SeqSLAM (Milford and Wyeth 2012) introduces a sequence-matching approach in which place recognition is based on the temporal alignment of sequential image frames.”
Robust visual semi-semantic loop closure detection by a covisibility graph and CNN features
Cascianelli, S., Costante, G., Bellocchio, E., Valigi, P., Fravolini, M. L., & Ciarfuglia, T. A. (2017) · Robotics and Autonomous Systems · doi:10.1016/j.robot.2017.03.004 · Cited by 54
Mapping, Localization and Path Planning for Image-Based Navigation Using Visual Features and Map
Thoma, J., Paudel, D. P., Chhatkuli, A., Probst, T., & Gool, L. V. (2019) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr.2019.00756 · Cited by 55
“As a comparison, in green we show the matches produced by SeqSLAM [19], the state of the art for capitalizing on sequential information to improve image matches.”
BackgroundMethodInfluentialSky segmentation with ultraviolet images can be used for navigation
Stone, T., Mangan, M., Ardin, P., & Webb, B. H. (2014) · Robotics: Science and Systems · doi:10.15607/rss.2014.x.047 · Cited by 49
“SeqSLAM was run in both directions for each of the resulting 19 route pairs.”
MethodInfluentialTopological local-metric framework for mobile robots navigation: a long term perspective
Tang, L., Wang, Y., Ding, X., Yin, H., Xiong, R., & Huang, S. (2018) · Autonomous Robots · doi:10.1007/s10514-018-9724-7 · Cited by 53
“The other is to find the topological localization through place recognition (Lowry et al. 2016; Churchill and Newman 2013; Milford and Wyeth 2012; Cummins and Newman 2008).”
BackgroundCNN Feature Boosted SeqSLAM for Real‐Time Loop Closure Detection
Bai, D., Wang, C., Zhang, B., Yi, X., & Yang, X. (2018) · Chinese Journal of Electronics · doi:10.1049/cje.2018.03.010 · Cited by 49
Overview of deep learning application on visual SLAM
Li, S., Zhang, D., Xian, Y., Li, B., Zhang, T., & Zhong, C. (2022) · Displays · doi:10.1016/j.displa.2022.102298 · Cited by 54
Zaffar, M., Khaliq, A., Ehsan, S., Milford, M. J., & McDonald-Maier, K. D. (2019) · arXiv (Cornell University) · doi:10.48550/arxiv.1903.09107 · Cited by 50
“Seq-SLAM [29] is a VPR technique using confusion matrix created by subtracting patch-normalized sequences of frames to find the best matched route.”
MethodSelf-citationAutoPlace: Robust Place Recognition with Single-chip Automotive Radar
Cai, K., Wang, B., & Lu, C. X. (2022) · Proceedings - IEEE International Conference on Robotics and Automation/Proceedings · doi:10.1109/icra46639.2022.9811869 · Cited by 56
Benchmarking 6-DOF Outdoor Visual Localization in Changing ConditionsPreprint
Sattler, T., Maddern, W., Toft, C., Torii, A., Hammarstrand, L., Stenborg, E., Safari, D., Okutomi, M., Pollefeys, M., Šivic, J., et al. (2017) · arXiv (Cornell University) · doi:10.48550/arxiv.1707.09092 · Cited by 51
“Requiring a matched sequence of images in the correct order significantly reduces false positive rates in comparison to single-image retrieval approaches, producing impressive results including direct day-night matches [46].”
BackgroundMethodResultInfluentialTransforming morning to afternoon using linear regression techniques
Lowry, S. M., Milford, M. J., & Wyeth, G. F. (2014) · Proceedings - IEEE International Conference on Robotics and Automation/Proceedings · doi:10.1109/icra.2014.6907432 · Cited by 46
“Localizing against image sequences [10, 15] rather than individual images can significantly improve visual localization ability when environments have experienced drastic change.”
BackgroundSelf-citationAugmenting Visual Place Recognition With Structural Cues
Oertel, A., Cieslewski, T., & Scaramuzza, D. (2020) · IEEE Robotics and Automation Letters · doi:10.1109/lra.2020.3009077 · Cited by 51
“We verify our method throughout several experiments by comparing its retrieval performance to state-of-the-art baselines such as NetVLAD [18], SeqSLAM [26], and DenseVLAD [27].”
