IP Library Granted Patent US 12,366,459
Granted Patent B2
US 12,366,459 · App. 18/208,257 · Granted Jul 22, 2025

Localization processing service and observed scene reconstruction service

Inventors: Brian Streem (Jackson Heights, NY); Ricardo Achilles Filho (Bauru, BR); Vijaykumar Sureshkumar (Jersey City, NJ); Marcelo Guimarães (Bauru, BR); Marcelo Eduardo Benencase (Cascavel, BR); Andresso da Silva (Campo Alegre, BR); Guilherme Soares Silvestre (Matão, BR); Gustavo Henrique Benso (Reginópolis, BR); Juan Carlos Conceição de Lima Sales (Campo Grande, BR); Natan Henrique Sanches (São Carlos, BR); Saulo Gabriel Felix (São Cristóvão, BR); Pedro Luiz Cason Caldato (Bauru, BR); Gustavo Henrique Stahl (Limeira, BR); Igor Kenzo Ishikawa Oshiro Nakashima (Campinas, BR)
Assignee: AEROCINE VENTURES, INC.
G01C21/3804G01C21/30G06T7/70G06T7/74G06T17/05G06V10/56G06V10/761G06V10/764G06V10/771G06V20/625G06T2207/20084
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 12,366,459
App. No.
18/208,257
Granted
Jul 22, 2025
Kind
B2
Abstract

Systems, methods, and computer-readable media for providing a localization processing service for enabling localization of a navigation network-restricted subsystem and for providing an observed scene reconstruction service are provided.

Claims (69)

1. A method of localizing an object in an environment using an observing subsystem comprising an image sensor component, a memory component, and a processing module communicatively coupled to the image sensor component and the memory component, the method comprising:

storing, with the memory component, a map feature database comprising a plurality of map feature entries, wherein:

each map feature entry of the plurality of map feature entries is respectively associated with a rendered map image of a plurality of rendered map images rendered from a georeferenced three-dimensional map; and

each rendered map image of the plurality of rendered map images is associated with a respective map location and a respective map orientation;

capturing, at a moment in time with the image sensor component, an observing image of the object;

determining, with the processing module, an observing orientation of the image sensor component at the moment in time;

determining, with the processing module, an observing location of the image sensor component at the moment in time;

defining, with the processing module, a similar image set, wherein:

the similar image set comprises the captured observing image and a particular rendered map image of the plurality of rendered map images; and

the defining comprises determining that:

the observing orientation of the captured observing image satisfies an orientation similarity comparison with the map orientation of the particular rendered map image; and

the observing location of the captured observing image satisfies a location similarity comparison with the map location of the particular rendered map image;

extracting, with a feature extractor model of the processing module, the following:

an image feature map from the captured observing image of the similar image set; and

a map feature map from the particular rendered map image of the similar image set;

performing, with a correlation module of the processing module, a tensor correlation between the extracted image feature map and the extracted map feature map to determine the position of the object on the particular rendered map image; and

based on the determined position of the object on the particular rendered map image, determining, with a raycasting module of the processing module, georeferenced three-dimensional coordinates of the object in the environment.

2. The method of claim 1 , wherein the particular rendered map image has a lower resolution than the captured observing image.

3. The method of claim 1 , wherein the feature extractor model comprises a convolutional neural network.

4. The method of claim 1 , wherein the determined position of the object on the particular rendered map image is indicative of the two-dimensional position of the object on the particular rendered map image.

5. The method of claim 1 , further comprising analyzing, with the processing module, the captured observing image to determine that the object is a pre-defined object of interest.

6. The method of claim 5 , wherein the extracting is in response to the determination that the object is a predefined object of interest.

7. The method of claim 1 , wherein the observing subsystem is navigation network-restricted throughout the duration between the capturing and the determining georeferenced three-dimensional coordinates of the object in the environment.

