Method for Road Sign Recognition

• US 20130058534A1
• Filed: 04/28/2011
• Published: 03/07/2013
• Est. Priority Date: 05/14/2010
• Status: Active Grant

• Current Assignee: Conti Temic Microelectronic GmbH (Continental AG)
• Original Assignee: Conti Temic Microelectronic GmbH (Continental AG)
• Inventors: Zobel, Matthias
• Granted Patent: US 8,953,842 B2
• Time in Patent Office: Days
• US Class Current: 382/103
• CPC Class Codes: G06V 10/421 by analysing segments intersecting the pattern
• CPC Class Codes: G06V 20/582 of traffic signs

Abstract

The invention relates to a method and to a device for the recognition of road signs (1). According to an inventive method for road sign recognition, images of a region surrounding the vehicle are acquired by means of a camera. The image data of the camera are analyzed, wherein at least one image region (10) that contains a potential road sign is determined. The image region is submitted to a first classification unit that identifies the potential road sign, on the basis of at least one class-specific feature, as a road sign (1) belonging to this particular class. After that, the brightness of the or the color intensity of the surface of the road sign (1) is analyzed along radially extending scanning beams (5) starting from a defined starting point. Potential contour points (6) of an information-bearing part (3) of the road sign (1) are determined from the results of brightness or color intensity analysis. A region of the surface of the road sign (1) is extracted as an information-bearing part (3) of the road sign (1) that includes all determined potential contour points (6). The extracted region is submitted to a second classification unit for semantic interpretation. The inventive method stands out due to a fast performance of the method on account of the brightness or color intensity analysis performed along scanning beams (5) that makes considerably lower demands on the control unit of a vehicle than two-dimensional edge detection operators (e.g., Sobel or Canny operators) that are usually employed for road sign recognition.

Claims

1. A method for road sign recognition, in which images of a region surrounding the vehicle are acquired by means of a camera,

the image data of the camera are analyzed, wherein at least one image region (10) that contains a potential road sign (1) is determined,

the image region (10) that contains the potential road sign (1) is submitted to a first classification unit that identifies the potential road sign (1), on the basis of at least one class-specific feature, as a road sign (1) belonging to this particular class, characterized in that

the brightness of the or the color intensity of the surface of the road sign (1) is analyzed along radially extending scanning beams (5) starting from a defined starting point within the surface of the road sign (1),

potential contour points (6) of an information-bearing part (3) of the road sign (1) are determined from the results of brightness or color intensity analysis,

a region of the surface of the road sign (1) is extracted as an information-bearing part (3) of the road sign that includes all determined potential contour points (6), and

the extracted region is submitted to a second classification unit for semantic interpretation.

2. The method according to claim 1, characterized in that the first classification unit defines the defined starting point for brightness analysis or color intensity analysis.

3. The method according to claim 1, characterized in that the center (8) of the surface of the road sign (1) is determined and used as a defined starting point.

4. The method according to claim 1, characterized in that brightness analysis or color intensity analysis comprises a one-dimensional gradient formation of the brightness values of or color intensity values of the pixels along the scanning beams (5).

5. The method according to claim 1, characterized in that the first classification unit determines the inner surface of the road sign (1) of a particular class and the brightness or color intensity analysis along the scanning beams (5) is only performed within the inner surface of the road sign (1).

6. The method according to claim 5, characterized in that that region of a bright-background road sign (1) for which the potential contour points (6) were determined as dark-to-bright transitions during brightness analysis performed along the scanning beams (5) from the starting point outward is extracted as an information-bearing part (3).

7. The method according to claim 5, characterized in that that region of a dark-background road sign (1) for which the potential contour points (6) were determined as bright-to-dark transitions during brightness analysis performed along the scanning beams (5) from the starting point outward is extracted as an information-bearing part (3).

8. The method according to claim 1, characterized in that the smallest rectangle (4) that includes all potential contour points (6) is that region of the road sign (1) which is extracted as an information-bearing part (3).

9. The method according to claim 1, characterized in that the surface of the road sign (1) is subjected to

a) brightness normalization and/or

b) noise suppression and/or

c) masking the region that surrounds the surface

prior to brightness or color intensity analysis.

10. The method according to claim 1, characterized in that the brightness or color intensity analysis of the surface of the road sign (1) is performed along a particular number of radially extending scanning beams (5), wherein the number of radially extending scanning beams (5) depends on the determined number of image regions with potential road signs (1) in a camera image or on the number of the road signs (1) identified in a camera image by the first classification unit.

11. The method according to claim 1, characterized in that

the brightness or color intensity analysis of the surface of the road sign (1) is performed along a predetermined minimum number of radially extending scanning beams (5),

the region extracted on the basis of brightness analysis or color intensity analysis is submitted to the second classification unit for semantic interpretation, and

if the second classification unit cannot classify the extracted region successfully, the number of scanning beams (5) is increased before brightness or color intensity analysis and extraction of a region are performed again.

12. The method according to claim 11, characterized in that the number of scanning beams (5) is only increased up to a defined maximum number.

13. The method according to claim 1, characterized in that all angles between adjacent scanning beams (5) have the same size.

14. The method according to claim 1, characterized in that twenty-four scanning beams (5) radially extending from the starting point outward are provided, wherein all angles between adjacent scanning beams (5) have a size of 15°

.

15. A device for performing the method for road sign recognition according to claim 1, said device comprising a camera for acquiring images of a region surrounding the vehicle and an image processing unit with a first classification unit that comprises means for determining, the information-bearing part (3) of the road sign (1) from the image data of the camera, and for forwarding the extracted region of the road sign (1) that comprises the information-bearing part (3) to a second classification unit for semantic interpretation.