MethodInfluentialVisual Route Recognition with a Handful of Bits
Milford, M. J. (2012) · Robotics: Science and Systems · doi:10.15607/rss.2012.viii.038 · Cited by 45
“The research presented here also builds on related work by the authors on localization using ‘whole of image’ appearance-based methods under extreme environmental change [16], in which it was shown that routes could be recognized over day to night, sunny to stormy and summer to winter transitions, albeit with image si…”
BackgroundSelf-citationSRAL: Shared Representative Appearance Learning for Long-Term Visual Place Recognition
Han, F., Yang, X., Deng, Y., Rentschler, M. E., Yang, D., & Zhang, H. (2017) · IEEE Robotics and Automation Letters · doi:10.1109/lra.2017.2662061 · Cited by 45
“(2)”
BackgroundMethodInfluentialThe Raincouver Scene Parsing Benchmark for Self-Driving in Adverse Weather and at Night
Tung, F., Chen, J., Meng, L., & Little, J. J. (2017) · IEEE Robotics and Automation Letters · doi:10.1109/lra.2017.2723926 · Cited by 43
“For example, Milford and Wyeth [20] developed a simultaneous localization and mapping method that achieves robust place recognition by match-”
MethodVysotska, O., Naseer, T., Spinello, L., Burgard, W., & Stachniss, C. (2015) · Proceedings - IEEE International Conference on Robotics and Automation/Proceedings · doi:10.1109/icra.2015.7139576 · Cited by 44
“This approach works well to match image sequences and it was shown that it outperforms methods such a SeqSLAM or FABMAP2, given substantial changes in the appearance of the scene.”
BackgroundMethodInfluentialLook Around You: Sequence-based Radar Place Recognition with Learned Rotational Invariance
Gadd, M., Martini, D. D., & Newman, P. M. (2020) · IEEE/ION Position Location and Navigation Symposium · doi:10.1109/plans46316.2020.9109951 · Cited by 52
“We have have further proven the utility of our rotationally invariant architecture – a crucial enabling factor of our SeqSLAM variant.”
BackgroundInfluentialData-Efficient Large Scale Place Recognition With Graded Similarity SupervisionPreprint
Leyva-Vallina, M., Strisciuglio, N., & Petkov, N. (2021) · arXiv (Cornell University) · doi:10.34894/w4ligp · Cited by 44
AutoMerge: A Framework for Map Assembling and Smoothing in City-Scale Environments
Yin, P., Zhao, S., Lai, H., Ge, R., Zhang, J., Choset, H., & Scherer, S. (2023) · IEEE Transactions on Robotics · doi:10.1109/tro.2023.3290448 · Cited by 54
“We notice that: 1) sequence matching [9] can provide high recall and accuracy over long consistent overlaps but not”
BackgroundMethodInfluentialLightweight, Viewpoint-Invariant Visual Place Recognition in Changing Environments
Lowry, S. M., & Andreasson, H. (2018) · IEEE Robotics and Automation Letters · doi:10.1109/lra.2018.2793308 · Cited by 43
“4(a) compares the performance of VLAD using 16384 bits to SeqSLAM [14], using the OpenSeqSLAM implementation with default parameters [1].”
BackgroundMethodInfluentialLin, S., Cheng, R., Wang, K., & Yang, K. (2018) · Sensors · doi:10.3390/s18082476 · Cited by 44
“As a feasible topological structure to globally optimize the matching results between query images and database images, data association graph [23] outperformed SeqSLAM [40] in higher localization precision.”
BackgroundLighting‐invariant Visual Teach and Repeat Using Appearance‐based Lidar
McManus, C., Furgale, P., Stenning, B. E., & Barfoot, T. D. (2012) · Journal of Field Robotics · doi:10.1002/rob.21444 · Cited by 44
“It is worth mentioning the recent place recognition systems of Milford and Wyeth (2012) (purely topological) and Badino et al. (2012) (hybrid topological/metric), which are able to localize against a database containing images of the same routes, but under extreme environmental changes.”
BackgroundCondition-invariant, top-down visual place recognition
Milford, M. J., Scheirer, W. J., Vig, E., Glover, A., Baumann, O., Mattingley, J. B., & Cox, D. (2014) · Proceedings - IEEE International Conference on Robotics and Automation/Proceedings · doi:10.1109/icra.2014.6907678 · Cited by 42
“%) at 100% precision was lower than in the initial SeqSLAM study (35%), place match coverage over the environment was more even, with the longest segment without place matches measuring approximately 280 metres, versus 1400 metres in the original SeqSLAM result [7].”
BackgroundMethodSelf-citationProbabilistic place recognition with covisibility maps
Stumm, E., Mei, C., & Lacroix, S. (2013) · 2013 IEEE/RSJ International Conference on Intelligent Robots and Systems · doi:10.1109/iros.2013.6696952 · Cited by 42
“Abstraction from single image location models has been addressed in the work of [2], [3], [11], [12].”
