Track feature detection using machine vision

Patent metadata

Field	Value
Publication number	US-11495009-B2
Application number	US-202016948883-A
Country	US
Kind code	B2
Filing date	Oct 5, 2020
Priority date	Mar 23, 2017
Publication date	Nov 8, 2022
Grant date	Nov 8, 2022

Abstract

The present disclosure generally relates to automated detection of railroad track features. Images of a railroad track are captured and analyzed to identify track features such as anchors, spikes, rail ties, tie plates, and joints. Various image processing techniques are utilized to accurately distinguish between track features and other objects in the captured images. Track features identified in the images are assigned identifiers and locations and stored in a database so that a status and/or condition of the track features may be monitored for maintenance purposes.

First claim

What is claimed is:

1. A railroad track feature detection system comprising:
a camera configured to capture a plurality of images,
one or more light sources; and
a computing apparatus comprising:
at least one memory comprising instructions; and
at least one processing device configured to execute the instructions, wherein the instructions cause the at least one processing device to perform operations comprising:
receiving a first image of the plurality of images;
detecting one or more imaged rail features in the first image;
calculating a confidence score for each detected rail feature in the first image; and
based on the confidence score of a first set of detected rail features being above a feature threshold, identifying the detected rail features in the first set and assigning feature identifiers to each of the identified rail features;
identifying a second set of the detected rail features as unwanted objects and filtering out the unwanted objects.

2. The railroad track feature detection system of claim 1 , wherein the camera and the one or more light sources are coupled to a frame of a rail vehicle and positioned such that the plurality of images include a rail track.

3. The railroad track feature detection system of claim 1 , wherein the feature identifiers includes:
feature identification information that corresponds to a respective railroad track feature of a database of railroad track features; and
location information.

4. The railroad track feature detection system of claim 3 , wherein the database of railroad track features includes an anchor, a spike, a rail tie, a tie plate, and a rail joint.

5. The railroad track feature detection system of claim 1 , wherein the operations comprise:
determining at least one color value of each pixel in the first image;
detecting one or more colored areas in the first image; and
calculating a confidence score for each of the detected color areas in the first image; and
based on the confidence score being above a color threshold, identifying the detected color areas as a color marker and assigning a color identifier to the color markers.

6. The railroad track feature detection system of claim 5 , wherein the color identifiers include:
color identification information that corresponds to a respective railroad track feature of a database of railroad track features; and
location information.

7. The railroad track feature detection system of claim 1 , wherein the operations comprise storing the feature identifiers of the identified rail features in the first image.

8. The railroad track feature detection system of claim 7 , wherein the operations comprise:
receiving a second image of the plurality of images;
detecting one or more imaged rail features in the second image;
calculate a confidence score for each detected rail feature in the second image; and
based on the confidence score being above the feature threshold, identifying the detected rail features of the second image and assigning feature identifiers to each of the identified rail features of the second image.

9. The railroad track feature detection system of claim 8 , wherein the operations comprise storing, in the at least one memory, the feature identifiers of the identified rail features in the second image.

10. The railroad track feature detection system of claim 1 , wherein the operations comprise performing a pattern detection technique to detect one or more imaged rail features in the first image.

11. The railroad track feature detection system of claim 10 , wherein the pattern detection technique includes weighting one or more image characteristics.

12. The railroad track feature detection system of claim 1 , wherein the operations comprise:
receiving a second image of the plurality of images;
detecting one or more imaged rail features in the second image;
identifying at least one of the second set of the detected rail features in the second image and filtering out the unwanted objects in the second image.

13. A non-transitory computer readable medium comprising instructions that, when executed by a processor, causes the processor to:
receive a plurality of images from a camera coupled to a frame of a rail vehicle, the camera positioned such that the plurality of images contain a rail track;
detect a plurality of imaged rail features in the plurality of images;
calculate a confidence score for each detected rail feature; and
based on the confidence score of the detected rail features being above a feature threshold, identify the detected rail features and assign feature identifiers to each of the identified rail features
identify a second set of the detected rail features and classifying the second set of the detected rail features as unwanted objects; and
ignore the unwanted objects.

14. The non-transitory computer readable medium of claim 13 , wherein the instructions further cause the processor to:
weight a plurality of image characteristics within the plurality of images; and
based on the weighted image characteristics, perform a pattern detection on the plurality of images to detect the plurality of imaged rail features.

15. The non-transitory computer readable medium of claim 13 , wherein the feature identifiers include:
feature identification information that corresponds to one railroad track feature of a database of railroad track features; and
location information.

16. The non-transitory computer readable medium of claim 15 , wherein the instructions further cause the processor to:
measure a distance between a first identified rail track feature and a second identified rail track feature; and
wherein:
the first identified rail track feature is contained in a first image of the plurality of images; and
the second identified rail track feature is contained in a second image of the plurality of images.

17. The non-transitory computer readable medium of claim 15 , wherein the database of railroad track features includes an anchor, a spike, a rail tie, a tie plate, and a rail joint.

18. The non-transitory computer readable medium of claim 13 , wherein the instructions further cause the processor to:
determine at least one color value of each pixel in the plurality of images;
detect one or more colored areas in the plurality of images; and
calculate a confidence score for each of the detected color areas; and
based on the confidence score being above a color threshold, identify the detected color areas as color markers and assign color identifiers to the identified color markers.

19. The non-transitory computer readable medium of claim 18 , wherein the color identifiers include:
color identification information that corresponds one railroad track features of a database of railroad track features; and
location information.

20. The non-transitory computer readable medium of claim 19 , wherein the instructions further cause the processor to:
measure a distance between a first color marker and a second color marker; and
wherein:
the first color marker is contained in a first image of the plurality of images; and
the second color marker is contained in a second image of the plurality of images.

Assignees

• Harsco Technologies LLC

Inventors

• Fernandez Javier

Classifications

• G06V10/56 Primary

  relating to colour · CPC title

• B61L23/042 Primary

  Track changes detection · CPC title

• G06V20/56

  exterior to a vehicle by using sensors mounted on the vehicle · CPC title

• H04N7/18

  Closed-circuit television [CCTV] systems, i.e. systems in which the video signal is not broadcast · CPC title

• B61L23/045

  Rail wear · CPC title