IP Library Granted Patent US 12,737,320
Granted Patent B2
US 12,737,320 · App. 16/915,778 · Granted Sep 15, 2026

Method of generating at-scale geospatial features of designated attribution and geometry

Inventors: Mitchell Pillarick, III (Bethalto, IL); Christen Welch (Byrnes Mill, MO); Christopher Black (Kirkwood, MO)
Assignee: The United States of America, as Represented by the Director of the National Geospatial-Intelligence Agency
G06F16/122G06N20/00
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 12,737,320
App. No.
16/915,778
Granted
Sep 15, 2026
Kind
B2
Abstract

A computer-implemented method is disclosed for generating large volumes of data for purposes such as testing data validation tools and training machine-learning models. Selecting a data schema and identifying its rules and conditions establish the characteristics of the data to be generated. The internal data structure can be organized into a multilevel nested hierarchy of bounding boxes that serve as separate containers for articles of data. The bounding boxes can be readily configured in a manner that generates a broad spectrum of data applicable to even the most complex or lengthy data schema, such as creating separate bounding boxes for data that complies with the rules of the schema, and data that fails to comply. By automating data generation in this manner, users can generate desired data at scale, instead of relying on time-intensive manual production efforts, and precisely tailor the generated data for the desired purposes.

Claims (43)

1 . A method of training a model using mimicking training data, the method comprising:

receiving, at a model, mimicking training data generated by a process comprising:

selecting, by a user, a schema having desired or required characteristics, related to geospatial features, for the mimicking training data to comply with the schema;

selecting, by the user, one or more evaluation protocols associated with the selected schema;

creating generator modules for the evaluation protocols;

creating an internal data structure to compile and record mimicking generated training data, the mimicking generated training data being generated from the generator modules after the schema is selected, the internal data structure comprising a plurality of bounding boxes wherein at least a first bounding box is arranged to only contain mimicking generated training data that comply with the schema and at least a second bounding box is arranged to only contain mimicking generated training data that do not comply with the schema;

executing the generator modules to create the mimicking generated training data according to the internal data structure, wherein the mimicking generated training data comprises geospatial features including map features; and

training the model using, at least in part, the mimicking generated training data where the training comprises evaluating whether the model can accurately identify errors in a validation tool.

2 . The method of claim 1 , wherein the plurality of bounding boxes are arranged logically in a hierarchy.

3 . The method of claim 1 , wherein the at least second bounding box contains a plurality of subordinate bounding boxes that each contain only mimicking generated training data that are non-compliant with at least one specified protocol of the selected schema.

4 . The method of claim 1 , wherein the one or more selected evaluation protocols dictates the size of the plurality of the bounding boxes.

5 . The method of claim 1 , wherein the one or more protocols forbids overlap of two or more features.

6 . The method of claim 1 , wherein the boundaries of at least two of the plurality of bounding boxes are defined by geocoordinates, and wherein a logical relationship between the at least two bounding boxes is established according to geocoordinates relative to the two bounding boxes.

7 . The method of claim 1 , wherein at least a third bounding box is generated according to metadata characteristics related to the schema.

8 . The method of claim 1 , wherein the mimicking generated training data comprises vectorized geospatial data and the schema evaluates geospatial features in the mimicking generated training data.

9 . A method of improving the accuracy of a model using mimicking training data, the method comprising:

receiving, at a trained model, mimicking training data generated by a process comprising:

selecting, by a user, a schema having desired or required characteristics, related to geospatial features, for the mimicking training data to comply with the schema;

selecting, by the user, one or more evaluation protocols associated with the selected schema;

creating generator modules for the evaluation protocols;

creating an internal data structure to compile and record mimicking generated training data, the mimicking generated training data being generated from the generator modules after the schema is selected, the internal data structure comprising a plurality of bounding boxes wherein at least a first bounding box is arranged to only contain generated data that comply with the schema and at least a second bounding box is arranged to only contain generated data that do not comply with the schema;

executing the generator modules to create the mimicking generated data according to the internal data structure, wherein the mimicking generated training data comprises geospatial features including map features; and

improving the accuracy of the model by further iteratively training the model using, at least in part, mimicking generated training data newly generated based on real-world data.

10 . The method of claim 9 , wherein the plurality of bounding boxes are arranged logically in a hierarchy.