8. The method of claim 1 , further comprising reconstructing a scene comprising the object in the environment using the observing subsystem and a reconstruction subsystem that is remote from the observing subsystem, wherein:

the observing subsystem further comprises a communication component communicatively coupled to the processing module;

the captured observing image comprises an image of the scene comprising the object; and

the reconstructing comprises:

based on the captured observing image, determining, with the processing module, the following attributes of the scene:

a state-based attribute comprising the determined georeferenced three-dimensional coordinates of the object in the environment; and

a categorical attribute of the object; and

transmitting, with the communication component from the observing subsystem to the reconstruction subsystem, the determined attributes of the scene.

9. The method of claim 8 , wherein no pixel data of the scene is transmitted from the observing subsystem to the reconstruction subsystem.

10. The method of claim 1 , further comprising reconstructing a scene comprising the object in the environment using the observing subsystem and a reconstruction subsystem that is remote from the observing subsystem, wherein:

the reconstruction subsystem comprises a reconstruction memory component, a reconstruction communication component, and a reconstruction processing module communicatively coupled to the reconstruction memory component and the reconstruction communication component;

the captured observing image comprises an image of the scene comprising the object; and

the reconstructing comprises:

receiving, with the reconstruction communication component at the reconstruction subsystem from the observing subsystem, the following attributes of the scene:

a state-based attribute comprising the determined georeferenced three-dimensional coordinates of the object in the environment; and

a categorical attribute of the object;

identifying, with the reconstruction processing module from a plurality of three-dimensional models stored in the reconstruction memory component, a particular three-dimensional model associated with the categorical attribute of the object; and

rendering, with the reconstruction processing module, a visualization of the scene by positioning the particular three-dimensional model in a reconstruction georeferenced three-dimensional map at the georeferenced three-dimensional coordinates of the state-based attribute.

11. The method of claim 10 , wherein no pixel data of the scene is transmitted from the observing subsystem to the reconstruction subsystem.

12. The method of claim 1 , wherein the image sensor component comprises at least one of an infrared sensor or a thermal sensor.

13. A method of reconstructing a scene comprising an object in an environment using a reconstruction subsystem and an observing subsystem that is remote from the reconstruction subsystem, the observing subsystem comprising an image sensor component, a communication component, and a processing module communicatively coupled to the image sensor component and the communication component, the method comprising:

capturing, at a moment in time with the image sensor component, an observing image of the scene comprising the object;

based on the captured observing image, determining, with the processing module, the following attributes of the scene:

a state-based attribute comprising georeferenced three-dimensional coordinates of the object in the environment with a raycasting module of the processing module; and

a categorical attribute of the object; and

transmitting, with the communication component from the observing subsystem to the reconstruction subsystem, the determined attributes of the scene.

14. The method of claim 13 , wherein the observing subsystem is navigation network-restricted during the capturing, the determining, and the transmitting.

15. The method of claim 13 , wherein the categorical attribute of the object comprises an object class of the object.

16. The method of claim 13 , wherein the categorical attribute of the object comprises a color of the object.

17. The method of claim 13 , wherein the object comprises an entity moving within the environment at the moment in time.

18. The method of claim 13 , wherein the object comprises a vehicle.

19. The method of claim 18 , wherein the categorical attribute of the object comprises a license plate number of the vehicle.

20. The method of claim 18 , wherein the categorical attribute of the object comprises at least one of a make or model of the vehicle.

21. The method of claim 13 , wherein the object comprises a human being.

22. The method of claim 13 , wherein the observing subsystem is a mobile subsystem.

23. The method of claim 22 , wherein the observing subsystem is a vehicle.

24. The method of claim 13 , wherein no pixel data of the scene is transmitted from the observing subsystem to the reconstruction subsystem.

25. The method of claim 13 , wherein the image sensor component comprises at least one of an infrared sensor or a thermal sensor.

26. A method of reconstructing a scene comprising an object in an environment using a reconstruction subsystem and an observing subsystem that is remote from the reconstruction subsystem, the reconstruction subsystem comprising a memory component, a communication component, and a processing module communicatively coupled to the memory component and the communication component, the method comprising:

receiving, with the communication component at the reconstruction subsystem from the observing subsystem, the following attributes of the scene:

a state-based attribute comprising georeferenced three-dimensional coordinates of the object in the environment; and

a categorical attribute of the object;

identifying, with the processing module from a plurality of three-dimensional models stored in the memory component, a particular three-dimensional model associated with the categorical attribute of the object;

rendering, with the processing module, a visualization of the scene by positioning the particular three-dimensional model in a georeferenced three-dimensional map at the georeferenced three-dimensional coordinates of the state-based attribute; and

presenting, with an output component of the reconstruction subsystem, at least a portion of the rendered visualization.

