IP Library › Granted Patent US 12,592,156
Granted Patent B2
US 12,592,156 · App. 18/312,444 · Granted Mar 31, 2026

Method and system for on-board localization

Inventors: Roland Brockers (Pasadena, CA); Friedrich Alexander Dietsche (Herrliberg, CH); Jeff Hugues Luc Delaune (Pasadena, CA); Pedro Proença (Pasadena, CA); Robert Alexander Hewitt (Tujunga, CA); Georgios Georgakis (Pasadena, CA)
Assignee: CALIFORNIA INSTITUTE OF TECHNOLOGY
G08G5/55G06V20/17G08G5/57
View Patent ↗
Loading inventors, assignments & file history…
Monitor This Case
Get email alerts when status or documents change.
Order Certified Copies
Most orders are placed with the USPTO same day — all within 24 business hours.
Order via The Patent Place →
Pre-filled with this patent's details
Quick Facts
Patent No.
US 12,592,156
App. No.
18/312,444
Filed
May 4, 2023
Granted
Mar 31, 2026
Kind
B2
Art Unit
3661
USPC
701/3
Abstract

A method and system provide for on-board localization in a unmanned aerial system (UAS). A map image is generated (using previously acquired images) of an area that the UAS is overflying. The map image is then processed by orthorectifying, referencing the map image in a global reference frame, and generating an abstract map by detecting features and locating the features in the global reference frame. The UAS is then localized by acquiring camera images during flight, selecting a subset of the camera images as localization images, detecting on-board image features (in the localization images), mapping features from the detected on-board image features to the abstract map, deleting outliers to determine an estimated 3D pose, and refining the 3D pose. The localized UAS then used to autonomously navigate the UAS.

Claims (146)

1 . A method for on-board localization of an unmanned aerial system (UAS), comprising:

(a) generating a map image of an area that the UAS is overflying, wherein:

(i) the map image is generated on-board the UAS using previously acquired images; and

(ii) the UAS is navigating in a GPS-denied environment at a distance where overflown terrain cannot be assumed to be flat;

(b) processing the map image by:

(i) orthorectifying the map image;

(ii) referencing the map image in a global reference frame by mapping pixel coordinates in the map image to the global reference frame;

(iii) generating an abstract map, wherein the generating comprises:

(A) detecting features enforcing a local minimum feature density in the map image; and

(B) the features in the abstract map are located within the global reference frame and assigned a 3D position in the global reference frame;

(c) localizing the UAS by performing on-board the UAS:

(i) acquiring camera images during flight by an on-board camera of the UAS;

(ii) selecting localization images from the camera images via a triage process;

(iii) detecting on-board image features in the localization images;

(iv) performing feature association from the detected on-board image features to the abstract map;

(v) deleting matching outliers to determine an estimate of a 3D pose of the UAS in the global reference frame;

(vi) refining the 3D pose of the UAS as an absolute pose of the UAS in the global reference frame; and

(d) utilizing the localized UAS to autonomously navigate and control the UAS.

2 . The method of claim 1 , wherein the previously acquired images are augmented by:

available orbital image maps;

images from a previous flight of a different aerial platform; or

images from a previous flight of the same aerial platform.

3 . The method of claim 1 , wherein the localizing further comprises:

using an on-board estimated pose of the UAS as a pose prior to predict where a feature appears in the map image using a pose error covariance plus a margin, to restrict a search range for matching in the global reference frame.

4 . The method of claim 1 , wherein:

the localizing:

identifies keyframe images from the previously acquired images;

stores the keyframe images in a database on-board the UAS;

utilizes a place recognition algorithm that determines an association between a current localization image from the localization images with one of the keyframe images in the database using a bag-of-words approach;

calculates the 3D pose of the UAS with respect to a previously recorded pose of the UAS when the keyframe image was taken;

improves an accuracy of the 3D pose by utilizing a loop-closure algorithm that uses the association to optimize all camera poses from the localization images and the keyframe images; and

updates the map image based on the optimized camera poses.

