patent.fyi
Applications  /  US 2026/0356407
United States Patent Application Publication

US 2026/0356407 A1

Published
Nov 12, 2026
Filed
Jan 26, 2026
Earliest date on the face
May 8, 2025
Classification
B61L 27/57
First published here
Nov 20, 2026

Generated automatically. Not reviewed by a person. Not legal advice or an opinion.

Section 1 · machine-generated

What if?

Questions about how individual parts of claim 1 might be changed. A question may describe something different from what the claim language recites, or a version of it; no view is offered on which.

1.1Manual entry of inspection findings

What ifWhat if the input module receives inspector selections from a form?
Claim language“receive a data frame comprising sensor data”

The inspector selects the vehicle number, axle position and observed condition from fixed lists at a trackside terminal. The inspection software maps the selected condition to a maintenance category and records the finding against the selected axle.

1.1Continuous analog temperature input

What ifWhat if the input module receives a continuous analog voltage from a trackside infrared detector?
Claim language“receive a data frame comprising sensor data”

An analog comparator monitors the detector voltage during each wheel passage and trips when the voltage rises above a calibrated temperature threshold. A wheel sensor gates the comparator and advances an axle counter so the alarm identifies the affected axle.

1.2Vehicle screening by peak temperature

What ifWhat if the system screens each vehicle using its highest measured bearing temperature and reports the result at vehicle level?
Claim language“determine a location on the vehicle”

A peak-hold circuit retains the highest reading from a trackside infrared detector during a vehicle passage. The controller compares that reading with a temperature limit and flags the vehicle for inspection if the limit is exceeded.

1.2Reference frame from a live camera

What ifWhat if the system captures a fresh reference frame from a corresponding wheel on the opposite side of the vehicle?
Claim language“retrieve a reference frame from a database of reference frames”

A wheel detector triggers opposing cameras as each axle passes. The processor pairs the two images by axle number and mirrors the opposite-side image to align the wheel profiles for comparison.

1.3Fault detection by geometric measurement

What ifWhat if the fault identification module compares wheel profiles using geometric measurements and numeric thresholds?
Claim language“a visual large language model”

A laser line profiler measures the wheel cross-section, and software aligns the measured profile with the reference profile at the flange root. The software calculates flange thickness and flags the location when the difference exceeds a configured tolerance.

Section 2 · machine-generated

Variations

Further arrangements of the same subject matter, described here as of the publication date at the foot of this page.

Other field of use

Reference inspection of elevator doors

A camera in the elevator cab supplies a frame when the doors reach the fully open position. The localisation module uses the floor identifier, door side and camera calibration to map image regions to sections of the landing door sill. It retrieves a stored frame for each section in the same door position. A visual large language model compares each observed region with its reference to determine whether the sill contains an obstruction or visible damage.

Component substitution

Surface comparison using point cloud registration

A laser profiler supplies a three-dimensional point cloud of a bogie bracket, and the localisation module retrieves the reference cloud using the vehicle type and bracket position. An iterative closest point algorithm aligns the clouds using mounting surfaces adjacent to the bracket face under inspection. The fault module calculates signed distances from measured points on the bracket face to the reference surface. It flags a fault when a contiguous region exceeds both the distance and area thresholds stored for that bracket.

Parameter or scale variation

Distance triggered imaging along the underframe

A trackside line-scan camera samples the underframe at 1 mm of vehicle travel per line over a speed range of 5 to 40 km/h. A trackside speed sensor supplies the travel estimate, and a leading axle detector establishes the longitudinal origin. The input module groups 500 consecutive lines into each frame and uses a stride of 450 lines to produce a 50 mm overlap. The localisation module retrieves the matching 500 mm reference frame by vehicle type and longitudinal position, and a visual large language model compares the pair for faults.

Combination with existing systems

Inspection linked to maintenance work orders

An inspection server and a fleet maintenance management system share vehicle identifiers and component position codes. The localisation module uses the position code and the component revision recorded in the maintenance system to retrieve the reference frame. When a visual large language model indicates a fault, the server submits the vehicle identifier, position code, timestamp and links to both frames to the maintenance system. The maintenance system appends the inspection record to an open work order for that component or opens a work order if none exists.

Automated or data-driven variant

Axle keypoints for vehicle frame localisation

A keypoint detector trained on labelled side-view images returns wheel-centre and bogie pivot coordinates in each incoming frame. The localisation module fits those points to a stored geometry template for the vehicle type using the camera calibration and known axle spacing. It maps the inspected region into vehicle coordinates and retrieves the reference frame indexed to that region. A visual large language model compares the observed and reference crops to determine whether the region contains a fault.