BackgroundMethodEncoding the description of image sequences: A two-layered pipeline for loop closure detection
Bampis, L., Amanatiadis, A. A., & Gasteratos, A. C. (2016) · IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) · doi:10.1109/iros.2016.7759667 · Cited by 45
Hou, Y., Zhang, H., & Zhou, S. (2017) · Autonomous Robots · doi:10.1007/s10514-017-9684-3 · Cited by 42
“(b) SeqSLAM (Milford and Wyeth 2012) is another wellknown VPR method.”
BackgroundMethodInfluentialFrom Night to Day: GANs Based Low Quality Image Enhancement
Meng, Y., Kong, D., Zhu, Z., & Zhao, Y. (2019) · Neural Processing Letters · doi:10.1007/s11063-018-09968-2 · Cited by 46
“We evaluate the performance of the proposed method and compare it against several standard techniques on an open available vehicle dataset [27].”
MethodImproving Visual Localization Accuracy in Dynamic Environments Based on Dynamic Region Removal
Cheng, J., Zhang, H., & Meng, M. Q. ‐. (2020) · IEEE Transactions on Automation Science and Engineering · doi:10.1109/tase.2020.2964938 · Cited by 42
“Typically, visual localization can be achieved by a simultaneous visual localization and mapping (visual SLAM) system [16]–[19] or visual odometry (VO) system [20]–[22].”
BackgroundJacobson, A., Zeng, F., Smith, D., Boswell, N., Peynot, T., & Milford, M. J. (2020) · Journal of Field Robotics · doi:10.1002/rob.21978 · Cited by 41
“The query and map images are downsampled and patch normalized to enhance illumination and pose invariance (Milford & Wyeth, 2012).”
MethodSelf-citationLong-term topological localisation for service robots in dynamic environments using spectral maps
Krajnik, T., Fentanes, J. P., Mozos, O. M., Duckett, T., Ekekrantz, J., & Hanheide, M. (2014) · 2014 IEEE/RSJ International Conference on Intelligent Robots and Systems · doi:10.1109/iros.2014.6943205 · Cited by 40
“[19] use sequential images to identify places in an across-seasons outdoor scenario and [20] use an “experience” based approach, which uses a set of images acquired at different times of the day to identify the same place.”
MethodFARLAP: Fast robust localisation using appearance priors
Pascoe, G., Maddern, W., Stewart, A. D., & Newman, P. M. (2015) · Proceedings - IEEE International Conference on Robotics and Automation/Proceedings · doi:10.1109/icra.2015.7140093 · Cited by 40
“A different approach is demonstrated by Seq-SLAM [11], wherein sequences of images are employed.”
MethodNeubert, P., & Protzel, P. (2016) · IEEE Robotics and Automation Letters · doi:10.1109/lra.2016.2517824 · Cited by 41
“Holistic image matching approaches compute a single descriptor for the whole image and showed to be very useful in combination with image sequences [1] and for single image”
BackgroundInfluentialIndoor Localization for Augmented Reality Devices Using BIM, Point Clouds, and Template Matching
Herbers, P., & König, M. (2019) · Applied Sciences · doi:10.3390/app9204260 · Cited by 40
“Even day and night changes can be enough to confuse a trained network, requiring even more training data [21].”
BackgroundRatSLAM: Using Models of Rodent Hippocampus for Robot Navigation and Beyond
Milford, M. J., Jacobson, A., Chen, Z., & Wyeth, G. F. (2016) · Springer tracts in advanced robotics · doi:10.1007/978-3-319-28872-7_27 · Cited by 45
“tion can be achieved at any time of day or night, during any weather, and in any season using sequences of visual images as small as 2 pixels in size [9-12].”
MethodSelf-citationRobust Multimodal Sequence-Based Loop Closure Detection via Structured Sparsity
Zhang, H., Han, F., & Wang, H. (2016) · Robotics: Science and Systems · doi:10.15607/rss.2016.xii.043 · Cited by 39
“Global features can encode whole image information and no dictionary-based quantization is required, which showed promising performance for long-term place recognition [2, 26, 27, 30, 33].”
BackgroundMethodInfluentialRobust Loop Closure Detection based on Bag of SuperPoints and Graph Verification
Yue, H., Miao, J., Yu, Y., Chen, W., & Wen, C. (2019) · IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) · doi:10.1109/iros40897.2019.8967726 · Cited by 43
“Milford et al. [23] proposed SeqSLAM, which believed that two images are judged as a loop closure only if there are continuous matched pairs.”
BackgroundMethodProbabilistic Visual Place Recognition for Hierarchical Localization
Xu, M., Sünderhauf, N., & Milford, M. J. (2020) · IEEE Robotics and Automation Letters · doi:10.1109/lra.2020.3040134 · Cited by 49
“Contrast normalization of the distance matrix used in [10] was not applied, resulting in further significant performance gains.”