5 . The method of claim 4 , wherein:

the selecting the localization images and the storing of the keyframe images are performed on-board and on-line during the flight.

6 . The method of claim 1 , further comprising:

feeding back results of the localizing into an on-board state estimator to directly eliminate drift and reduce error of on-board state estimates continuously in real-time to improve accuracy of UAS navigation; and

utilizing an odometry frame for on-board estimated poses;

utilizing a control frame to avoid position deltas that would lead to control outputs that would make the UAS unstable.

7 . The method of claim 1 , further comprising:

coping with a changing resolution between a time the map image was taken and a time the localization images were acquired on-board by:

deploying a multi-resolution matching approach that incorporates feature mapping at multiple image pyramid levels in the map image and in the localization images.

8 . The method of claim 1 , further comprising:

coping with a changing illumination between a time the map image was taken and a time the localization images were acquired on-board by:

extending the map image with images taken of a same location at a different time of day to produce features that each have one or more descriptors extracted from the images taken of the same location, wherein the one more descriptors can be used to match the localization images within a larger range of illumination compared to the non-extended map image.

9 . The method of claim 1 , further comprising:

coping with a changing illumination between a time the map image was taken and a time the localization images were acquired on-board by:

using a digital elevation map (DEM) to render virtual shadows on a terrain for different times of day and use the rendered virtual shadows to alter a brightness of pixels in the map image in order to derive descriptors that are adapted to a simulated illumination regime at the different times of day.

10 . The method of claim 1 , further comprising:

using a machine-learning algorithm to match features agnostic to illumination and scale changes.

11 . An unmanned aerial system (UAS) comprising:

(a) a UAS vehicle comprising:

(i) one or more rotors for operating the UAS vehicle aerially;

(ii) a processor;

(iii) software executed by the processor for conducting onboard autonomy of the UAS vehicle, wherein the software:

(A) generates a map image of an area that the UAS is overflying, wherein:

(1) the map image is generated on-board the UAS using previously acquired images; and

(2) the UAS is navigating in a GPS-denied environment at a distance where overflown text cannot be assumed to be flat;

(B) processes the map image by:

(1) orthorectifying the map image;

(2) referencing the map image in a global reference frame by mapping pixel coordinates in the map image to the global reference frame;

(3) generating an abstract map, wherein the generating comprises:

 detecting features enforcing a local minimum feature density in the map image; and

 the features in the abstract map are located within the global reference frame and assigned a 3D position in the global reference frame;

(C) localizes the UAS by performing on-board the UAS:

(1) acquiring camera images during flight by an on-board camera of the UAS;

(2) selecting localization images from the camera images via a triage process;

(3) detecting on-board image features in the localization images;

(4) performing feature association from the detected on-board image features to the abstract map;

(5) deleting matching outliers to determine an estimate of a 3D pose of the UAS in the global reference frame;

(6) refining the 3D pose of the UAS as an absolute pose of the UAS in the global reference frame; and

(D) utilizes the localized UAS to autonomously navigate and control the UAS.

12 . The UAS of claim 11 , wherein the previously acquired images are augmented by:

available orbital image maps;

images from a previous flight of a different aerial platform; or

images from a previous flight of the same aerial platform.

13 . The UAS of claim 11 , wherein the software localizes by further:

using an on-board estimated pose of the UAS as a pose prior to predict where a feature appears in the map image using a pose error covariance plus a margin, to restrict a search range for matching in the global reference frame.

14 . The UAS of claim 11 , wherein:

the software localizes by:

identifying keyframe images from the previously acquired images;

storing the keyframe images in a database on-board the UAS;

utilizing a place recognition algorithm that determines an association between a current localization image from the localization images with one of the keyframe images in the database using a bag-of-words approach;

calculating the 3D pose of the UAS with respect to a previously recorded pose of the UAS when the keyframe image was taken;

improving an accuracy of the 3D pose by utilizing a loop-closure algorithm that uses the association to optimize all camera poses from the localization images and the keyframe images; and

updating the map image based on the optimized camera poses.