Temperature matched references for axle bearings

A trackside infrared camera supplies a radiometric frame, and the input module records ambient temperature and vehicle speed with the frame. The localisation module uses axle passage timestamps and a calibrated camera view to assign each bearing region to an axle number and vehicle side. It retrieves the reference frame indexed by bearing position, vehicle type, ambient temperature band and speed band. The fault module renders both bearing crops on the same temperature scale and passes them to a visual large language model to determine whether the observed bearing shows an abnormal heat pattern.

Specific version of a part

Triggered underframe camera input

A wheel detector triggers a global shutter camera and a pulsed light as an axle passes over the inspection pit. The input module receives each grayscale image with its trigger timestamp and frame number.

Reference lookup by axle position

The localisation module counts wheel detector pulses from the leading end of the vehicle and uses the camera identifier to determine the side. It retrieves a reference image from a database indexed by vehicle type, axle ordinal and side.

Paired images for brake inspection

The fault identification module supplies a current brake assembly image and a reference image from the same position as separate inputs to the visual large language model. A fixed prompt directs the model to assess whether the retaining clip is missing or displaced and return a categorical result.

Paired thermal images for bearing inspection

The fault identification module supplies current and reference thermal images of the same axlebox position to the visual large language model, with both images rendered on the same temperature scale. The model returns an overheating classification based on the extent and location of the hot region relative to the reference.

Section 3 · quoted from USPTO records

Other work

Passages reproduced word for word from two earlier-published U.S. patent documents, placed beside the language of claim 1. No characterization is offered. Each passage is marked A or B for the document it comes from; ¶ numbers are the paragraph numbers in the USPTO’s published full text of that document.

Document AUS 2007/0014467 A1
Published
Jan 18, 2007
Filed
Jul 18, 2005
Class
G06K 9/00
Document BUS 11,281,905 B2
Published
Mar 22, 2022
Filed
Sep 12, 2019
Class
G06K 9/00637
Claim 1
Document A
Document B
1.1a data input module configured to: receive a data frame comprising sensor data indicative of a status of the vehicle;
Once a set of workpiece image acquisition instructions are defined, in various exemplary embodiments of the present invention, the control system 120 executes the instructions and commands the camera system 260 to capture one or more images of the workpiece 20 according to the instructions. The control system 120 will then, under control of the controller 125, input the captured image(s) through the input/output interface 130 and store the captured image(s) in the memory 140.
A · US 2007/0014467 A1 · ¶ 40
In operation 502, a live image 126 received from a UAV 114, for example, as an image file, is read in.
B · US 11,281,905 B2 · ¶ 54
1.2a localisation module configured to: determine a location on the vehicle which corresponds to the sensor data in the data frame, and use the location to retrieve a reference frame from a database of reference frames, the reference frame comprising reference sensor data which corresponds to the determined location on the vehicle; and
In operation 504, a corresponding FOD-free reference image 128 file is located in library 120 and read in. In one example, the FOD-free reference image 128 file includes identifying information allowing it to be identified as corresponding to the live image 1126 file, for example, identifying information, such a location on the runway from the start of the runway.
B · US 11,281,905 B2 · ¶ 54
1.3a fault identification module configured to: provide the data frame and the reference frame to a visual large language model, VLM, to determine from the data frame and the reference frame if the vehicle has a fault in the determined location.
In one embodiment, live images 126 received from UAVs 114 are compared to corresponding FOD-free reference images 128 read in from library 120 to determine if candidate FOD is detected.
B · US 11,281,905 B2 · ¶ 53

An empty cell means only that no passage was reproduced there.

Statements in these documents
Accordingly, a need exists for a template matching method, which provides an increased speed advantage while also ensuring robust template matching and high accuracy template position results for a wide variety of workpiece features.
A · US 2007/0014467 A1 · ¶ 7
The novel template decomposition technique subdivides the defined template into a number of segments that are sufficient to ensure robust template matching and high accuracy template position results for a wide variety of workpiece features, while doing so with reduced and/or approximately optimum processing time.
A · US 2007/0014467 A1 · ¶ 8
These parts remain on the runway and can cause damage to tires of other aircraft, hit the fuselage or windshield/canopy, or get sucked up into an engine.
B · US 11,281,905 B2 · ¶ 3
Identifying information and images of candidate FOD are displayed to the user allowing the user to determine whether to send a team to remove the FOD from the runway.
B · US 11,281,905 B2 · ¶ 5
Cite aspatent.fyi, US 2026/0356407 A1, claim 1, first published 2026-11-20T01:05:31-05:00, https://patent.fyi/US20260356407/, SHA-256 311f1f9c451bdb1a659ab353b13572ffb7ba5e994bceefd8b1f4cb395926398d
First published
Permanent addresshttps://patent.fyi/US20260356407/
Content fingerprint (SHA-256)311f1f9c451bdb1a659ab353b13572ffb7ba5e994bceefd8b1f4cb395926398d
Methodv2026-11-18.2 · automated, not reviewed by a person
Correctionscorrections@patent.fyi