11 . The method of claim 9 , wherein the at least second bounding box contains a plurality of subordinate bounding boxes that each contain only mimicking generated training data that are non-compliant with at least one specified protocol of the selected schema.

12 . The method of claim 9 , wherein the one or more evaluation protocols dictates the size of the plurality of bounding boxes.

13 . The method of claim 9 , wherein the one or more protocols forbids overlap of two or more features.

14 . The method of claim 9 , wherein the boundaries of at least two of the plurality of bounding boxes are defined by geocoordinates, and wherein a logical relationship between the at least two bounding boxes is established according to geocoordinates relative to the two bounding boxes.

15 . The method of claim 9 , wherein at least a third bounding box is generated according to metadata characteristics related to the schema.

16 . The method of claim 9 , wherein the mimicking generated training data comprises vectorized geospatial data and the schema evaluates geospatial features in the mimicking generated data.

17 . A method of creating mimicking training data, the method comprising:

selecting, a schema having desired or required characteristics, related to geospatial features, for the mimicking training data to comply with the schema;

selecting, one or more evaluation protocols associated with the selected schema;

creating generator modules for the evaluation protocols;

creating an internal data structure to compile and record mimicking generated training data, the mimicking generated training data being generated from the generator modules after the schema is selected, the internal data structure comprising a plurality of bounding boxes wherein at least a first bounding box is arranged to only contain mimicking generated training data that comply with the schema and at least a second bounding box is arranged to only contain mimicking generated training data that do not comply with the schema;

executing the generator modules to create the mimicking generated training data according to the internal data structure, wherein the mimicking generated training data comprises geospatial features including map features; and

training the model using one or more simulations, provided by the mimicking generated training data, as training data.

18 . The method of claim 17 , wherein the plurality of bounding boxes are arranged logically in a hierarchy.

19 . The method of claim 17 , wherein the at least second bounding box contains a plurality of subordinate bounding boxes that each contain only mimicking generated training data that are non-compliant with at least one specified protocol of the selected schema.

20 . The method of claim 17 , wherein the one or more selected evaluation protocols dictates the size of the plurality of bounding boxes.

21 . The method of claim 17 , wherein the one or more protocols forbids overlap of two or more features.

22 . The method of claim 17 , wherein at least a third bounding box is generated according to metadata characteristics related to the schema.

23 . The method of claim 17 , wherein the mimicking generated training data comprises vectorized geospatial data and the schema evaluates geospatial features in the mimicking generated training data.

Assignments (1)
ASSIGNMENT OF ASSIGNOR'S INTEREST Recorded Aug 5, 2026
From: PILLARICK, MITCHELL EDWARD, III; WELCH, CHRISTEN; BLACK, CHRISTOPHER D.
To: UNITED STATES OF AMERICA AS REPRESENTED BY THE DIRECTOR OF THE NATIONAL GEOSPATIAL-INTELLIGENCE AGENCY
Reel/Frame 075531/0970 →
Continuity (2)
Provisional Application 62868268 · Jun 28, 2019
Related Publication 20200409905A1 · Dec 31, 2020
References Cited (14)
US 7239320B1 · Hall · 2007 [cited by examiner]
US 8965906B1 · Werth · 2015 [cited by examiner]
US 10558224B1 · Lin · 2020 [cited by examiner]
US 20060104511A1 · Guo · 2006 [cited by examiner]
US 20080195584A1 · Nath · 2008 [cited by examiner]
US 20080301570A1 · Milstead · 2008 [cited by examiner]
US 20110316854A1 · Vandrovec · 2011 [cited by examiner]
US 20170124230A1 · Liu · 2017 [cited by examiner]
US 20170161341A1 · Hrabovsky · 2017 [cited by examiner]
US 20180165364A1 · Mehta · 2018 [cited by examiner]
US 20190028843A1 · Santarone · 2019 [cited by examiner]
US 20200143499A1 · Erdem · 2020 [cited by examiner]
“What is Geospatial Data?—Geospatial Data—AWS” Amazon Web Services, https://aws.amazon.com/what-is/geospatial-data downloaded Oct. 17, 2025. [cited by applicant]
“What is spatial data and how does it work?” Definition from TechTarget, Rahul Awati et al, published Apr. 9, 2024, downloaded Oct. 17, 2025 from https://www.techtarget.com. [cited by applicant]