15 . The UAS of claim 14 , wherein:

the selecting the localization images and the storing of the keyframe images are performed on-board and on-line during the flight.

16 . The UAS of claim 11 , wherein the software further:

feeds back results of the localizing into an on-board state estimator to directly eliminate drift and reduce error of on-board state estimates continuously in real-time to improve accuracy of UAS navigation; and

utilizes an odometry frame for on-board estimated poses;

utilizes a control frame to avoid position deltas that would lead to control outputs that would make the UAS unstable.

17 . The UAS of claim 11 , wherein the software further:

copes with a changing resolution between a time the map image was taken and a time the localization images were acquired on-board by:

deploying a multi-resolution matching approach that incorporates feature mapping at multiple image pyramid levels in the map image and in the localization images.

18 . The UAS of claim 11 , wherein the software further:

copes with a changing illumination between a time the map image was taken and a time the localization images were acquired on-board by:

extending the map image with images taken of a same location at a different time of day to produce features that each have one or more descriptors extracted from the images taken of the same location, wherein the one or more descriptors can be used to match the localization images within a larger range of illumination compared to the non-extended map image.

19 . The UAS of claim 11 , wherein the software further:

copes with a changing illumination between a time the map image was taken and a time the localization images were acquired on-board by:

using a digital elevation map (DEM) to render virtual shadows on a terrain for different times of day and use the rendered virtual shadows to alter a brightness of pixels in the map image in order to derive descriptors that are adapted to a simulated illumination regime at the different times of day.

20 . The UAS of claim 11 , wherein the software further:

uses a machine-learning algorithm to match features agnostic to illumination and scale changes.

21 . A method for on-board localization of an unmanned aerial system (UAS), comprising:

(a) generating a map image of an area that the UAS is overflying, wherein the map image is generated using previously acquired images;

(b) processing the map image by:

(i) orthorectifying the map image;

(ii) referencing the map image in a global reference frame by mapping pixel coordinates in the map image to the global reference frame;

(iii) generating an abstract map, wherein the generating comprises:

(A) detecting features enforcing a local minimum feature density in the map image; and

(B) the features in the abstract map are located within the global reference frame and assigned a 3D position in the global reference frame;

(c) localizing the UAS by:

(i) acquiring camera images during flight by an on-board camera of the UAS;

(ii) selecting localization images from the camera images via a triage process;

(iii) detecting on-board image features in the localization images;

(iv) performing feature association from the detected on-board image features to the abstract map;

(v) deleting matching outliers to determine an estimate of a 3D pose of the UAS in the global reference frame;

(vi) refining the 3D pose of the UAS as an absolute pose of the UAS in the global reference frame;

(d) utilizing the localized UAS to autonomously navigate and control the UAS; and

(e) coping with a changing environment between a time the map image was taken and a time the localization images were acquired on-board by:

(i) using temporal integration to identify areas of change within the map image and invalidate such areas; and

(ii) processing the map image to:

(A) identify areas of feature distribution that are below a defined distribution threshold reflecting potential obstructions or false feature appearance due to introduced artificial texture; and

(B) invalidate the identified areas.

22 . The method of claim 21 , wherein the coping with the changing environment further comprises:

the UAS adapting a motion plan to avoid areas with map feature density below a density threshold or areas with map coverage below a coverage threshold.

23 . The method of claim 21 , wherein the previously acquired images are augmented by:

available orbital image maps;

images from a previous flight of a different aerial platform; or

images from a previous flight of the same aerial platform.

24 . The method of claim 21 , further comprising:

uses a machine-learning algorithm to match features agnostic to illumination and scale changes.

