Method and system for detecting a signal structure from a moving video platform

• US 8,229,170 B2
• Filed: 07/31/2008
• Issued: 07/24/2012
• Est. Priority Date: 07/31/2008
• Status: Active Grant

• Current Assignee: General Electric Company
• Original Assignee: General Electric Company
• Inventors: Sebastian, Thomas Baby, Tu, Peter Henry
• Primary Examiner(s): Fitzpatrick, Atiba O
• Application Number:
  US12/183,326
• Publication Number:
  US 20100027841A1
• Time in Patent Office: 1,454 Days
• Field of Search: 382100-107, 348113-120, 356 27- 30, 246473 R-473 A
• US Class Current: 382/104
• CPC Class Codes: B61L 23/041 Obstacle detection
• CPC Class Codes: B61L 3/065 controlling optically
• CPC Class Codes: G06V 10/25 Determination of region of interest [ROI] or a volume of interest [VOI]

Abstract

The present invention aims at providing a method for detecting a signal structure from a moving vehicle. The method for detecting signal structure includes capturing an image from a camera mounted on the moving vehicle. The method further includes restricting a search space by predefining candidate regions in the image, extracting a set of features of the image within each candidate region and detecting the signal structure accordingly.

Claims

1. A method for detecting a signal structure from a moving vehicle comprising:

capturing an image from a camera mounted on the moving vehicle;

restricting a search space by predefining candidate regions in the image based on locations of signal structures along railway tracks stored in a memory;

extracting a set of features of the image within each candidate region; and

detecting the signal structure;

wherein restricting the search space by predefining candidate regions in the image comprises mapping a set of measurement in world coordinates to a set of measurement in image coordinates based on a projection matrix.

2. The method of claim 1, wherein the moving vehicle comprises a railroad train.

3. The method of claim 1, wherein restricting the search space comprises detecting a path of the vehicle.

4. The method of claim 1, wherein the candidate region is provided by calibrating the camera.

5. The method of claim 1, wherein the set of measurement comprises the height of the signal structure with respect to the ground plane.

6. The method of claim 4, wherein the candidate region comprises a candidate box.

7. The method of claim 6, wherein the candidate box size is approximately equivalent to the signal structure size.

8. The method of claim 1, wherein detecting the signal structure comprises classifying the set of features as a signal structure image region or a non-signal structure image region.

9. The method of claim 8, comprising classifying extracted features using a machine learning technique.

10. The method of claim 9, wherein the machine learning technique comprises support vector machine learning technique or boosting technique.

11. The method of claim 1, further comprising tracking the signal structure in the image to eliminate false detections.

12. A method for detecting a signal structure from a moving vehicle comprising:

capturing a video from a camera mounted on the moving vehicle;

extracting an image from the video;

restricting a search space by predefining candidate regions in the image based on locations of signal structures along railway tracks stored in a memory;

extracting a set of features of the image within each candidate region;

detecting the signal structure; and

tracking the detection of the signal structure to eliminate a false detection;

wherein restricting the search space by predefining candidate regions in the image comprises mapping a set of measurement in world coordinates to a set of measurement in image coordinates based on a projection matrix.

13. A system for signal structure detection comprising:

a camera configured to capture video or images from a moving vehicle;

a descriptor configured to extract a set of features from at least one predefined candidate region in the image; and

a classifier configured to classify the set of features as a signal structure image or a non-signal structure image;

wherein the descriptor is further configured to map a set of measurement in world coordinates to a set of measurement in image coordinates based on a projection matrix.

14. The system of claim 13, wherein the camera comprises a video camera or a high-speed digital camera.

15. The system of claim 13, wherein the camera comprises a pan-tilt-zoom camera.

16. The system of claim 13, wherein the moving vehicle comprises a railroad train.

17. The system of claim 13, wherein the candidate region is provided by camera calibration.

18. The system of claim 13, wherein the descriptor and the classifier are implemented by appropriate programming of a digital processor.

19. The system of claim 13, wherein the classifier comprises a machine learning technique algorithm.

20. The system of claim 19, wherein the machine learning technique algorithm comprises adaptive boosting algorithm.

21. The system of claim 19, wherein the machine learning technique algorithm comprises support vector machine algorithm.

22. A non-transitory computer-readable storage medium comprising computer-readable instructions of a computer program that, when executed by a processor, cause the processor to perform a method for detecting a signal structure in an image comprising:

restricting a search space in the image by predefining candidate regions in the image based on locations of signal structures along railway tracks stored in a memory;

extracting a set of features of the image within the candidate region; and

detecting a signal structure;

wherein restricting the search space by predefining candidate regions in the image comprises mapping a set of measurement in world coordinates to a set of measurement in image coordinates based on a projection matrix.

23. The non-transitory computer-readable storage medium of claim 22, further comprising a code for classifying the set of features as a signal structure image or a non-signal structure image.