Granted Patent B2

US 9,030,559 · App. 13/596,608 · Granted May 12, 2015

Abstract

Claims (60)

1. A method of automatically detecting a constrained parametric curve over a set of one or more images, said method comprising:

obtaining a set of one or more binary images of a scene wherein pixels of the binary images are designated as one of an edge pixel or not an edge pixel;

processing at least one of the binary images, said processing including:

applying a Hough transform to the binary image being processed to generate an accumulator array;

determining one or more Hough peaks from the accumulator array;

selecting one or more Hough peaks subject to a set of one or more constraints; and

determining Hough curve segments for the selected Hough peaks;

grouping the Hough curve segments into one or more clusters;

selecting from the one or more clusters a cluster having a greatest number of Hough curve segments grouped therein; and

fitting a curve to the Hough curve segments grouped in the selected cluster, said fit curve representing the constrained parametric curve being detected.

2. The method of claim 1 , wherein said processing is applied to a plurality of the binary images individually and said grouping is applied to the Hough curve segments collectively resulting from said processing of the plurality of the binary images.

3. The method of claim 1 , said method further comprising:

obtaining a set of one or more gray scale images of the scene; and

generating the obtained set of one or more binary images by applying edge detection to the set of one or more gray scale images.

4. The method of claim 3 , said method further comprising:

obtaining a set of one or more color images of the scene; and

converting the set of one or more color images into the obtained set of one or more gray scale images based on luminance.

5. The method of claim 4 , wherein the obtained set of one or more color images are defined in a first color space, said first color space being a Red-Green-Blue (RGB) color space.

6. The method of claim 1 , said method further comprising:

obtaining a set of one or more color or gray scale images of the scene; and

generating the obtained set of one or more binary images by applying thresholds on computed features of curves comprising at least one of color, brightness, texture and thickness to extract said edges.

7. The method of claim 6 , further comprising performing image processing comprising at least one of de-noising, anti-aliasing, contrast enhancement, equalization, erosion and dilation to facilitate said feature computation.

8. The method of claim 6 , wherein the obtained set of one or more color images are defined in a first color space, said first color space being a Red-Green-Blue (RGB) color space.

9. The method of claim 1 , wherein said grouping is achieved by applying k-means clustering to the Hough curve segments.

10. The method of claim 1 , wherein the set of constraints effectively limits the selected Hough peaks to those Hough peaks which correspond to Hough curve segments which are straight lines having an orientation within a designated range.

11. The method of claim 10 , wherein the designated range is plus or minus 10 degrees from horizontal.

12. The method of claim 10 , wherein the Hough curve segments are defined by a first parameter representing a distance of a respective Hough curve segment from an origin and a second parameter representing an angle formed by a vector extending from the origin to the respective Hough curve segment.

13. The method of claim 12 , wherein the set of constrains includes the angle being less than −80 degree or greater than 80 degrees.

14. The method of claim 1 , wherein the scene includes a stop bar on a roadway, and said constrained parametric curve being detected corresponds to said stop bar.

15. The method of claim 1 , wherein said obtained set of one or more binary images are derived from a set of one or more images captured over time by a camera having a substantially fixed frame.

16. An apparatus that executes the method of claim 1 .

17. A non-transitory machine-readable medium including a computer program which when executed performs the method of claim 1 .

18. A camera system comprising:

a camera that obtains over time a set of one or more images of a scene; and

an image processor that:

derives a set of one or more binary images from the set of one or more image obtained by the camera, wherein pixels of the binary images are designated as one of an edge pixel or not an edge pixel;

processes at least one of the binary images by:

applying a Hough transform to the binary image being processed to generate an accumulator array;

determining one or more Hough peaks from the accumulator array;

selecting one or more Hough peaks subject to a set of one or more constraints; and

determining Hough curve segments for the selected Hough peaks;

groups the Hough curve segments into one or more clusters;

selects from the one or more clusters a cluster having a greatest number of Hough curve segments grouped therein; and

fits a curve to the Hough curve segments grouped in the selected cluster, said fit curve representing a constrained parametric curve being detected by said image processor.

19. The camera system of claim 18 , wherein said image processor processes a plurality of the binary images individually and groups the Hough curve segments collectively resulting therefrom.

20. The camera system of claim 18 , wherein said image processor derives the set of one or more binary images by:

applying edge detection to a set of one or more gray scale images.

21. The camera system of claim 20 , wherein the set of one or more images obtained by the camera are color images, and the image processor is further provisioned to convert the color images into the set of one or more gray scale images based on luminance.

22. The camera system of claim 21 , wherein the color images are defined in a first color space, said first color space being a Red-Green-Blue (RGB) color space.

23. The camera system of claim 18 , wherein said image processor derives the set of one or more binary images by:

applying thresholds on computed features of curves comprising at least one of color, brightness, texture and thickness to extract said edges from a set of one or more color or gray scale images.

24. The camera system of claim 23 , wherein said image processor performs image pre-processing steps comprising at least one of de-noising, anti-aliasing, contrast enhancement, equalization, erosion and dilation to facilitate said feature computation.

25. The camera system of claim 23 , wherein the color images are defined in a first color space, said first color space being a Red-Green-Blue (RGB) color space.

26. The camera system of claim 18 , wherein said image processor groups the Hough curve segments by applying k-means clustering to the Hough curve segments.

27. The camera system of claim 18 , wherein the set of constraints effectively limits the selected Hough peaks to those Hough peaks which correspond to Hough curve segments which are straight lines having an orientation within a designated range.

28. The camera system of claim 27 , wherein the designated range is plus or minus 10 degrees from horizontal.

29. The camera system of claim 27 , wherein the Hough curve segments are defined by a first parameter representing a distance of a respective Hough curve segment from an origin and a second parameter representing an angle formed by a vector extending from the origin to the respective Hough curve segment.

30. The camera system of claim 29 , wherein the set of constrains includes the angle being less than −80 degree or greater than 80 degrees.

31. The camera system of claim 18 , wherein the scene includes a stop bar on a roadway, and said constrained parametric curve being detected corresponds to said stop bar.

32. The camera system of claim 18 , wherein said camera has a substantially fixed frame during times when the one or more images are obtained by the camera.

Assignments (9)