25 . The method of claim 21 , wherein the localizing further comprises:

using an on-board estimated pose of the UAS as a pose prior to predict where a feature appears in the map image using a pose error covariance plus a margin, to restrict a search range for matching in the global reference frame.

26 . The method of claim 21 , further comprising:

feeding back results of the localizing into an on-board state estimator to directly eliminate drift and reduce error of on-board state estimates continuously in real-time to improve accuracy of UAS navigation; and

utilizing an odometry frame for on-board estimated poses;

utilizing a control frame to avoid positions deltas that would lead to control outputs that would make the UAS unstable.

27 . The method of claim 21 , further comprising:

coping with a changing resolution between a time the map image was taken and a time the localization images were acquired on-board by:

deploying a multi-resolution matching approach that incorporates feature mapping at multiple image pyramid levels in the map image and in the localization images.

28 . The method of claim 21 , further comprising:

coping with a changing illumination between a time the map image was taken and a time the localization images were acquired on-board by:

extending the map image with images taken of a same location at a different time of day to produce features that each have one or more descriptors extracted from the images taken of the same location, wherein the one or more descriptors can be used to match the localization images within a larger range of illumination compared to the non-extended map image.

29 . The method of claim 21 , further comprising:

coping with a changing illumination between a time the map image was taken and a time the localization images were acquired on-board by:

using a digital elevation map (DEM) to render virtual shadows on a terrain for different times of day and use the rendered virtual shadows to alter a brightness of pixels in the map image in order to derive descriptors that are adapted to a simulated illumination regime at the different times of day.

