Search

EP-3602403-B1 - TRACK FEATURE DETECTION USING MACHINE VISION

EP3602403B1EP 3602403 B1EP3602403 B1EP 3602403B1EP-3602403-B1

Inventors

  • FERNANDEZ, JAVIER

Dates

Publication Date
20260506
Application Date
20180322

Claims (15)

  1. A railroad track feature detection system (100) comprising: a camera (120) operatively coupled to the enclosure and configured to capture images of the railroad track (106) as a railroad chassis (102) is propelled along the railroad track (106); at least one light source (124); and a computing apparatus (200) comprising: at least one memory (210) comprising instructions; and at least one processing device (208, 302) configured for executing the instructions, wherein the instructions cause the at least one processing device (208) to perform the operations of: capturing, using the camera (120), an image (402) of the railroad track (106), wherein the at least one light source (124) is used to capture the image (402), the image (402) being composed of pixels; determining, using a graphical processing unit (GPU) (316) comprised in the at least one processing device (208, 302), at least one color value of each pixel comprised in the image (402); identifying, using a visual recognition unit (318) comprised in the at least one processing device (302), an object in the image (402) based on the determined color values; and assigning, using a tagging unit (320) comprised in the at least one processing device (302), an identifier associated with a railroad track feature (408) and a location along the railroad track (106) to the identified object in a database.
  2. The railroad track feature detection system of claim 1, wherein the railroad track feature (408) is an anchor or a spike; and the at least one color value includes a hue/saturation/value, HSV, value or a red-green-blue, RGB, value.
  3. The railroad track feature detection system of claim 1, wherein the operations further comprise: assigning a first range of hue values to correspond to a first color; assigning a second range of hue values to correspond to a second color; and assigning a third range of hue values to correspond to a third color.
  4. The railroad track feature detection system of claim 1, wherein the railroad track (106) includes at least one visual marker corresponding to a railroad track feature (408).
  5. The railroad track feature detection system of claim 1, wherein the image (402) of the railroad track (106) comprises an illuminated portion of the railroad track (106); and the instructions cause the at least one processing device (302) to further perform the operations of: detecting one or more shadow portions (410) in the image (402), the shadow portions (410) corresponding to a shadow (410) cast by the railroad track feature (408); measuring the shadow portions (410); and identifying the track feature (408) based on the shadow portions (410).
  6. The railroad track feature detection system of claim 1, wherein the operations further comprise identifying a contour based on the determined color values of adjacent pixels being greater than a predetermined threshold.
  7. The railroad track feature detection system of claim 6, wherein: the light source (124) is operable to cast a shadow (410) of a railroad track feature (408); and the processing device (302) is configured to measure the shadow (410) and identify the track feature (408).
  8. A method for detecting a railroad track feature (408), comprising: capturing, using a camera operatively coupled to the enclosure and configured to capture images of the railroad track as a railroad chassis is propelled along the railroad track (120) and a light source (124), an image (402) of a railroad track (106) as a railroad chassis (102) is propelled along the railroad track (106), the image being composed of pixels; determining at least one color value of each pixel comprised in the image (402); identifying an object in the image (402) based on the determined color values; and assigning an identifier associated with a railroad track feature (408) and a location along the railroad track to the identified object.
  9. The method of claim 8, further comprising positioning the light source (124) at an angle to the camera (120).
  10. The method of claim 9, further comprising using the light source (124) to cast a shadow (410) of the track feature (408).
  11. The method of claim 10, further comprising measuring the shadow (410) to identify the railroad track feature (408).
  12. The method of claim 8, further comprising disposing a color marker on the railroad track (106).
  13. The method of claim 12, further comprising detecting the color marker on the railroad track (106) and assigning an identifier and a location to the color marker.
  14. The method of claim 8, further comprising: identifying multiple color ranges, each color range of the multiple color ranges including a corresponding range of color values that correspond to at least one respective color; for at least one pixel, performing a comparison of the at least one color value of the pixel to the color values to at least one of the multiple color ranges; and identifying an object in the image (402) based on the comparison.
  15. The method of claim 8, further comprising filtering unwanted objects in the captured image (402).