27. The method of claim 26 , wherein no pixel data of the scene is received from the observing subsystem by the reconstruction subsystem.

Assignments (1)
ASSIGNMENT OF ASSIGNOR'S INTEREST Recorded Dec 10, 2024
From: STREEM, BRIAN; FILHO, RICARDO ACHILLES; SURESHKUMAR, VIJAYKUMAR; GUIMARÃES, MARCELO; BENENCASE, MARCELO EDUARDO; DA SILVA, ANDRESSO; SILVESTRE, GUILHERME SOARES; BENSO, GUSTAVO HENRIQUE; DE LIMA SALES, JUAN CARLOS CONCEIÇÃO; SANCHES, NATAN HENRIQUE; FELIX, SAULO GABRIEL; CALDATO, PEDRO LUIZ CASON; STAHL, GUSTAVO HENRIQUE; NAKASHIMA, IGOR KENZO ISHIKAWA OSHIRO
To: AEROCINE VENTURES, INC.
Reel/Frame 069564/0013 →
Continuity (3)
Provisional Application 63350856 · Jun 9, 2022
Provisional Application 63422564 · Nov 4, 2022
Related Publication 20230400327A1 · Dec 14, 2023
References Cited (32)
US 7822264B2 · Balslev et al. · 2010 [cited by applicant]
US 8509488B1 · Enge · 2013 [cited by applicant]
US 9467660B1 · Edelstein · 2016 [cited by examiner]
US 9483842B2 · Haglund et al. · 2016 [cited by applicant]
US 9964955B2 · Keivan et al. · 2018 [cited by applicant]
US 10515458B1 · Yakimenko et al. · 2019 [cited by applicant]
US 10992755B1 · Tran · 2021 [cited by examiner]
US 11355019B2 · Streem et al. · 2022 [cited by applicant]
US 20100253598A1 · Szczerba · 2010 [cited by examiner]
US 20120184845A1 · Ishikawa et al. · 2012 [cited by applicant]
US 20130243250A1 · France et al. · 2013 [cited by applicant]
US 20140172296A1 · Shtukater · 2014 [cited by applicant]
US 20170278402A1 · Yalla et al. · 2017 [cited by applicant]
US 20190088030A1 · Masterson · 2019 [cited by examiner]
US 20190355173A1 · Gao · 2019 [cited by examiner]
US 20200218690A1 · Huston et al. · 2020 [cited by applicant]
US 20200279121A1 · Stanimirovic et al. · 2020 [cited by applicant]
US 20200300637A1 · Chiu et al. · 2020 [cited by applicant]
US 20210095992A1 · Kitaura et al. · 2021 [cited by applicant]
US 20210133929A1 · Ackerson et al. · 2021 [cited by applicant]
US 20210150203A1 · Liu et al. · 2021 [cited by applicant]
US 20210240197A1 · Shalev-Shwartz et al. · 2021 [cited by applicant]
US 20220029941A1 · Holmes · 2022 [cited by applicant]
US 20220398775A1 · Streem et al. · 2022 [cited by applicant]
US 20230087702A1 · Carnahan · 2023 [cited by examiner]
US 20230168106A1 · Velkavrh · 2023 [cited by examiner]
US 20230314162A1 · Mori · 2023 [cited by examiner]
Facebook AI Similarity Search: https://github.com/facebookresearch/faiss. [cited by applicant]
Devil is in the Edges: Learning Semantic Boundaries from Noisy Annotations, Acuna et al. (2019). [cited by applicant]
VLASE: Vehicle Localization by Aggregating Semantic Edges: https://github.com/sagachat/VLASE. [cited by applicant]
NetVLAD: CNN architecture for weakly supervised place recognition, Arandjelovic et al. (2016). [cited by applicant]
SuperPoint: Self-Supervised Interest Point Detection and Description, DeTone et al. (2017). [cited by applicant]