Assignments (1)
ASSIGNMENT OF ASSIGNOR'S INTEREST Recorded Jun 6, 2023
From: BROCKERS, ROLAND; DIETSCHE, FRIEDRICH ALEXANDER; DELAUNE, JEFF HUGUES LUC; PROENÇA, PEDRO; HEWITT, ROBERT ALEXANDER; GEORGAKIS, GEORGIOS
To: CALIFORNIA INSTITUTE OF TECHNOLOGY
Reel/Frame 063866/0469 →
Continuity (3)
Provisional Application 63441020 · Jan 25, 2023
Provisional Application 63338381 · May 4, 2022
Related Publication 20230360547A1 · Nov 9, 2023
References Cited (79)
US 5146228A · Irani · 1992 [cited by examiner]
US 6853332B1 · Brookes · 2005 [cited by examiner]
US 10037028B2 · Loianno · 2018 [cited by examiner]
US 10515458B1 · Yakimenko · 2019 [cited by examiner]
US 11107275B2 · Millin · 2021 [cited by examiner]
US 11244500B2 · Marschner · 2022 [cited by examiner]
US 11808578B2 · Wissler · 2023 [cited by examiner]
US 20090210092A1 · Park · 2009 [cited by examiner]
US 20090285450A1 · Kaiser · 2009 [cited by examiner]
US 20160195878A1 · Peleg · 2016 [cited by examiner]
US 20170084037A1 · Barajas Hernandez · 2017 [cited by examiner]
US 20190089602A1 · Bagade · 2019 [cited by examiner]
US 20200202143A1 · Viswanathan · 2020 [cited by examiner]
US 20200240790A1 · Behrendt · 2020 [cited by examiner]
US 20200334857A1 · Garud · 2020 [cited by examiner]
US 20210318140A1 · Tang · 2021 [cited by examiner]
US 20210366148A1 · Gliner · 2021 [cited by examiner]
US 20220164595A1 · Zhou · 2022 [cited by examiner]
US 20220180476A1 · Poddar · 2022 [cited by examiner]
US 20220196432A1 · Buda · 2022 [cited by examiner]
US 20230112417A1 · Schildwaechter · 2023 [cited by examiner]
US 20240257387A1 · Peterfreund · 2024 [cited by examiner]
EP 2829940A2 · 2015 [cited by examiner]
WO WO2024074595A1 · 2024 [cited by examiner]
Couturier, A. et al. “A review on absolute visual localization for UAV”. Robotics and Autonomous Systems, 2021, pp. 1-17, 103666, vol. 135 [online], [retrieved on Apr. 18, 2025]. Retrieved from the Internet <URL: https:… [cited by examiner]
Vaswani, A., et al., “Attention Is All You Need”, 31st Conference on Neural Information Processing Systems (NIPS 2017), pp. 1-11. [cited by applicant]
Alcantarilla, P.F., et al., “Fast Explicit Diffusion for Accelerated Features in Nonlinear Scale Spaces”, IEEE Transactions on Pattern Analysis and Machine Intelligence, 2013, pp. 1281-1298, vol. 34, No. 7. [cited by applicant]
Sivic, J., et al., “Video Google: A Text Retrieval Approach to Object Matching in Videos”, Proceedings of the IEEE International Conference on Computer Vision, 2003, pp. 1470-1477, vol. 2. [cited by applicant]
Baeza-Yates, R., et al., “Modern Information Retrieval”, ACM Press, 1999, pp. 1-38, vol. 463. [cited by applicant]
Bapst, J., et al., “Mars Science Helicopter—Compelling Science Enabled by an Aerial Platform”, National Aeronautics and Space Administration, Planetary Science and Astrobiology Decadal Survey 2023-2032, 2021, pp. 1-8. [cited by applicant]
Rapin, W, et al., “Critical knowledge gaps in the Martian geological record: A rationale for regional-scale in situ exploration by rotorcraft mid-air deployment”, Planetary Science and Astrobiology Decadal Survey 202320… [cited by applicant]
Couturier, A., et al., “A review on absolute visual localization for UAV”, Robotics and Autonomous Systems, 2021, pp. 1-17, 103666, vol. 135. [cited by applicant]
Delaune, J., et al., “Visual-inertial navigation for pinpoint planetary landing using scale-based landmark matching”, Robotics and Autonomous Systems, 2016, pp. 63-82, vol. 78. [cited by applicant]
Conte, G., et al., “Vision-Based Unmanned Aerial Vehicle Navigation Using Geo-Referenced Information”, EURASIP Journal on Advances in Signal Processing, 2009, pp. 1-18. [cited by applicant]
Yol, A., et al., “Vision-based Absolute Localization for Unmanned Aerial Vehicles”, IEEE/RSJ Int. Conf. on Intelligent Robots and Systems, IROS'14, Sep. 2014, pp. 1-6. [cited by applicant]