BackgroundMethodInfluentialSelf-citationFusionVLAD: A Multi-View Deep Fusion Networks for Viewpoint-Free 3D Place Recognition
Yin, P., Xu, L., Zhang, J., & Choset, H. (2021) · IEEE Robotics and Automation Letters · doi:10.1109/lra.2021.3061375 · Cited by 42
“Average precision score of sequence-matching [21] results on the NCLT dataset.”
MethodInfluentialNeuromorphic Perception and Navigation for Mobile Robots: A Review
Novo, Á. M., Villanueva, F. L., de Marina, H. G., Romero, S. F., & Barranco, F. (2024) · ACM Computing Surveys · doi:10.1145/3656469 · Cited by 47
“This method runs in real-time outperforming the state-of-the-art but, it only tracks a single object.”
MethodSequence searching with CNN features for robust and fast visual place recognition
Bai, D., Wang, C., Zhang, B., Yi, X., & Yang, X. (2017) · Computers & Graphics · doi:10.1016/j.cag.2017.07.019 · Cited by 38
Network Uncertainty Informed Semantic Feature Selection for Visual SLAM
Ganti, P., & Waslander, S. L. (2019) · Canadian Conference on Computer and Robot Vision · doi:10.1109/crv.2019.00024 · Cited by 38
“Traditional feature detectors and descriptors, such as SIFT [2], SURF [3], FAST [4], or ORB [5] aim to satisfy the first 3 criteria, while appearance based methods such as FAB-MAP [6] or SeqSLAM [7] aim to fulfill criteria 4 and 5.”
MethodExploring Performance Bounds of Visual Place Recognition Using Extended Precision
Ferrarini, B., Waheed, M., Waheed, S., Ehsan, S., Milford, M. J., & McDonald-Maier, K. D. (2020) · IEEE Robotics and Automation Letters · doi:10.1109/lra.2020.2969197 · Cited by 45
“ence of dynamic objects [7], [55], viewpoint [19], [28] and seasonal variations [35], [38].”
BackgroundSelf-citationVisual Place Recognition by spatial matching of high-level CNN features
Camara, L. G., & Přeučil, L. (2020) · Robotics and Autonomous Systems · doi:10.1016/j.robot.2020.103625 · Cited by 47
Fast and Robust Bio-inspired Teach and Repeat Navigation
Dall’Osto, D., Fischer, T., & Milford, M. J. (2021) · IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) · doi:10.1109/iros51168.2021.9636334 · Cited by 38
“Downscaling significantly reduces the memory required to store the route, and patch normalisation [34] affords the system some robustness to lighting variation.”
BackgroundSelf-citationYu, L., Yang, T., & Chan, A. B. (2018) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2018.2845371 · Cited by 38
Robust Perception Under Adverse Conditions for Autonomous Driving Based on Data Augmentation
Zheng, Z., Cheng, Y., Xin, Z., Yu, Z., & Zheng, B. (2023) · IEEE Transactions on Intelligent Transportation Systems · doi:10.1109/tits.2023.3297318 · Cited by 47
“3) Rainy Night: We also perform experiments under various settings on the challenging Alderley dataset [23].”
Method
Metadata
- Authors
- Milford, M. J., & Wyeth, G. F.
- Affiliations
- Queensland University of Technology (AU)
- Venue
- 2012 IEEE International Conference on Robotics and Automation
- Publisher
- IEEE
- Published
- 2012-05
- Pages
- 1643-1649
- ISSN
- 1050-4729 (ISSN-L)
- OpenAlex
- W2110405746
- Cited by
- 981
- References
- 22
- Citations by year
- 2026: 22 · 2025: 44 · 2024: 66 · 2023: 61 · 2022: 82 · 2021: 115 · 2020: 110 · 2019: 125 · 2018: 99 · 2017: 74 · 2016: 52 · 2015: 61 · 2014: 38 · 2013: 25 · 2012: 7
- FWCI
- 78.04
- Type
- conference paper
- Language
- en
- Open access
- closed
- Locations
- Proceedings - IEEE International Conference on Robotics and Automation/Proceedings (publishedVersion)
- Topics
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Robotics and Sensor-Based Localization
- Physical Sciences › Engineering › Automotive Engineering › Autonomous Vehicle Technology and Safety
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Advanced Image and Video Retrieval Techniques
- Keywords
- visual SLAM, visual navigation, route recognition, seasonal variation, heavy rainfall, vehicle-mounted camera, trajectory matching
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.