IP Library Granted Patent US 12,657,932
Granted Patent B2
US 12,657,932 · App. 18/287,801 · Granted Jun 16, 2026

Anomalous road signs

Inventors: Matthew Thornton (San Diego, CA); Venkata Sreekanta Reddy Annapureddy (San Diego, CA); David Jonathan Julian (San Diego, CA); Arvind Yedla (La Jolla, CA)
Assignee: NETRADYNE, INC.
G06V20/582G06V10/774G06V10/776G06V10/82
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 12,657,932
App. No.
18/287,801
Granted
Jun 16, 2026
Kind
B2
Abstract

Systems and methods are provided for detecting and classifying driving behaviors, as well as techniques for improving such systems and methods, which may include identifying geographic locations where anomalous rates of non-compliant driving behaviors are observed, such as in the vicinity of certain traffic signs, traffic lights, and road markings.

Claims (48)

1 . A method of identifying training images, the method comprising:

determining a group compliance score for each driver group of a plurality of driver groups;

determining a location non-compliance score for a geographic location for a first driver group of the plurality of driver groups, wherein the group compliance score of the first driver group ranks among a top portion of the group compliance scores of the plurality of driver groups, wherein the geographic location is associated with the driving scenario;

selecting, image data associated with the geographic location based on the location non-compliance score exceeding a threshold value; and

storing the image data in a machine learning database for training a machine learning model.

2 . The method of claim 1 , further comprising:

receiving event data associated with an event related to a driving behavior from a device coupled to a vehicle, wherein the event data is captured when the vehicle was driven through the geographic location by a driver.

3 . The method of claim 2 , wherein the event data comprises event image data associated with the driving behavior, and further comprising:

processing a portion of the event image data with a neural network to produce inference data; and

detecting that the geographic location is associated with the driving scenario based on the inference data.

4 . The method of claim 3 , further comprising:

determining that the detection of the driving scenario at the geographic location was erroneous;

storing, in the machine learning database, the event image data together with an indication that the event image data does not reflect an example of the driving scenario; and

training the neural network with the machine learning database to detect the driving scenario.

5 . The method of claim 3 , wherein the driving scenario is a stop sign scenario; wherein detecting that the geographic location is associated with a stop sign scenario is based at least in part on a prediction by the neural network that the event image data contains an image of a traffic sign and that the traffic sign is a stop sign; and wherein determining that the detection of the stop sign scenario was erroneous comprises determining that the traffic sign is not a stop sign.

6 . The method of claim 3 , further comprising:

presenting a user of a safety management system with an option to ignore detections of non-compliant driving behaviors at the geographic location.

7 . The method of claim 3 , further comprising:

presenting a user of a safety management system with an option to emphasize detections of non-compliant driving behaviors at the geographic location.

8 . The method of claim 2 , further comprising:

receiving a characterization of the driving behavior exhibited by the driver at the geographic location based on the event data.

9 . The method of claim 8 , wherein the event data comprises event image data associated with the driving behavior, and wherein the driving behavior is characterized as non-compliant and further comprising:

determining that characterization of the driving behavior as non-compliant was erroneous;

storing, in the machine learning database, the event image data together with an indication that the event image data does not correspond to an example of non-compliant driving behavior; and

training a machine learning model with the machine learning database to characterize driving behavior as compliant or non-compliant.

10 . The method of claim 2 , wherein the driver is a member of the first driver group.

11 . The method of claim 2 , wherein the event data comprises event image data associated with the driving behavior, and further comprising:

storing, in the machine learning database, the event image data together with an indication that the event image data corresponds to a true positive example of non-compliant driving behavior; and

training a machine learning model with the machine learning database to characterize driving behavior as compliant or non-compliant.

12 . The method of claim 1 , further comprising:

storing, in the machine learning database, event image data together with an indication that the event image data corresponds to a true positive or a false positive example of non-compliant driving behavior; and

training the machine learning model with the machine learning database to characterize a driving behavior as compliant or non-compliant.

13 . The method of claim 1 , wherein selecting the image data is further based on:

determining that the location non-compliance score for the geographic location is higher than at least one of:

a group non-compliance score, wherein the group non-compliance score is a complement to the group compliance score; or

location non-compliance scores associated with the first driver group at other geographic locations.

14 . The method of claim 1 , wherein the driving scenario is a stop sign scenario, and further comprising:

detecting, based at least in part on the event data, that the vehicle came to a complete stop;

and wherein determining that the characterization of the driving behavior as non-compliant was erroneous is based at least in part on the detection that the vehicle came to a complete stop.

15 . The method of claim 1 , wherein the driving scenario is a stop sign scenario, and further comprising:

determining, based at least in part on the event data, that the vehicle was not expected to come to a complete stop in the presence of a detected stop sign because the stop sign did not apply to the direction that the vehicle travelled through an intersection near the detected stop sign.

16 . The method of claim 1 , wherein the driving scenario is a stop sign scenario, and further comprising:

determining, based at least in part on the event data, that the vehicle was not expected to come to a complete stop in the presence of a detected stop sign because there was a temporary override of traffic rules near the detected stop sign.

17 . A computer program product comprising a non-transitory computer-readable medium having instructions stored thereon, the instructions being executable by one or more processors configured to:

determine a group compliance score for each driver group of a plurality of driver groups;

determine a location non-compliance score for a geographic location for a first driver group of the plurality of driver groups, wherein the group compliance score of the first driver group ranks among a top portion of the group compliance scores of the plurality of driver groups, wherein the geographic location is associated with the driving scenario;

select image data associated with the geographic location based on the location non-compliance score exceeding a threshold value; and

store the image data in a machine learning database for training a machine learning model.

Assignments (3)
SECURITY INTEREST Recorded Apr 6, 2026
From: NETRADYNE, INC.
To: FIRST-CITIZENS BANK & TRUST COMPANY
Reel/Frame 075435/0670 →
SECURITY INTEREST Recorded Apr 6, 2026
From: NETRADYNE, INC.
To: FIRST-CITIZENS BANK & TRUST COMPANY
Reel/Frame 075359/0194 →
ASSIGNMENT OF ASSIGNOR'S INTEREST Recorded Oct 23, 2023
From: THORNTON, MATTHEW; ANNAPUREDDY, VENKATA SREEKANTA REDDY; JULIAN, DAVID JONATHAN; YEDLA, ARVIND
To: NETRADYNE, INC.
Reel/Frame 065304/0991 →
Continuity (2)
Provisional Application 63178525 · Apr 22, 2021
Related Publication 20240193960A1 · Jun 13, 2024
References Cited (9)
US 10830605B1 · Chintakindi · 2020 [cited by examiner]
US 11770505B1 · Nguyen · 2023 [cited by examiner]
US 20180025636A1 · Boykin et al. · 2018 [cited by applicant]
US 20190265712A1 · Satzoda · 2019 [cited by examiner]
US 20200074326A1 · Balakrishnan et al. · 2020 [cited by applicant]
US 20200079387A1 · Krishna et al. · 2020 [cited by applicant]
US 20200097741A1 · Takeyasu · 2020 [cited by examiner]
WO 2017123665A1 · 2017 [cited by applicant]
International Search Report and Written Opinion in International Patent Application No. PCT/US2022/025908 dated Jul. 15, 2022 (14 pages). [cited by applicant]