Mourikis, A.I., et al., “A Multi-State Constraint Kalman Filter for Vision-aided Inertial Navigation”, Proceedings of the IEEE International Conference on Robotics and Automation, 2007, pp. 3565-3572. [cited by applicant]
Matthies, L., et al., “Terrain Relative Navigation for Guided Descent on Titan”, 2020 IEEE Aerospace Conference, 2020, pp. 1-16. [cited by applicant]
Mourikis, A.I., et al., “Vision-Aided Inertial Navigation for Spacecraft Entry, Descent, and Landing”, IEEE Transactions On Robotics, Apr. 2009, pp. 264-280, vol. 25, No. 2. [cited by applicant]
Lowe, D., “Distinctive Image Features from Scale-Invariant Keypoints”, International Journal of Computer Vision, 2004, pp. 91-110, vol. 60. [cited by applicant]
Pham, B.V., et al., “Vision-Based Absolute Navigation for Descent and Landing”, Journal of Field Robotics, 2012, pp. 627-647, vol. 29, No. 4. [cited by applicant]
Cheng, Y., et al., “Landmark Based Position Estimation for Pinpoint Landing on Mars”, Proceedings of the 2005 IEEE International Conference on Robotics and Automation, 2005, pp. 1-6. [cited by applicant]
Dusmanu, M., et al., “D2-Net: A Trainable CNN for Joint Description and Detection of Local Features”, Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2019, pp. 8092-8101. [cited by applicant]
Detone, D., et al., “SuperPoint: Self-Supervised Interest Point Detection and Description”, CVPR 2018 Deep Learning for Visual SLAM Workshop, 2018, pp. 1-13. [cited by applicant]
Jin, Y., et al., “Image Matching Across Wide Baselines: From Paper to Practice”, International Journal of Computer Vision, 2021, pp. 517-547, vol. 129. [cited by applicant]
Delaune, J., et al., “Extended Navigation Capabilities for a Future Mars Science Helicopter Concept”, IEEE Aerospace Conference, 2020, pp. 1-10. [cited by applicant]
Delaune, J., et al., “Range-Visual-Inertial Odometry: Scale Observability Without Excitation”, IEEE Robotics and Automation Letters, Apr. 2021, pp. 2421-2428, vol. 6, No. 2. [cited by applicant]
Shah, S., et al., “AirSim: High-Fidelity Visual and Physical Simulation for Autonomous Vehicles”, Field and Service Robotics, 2017, pp. 621-635. [cited by applicant]
Bay, H., et al., “SURF: Speeded Up Robust Features”, European Conference on Computer Vision, 2006, LNCS 3951, pp. 404-417. [cited by applicant]
Rublee, E., et al., “ORB: an efficient alternative to SIFT or SURF”, International Conference on Computer Vision, 2011, pp. 1-8. [cited by applicant]
Mikolajczyk, K., et al., “Scale & Affine Invariant Interest Point Detectors”, International Journal of Computer Vision, 2004, pp. 63-86, vol. 60, No. 1. [cited by applicant]
Brockers, R., et al., “Autonomous Safe Landing Site Detection for a Future Mars Science Helicopter”, IEEE Aerospace Conference, Mar. 2021, pp. 1-8. [cited by applicant]
Delaune, J., et al., “xVIO: A Range-Visual-Inertial Odometry Framework”, Jet Propulsion Laboratory, California Institute of Technology, 2020, pp. 1-67. [cited by applicant]
Brockers, R., et al., “On-board Absolute Localization Based on Orbital Imagery for a Future Mars Science Helicopter”, IEEE Aerospace Conference, Mar. 2022, pp. 1-11. [cited by applicant]
Clemente, L.A., et al., “Mapping Large Loops with a Single Hand-Held Camera”, Robotics: Science and Systems, 2007, pp. 1-8. [cited by applicant]
Williams, B., et al., “Real-Time SLAM Relocalisation”, Proceedings of the IEEE International Conference on Computer Vision, 2007, pp. 1-8. [cited by applicant]
Nister, D., et al., “Scalable Recognition with a Vocabulary Tree”, Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition, 2006, pp. 2161-2168, vol. 2. [cited by applicant]
Jegou, H., et al., “Hamming Embedding and Weak Geometric Consistency for Large Scale Image Search”, European Conference on Computer Vision, 2008, pp. 304-317, vol. 5302. [cited by applicant]
Philbin, J., et al., “Lost in Quantization: Improving Particular Object Retrieval in Large Scale Image Databases”, Proceeding of the IEEE Conference on Computer Vision and Pattern Recognition, 2008, pp. 1-8. [cited by applicant]