Description

BACKGROUND Railroads are typically constructed to include a pair of elongated, substantially parallel rails held in place using spikes, anchors, rail ties, and various other hardware (e.g., track features). Over time, normal wear and tear on the railroad may require maintenance and/or replacement of track features. Traditional methods of maintaining rails and accompanying hardware often include manual inspection of the railroad tracks, which is very labor and time intensive. As such, automation of railroad track inspection is desired.The following documents are relevant prior art: GB 2 473 534 A (HARSCO CORP [US]) 16 March 2011, the article by CUCCHIARA R ET AL: "DETECTING MOVING OBJECTS, GHOSTS, AND SHADOWS IN VIDEO STREAMS", IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, vol. 25, no. 10, 1 October 2003, pages 1337-1342, ISSN: 0162-8828, DOI: 10.1109/TPAMI.2003.1233909, and the article by MOSIN RUSSELL ET AL: "An evaluation of moving shadow detection techniques", COMPUTATIONAL VISUAL MEDIA, vol. 2, no. 3, 19 August 2016, pages 195-217, ISSN: 2096-0433, DOI: 10.1007/s41095-016-0058-0. BRIEF SUMMARY The present disclosure generally relates to a railroad track feature detection system. The system couples with an underside surface or an outside surface of a railroad chassis. The system utilizes a camera pointed at the rails on which the railroad chassis operates to capture images of the railroad track. A high intensity light is used in tandem with the camera so as to illuminate the railroad track and track features during image capture. Illuminating railroad track surfaces provides sufficient lighting for discerning between track discoloration and deliberate colored markers that are used to identify the location of various track features. Additionally, illuminating track features such as anchors and spikes casts shadows upon the rail surface. During image processing, these shadows help identify track features in the captured images. A variety of image processing techniques are deployed. For example, color values for each pixel in a captured image may be determined and used to identify an object or an object edge in the image. Track features identified in captured images are assigned identifiers associated with a type of track feature as well as location information such as global positioning system (GPS) coordinates. Once input into a database, this assigned information may enable efficient tracking and monitoring of a status and/or condition of track features for maintenance purposes. BRIEF DESCRIPTION OF THE DRAWINGS Reference is now made to the following descriptions taken in conjunction with the accompanying drawings. FIGURE 1 illustrates an exemplary railroad track feature detection system according to the present disclosure.FIGURE 2 illustrates an exemplary computing system according to the present disclosure.FIGURE 3A illustrates an exemplary computing environment according to the present disclosure.FIGURE 3B illustrates an exemplary connectivity diagram of the computing environment of FIGURE 3A.FIGURE 4A illustrates exemplary visual processing of a captured railroad track image according to the present disclosure.FIGURE 4B illustrates exemplary visual processing of a captured railroad track image according to the present disclosure.FIGURE 4C illustrates exemplary visual processing of a captured railroad track image according to the present disclosure.FIGURE 5A illustrates an exemplary captured image and an associated processed image of a first anchor type according to the present disclosure.FIGURE 5B illustrates an exemplary captured image and an associated processed image of a second anchor type according to the present disclosure.FIGURE 5C illustrates an exemplary captured image and an associated processed image of a third anchor type according to the present disclosure.FIGURE 5D illustrates an exemplary captured image and an associated processed image of a fourth anchor type according to the present disclosure.FIGURE 5E illustrates an exemplary captured image and an associated processed image of a fifth anchor type according to the present disclosure. DETAILED DESCRIPTION Various embodiments of a railroad track feature detection system are described according to the present disclosure. It is to be understood, however, that the following explanation is merely exemplary in describing the devices and methods of the present disclosure. Accordingly, several modifications, changes, and substitutions are contemplated. As shown in FIGURE 1, a railroad track feature detection system 100 may include a railroad chassis 102. In some embodiments, the railroad chassis 102 may be associated with an unmanned rail maintenance vehicle (e.g., a drone vehicle), a manned rail maintenance vehicle, and/or another rail vehicle. The railroad chassis 102 may include wheels 104 for interfacing with underlying railroad tracks (e.g., rails 106). The railroad chassis 102 may be towed behind another rail vehicle