Cummins, M., et al., “FAB-MAP: Probabilistic Localization and Mapping in the Space of Appearance”, The International Journal of Robotics Research, Jun. 2008, pp. 647-665, vol. 27, No. 6. [cited by applicant]
Calonder, M., et al., “BRIEF: Binary Robust Independent Elementary Features”, European Conference on Computer Vision, 2010, pp. 778-792, vol. 6314. [cited by applicant]
Galvez-Lopez, D., et al., “Bags of Binary Words for Fast Place Recognition in Image Sequences”, IEEE Transactions on Robotics, 2012, pp. 1188-1197, vol. 28, No. 5. [cited by applicant]
Mur-Artal, R., et al., “ORB-SLAM: A Versatile and Accurate Monocular SLAM System”, IEEE Transactions On Robotics, Oct. 2015, pp. 1147-1163, vol. 31, No. 5. [cited by applicant]
Mur-Artal, R., et al., “ORB-SLAM2: An Open-Source SLAM System for Monocular, Stereo, and RGB-D Cameras”, IEEE Transactions On Robotics, Oct. 2017, pp. 1255-1262, vol. 33, No. 5. [cited by applicant]
Campos, C., et a., “ORB-SLAM3: An Accurate Open-Source Library for Visual, Visual-Inertial, and Multimap SLAM”, IEEE Transactions On Robotics, Dec. 2021, pp. 1874-1890, vol. 37, No. 6. [cited by applicant]
Qin, T., et al., “VINS-Mono: A Robust and Versatile Monocular Visual-Inertial State Estimator”, IEEE Transactions On Robotics, Aug. 2018, pp. 1004-1020, vol. 34, No. 4. [cited by applicant]
Sumikura, S., et al., “OpenVSLAM: A Versatile Visual SLAM Framework”, MM 2019—Proceedings of the 27th ACM International Conference on Multimedia, 2019, pp. 2292-2295. [cited by applicant]
Schops, T., et al., “BAD SLAM: Bundle Adjusted Direct RGB-D SLAM”, Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition, Jun. 2019, pp. 134-144. [cited by applicant]
Giubilato, R., et al., “Multi-Modal Loop Closing in Unstructured Planetary Environments with Visually Enriched Submaps”, 2021 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2021, pp. 8758-87… [cited by applicant]
Arandjelovic, R., et al., “NetVLAD: CNN architecture for weakly supervised place recognition”, IEEE Transactions on Pattern Analysis and Machine Intelligence, 2018, pp. 1437-1451, vol. 40, No. 6. [cited by applicant]
Naseer, T., et al., “Semantics-aware Visual Localization under Challenging Perceptual Conditions”, 2017 IEEE International Conference on Robotics and Automation (ICRA), 2017, pp. 2614-2620. [cited by applicant]
Latif, Y., et al., “Addressing Challenging Place Recognition Tasks using Generative Adversarial Networks”, 2018 IEEE International Conference on Robotics and Automation (ICRA), 2018, pp. 2349-2355. [cited by applicant]
Nister, D., “An Efficient Solution to the Five-Point Relative Pose Problem”, IEEE Transactions On Pattern Analysis and Machine Intelligence, Jun. 2004, pp. 756-770, vol. 26, No. 6. [cited by applicant]
Muja, M., et al., “Fast Approximate Nearest Neighbors With Automatic Algorithm Configuration”, Proceedings of the 4th International Conference on Computer Vision Theory and Applications, 2009, pp. 331-340, vol. 2, No. 2. [cited by applicant]
Mur-Artal, R., et al., “Fast Relocalisation and Loop Closing in Keyframe-Based SLAM”, Proceedings of the IEEE International Conference on Robotics and Automation, 2014, pp. 846-853. [cited by applicant]
Bay, H., et al., “Speeded-Up Robust Features (SURF)”, Computer Vision and Image Understanding, 2008, pp. 346-359, vol. 110, No. 3. [cited by applicant]
Alahi, A., et al., “FREAK: Fast Retina Keypoint”, Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition, 2012, pp. 510-517. [cited by applicant]
Leutenegger, S., et al., “BRISK: Binary Robust Invariant Scalable Keypoints”, Proceedings of the IEEE International Conference on Computer Vision, 2011, pp. 2548-2555. [cited by applicant]
Alcantarilla, “Fast Explicit Diffusion for Accelerated Features in Nonlinear Scale Spaces”, IEEE Transactions on Pattern Analysis and Machine Intelligence, 2013, pp. 1281-1298, vol. 34, No. 7. [cited by applicant]
Lin, T-Y, et al., “Feature Pyramid Networks for Object Detection”, CVPR 2017, pp. 1-9. [cited by applicant]