Traffic sign detecting method and traffic sign detecting device

• US 8,983,136 B2
• Filed: 09/10/2012
• Issued: 03/17/2015
• Est. Priority Date: 09/23/2011
• Status: Active Grant

• Current Assignee: Ricoh Company Limited
• Original Assignee: Ricoh Company Limited
• Inventors: Liu, Liyan, Wang, Xiaomeng, Pan, Pingping
• Primary Examiner(s): Drennan, Barry
• Assistant Examiner(s): Harandi, Siamak
• Application Number:
  US13/608,401
• Publication Number:
  US 20130077830A1
• Time in Patent Office: 918 Days
• Field of Search: None
• US Class Current: 382/104
• CPC Class Codes: G06V 20/582 of traffic signs

Abstract

Disclosed are a method and a device for detecting traffic signs in an input image camera. The method comprises a color space converting step of converting the input image into a HSV color space image; a filtering step of filtering, based on a predetermined pass range of a standard color of each of the traffic signs, the HSV color space image to obtain a filtered image, and then generating one or more connected domains based on one or more regions in the filtered image; a removing step of removing, based on a standard rule of the corresponding traffic sign, at least one of the generated connected domains, not being the corresponding traffic sign, and letting others of the generated connected domains be candidate traffic sign domains; and a recognition step of recognizing, based on a feature of each of the candidate traffic sign domains, the corresponding traffic sign.

Claims

1. A method of detecting one or more traffic signs in an input image captured by a camera, comprising:

a color space converting step of converting the input image into a HSV color space image;

a filtering step including,

filtering, based on a predetermined pass range of a standard color of each of the traffic signs, the HSV color space image so as to obtain a filtered image,

generating one or more connected domains, the connected domains being regions of the filtered image that correspond to the standard color, and

adding the connected domains onto the input image;

a removing step of removing, based on a standard rule of the corresponding traffic signs, one or more of the generated connected domains, not being the corresponding traffic sign; and

a recognition step of recognizing, based on a feature of each of candidate traffic sign domains, the corresponding traffic sign,

the candidate traffic sign domains being the domains, from among the connected domains, that are not removed in the removing step.

2. The method according to claim 1, further comprising:

a traffic sign feature tree creating step of creating a traffic sign feature tree by

classifying standard traffic signs into large classes according to color features of outer outlines of the standard traffic signs;

classifying, for each of the large classes, members of the large class into small classes according to shape features of the outer outlines of the members of the large class; and

classifying, for each of the small classes, members of the small class into subclasses according to shape features of inner outlines of members of the small class.

3. The method according to claim 2, wherein, the recognition step comprises:

a large class determining step of determining, based on the standard color of the corresponding traffic sign of the corresponding candidate traffic sign domain, a large class to which the corresponding candidate traffic sign domain belongs;

a small class determining step of determining, based on an outer outline shape feature of the corresponding candidate traffic sign domain, a small class to which the corresponding candidate traffic sign domain belongs, by utilizing a pre-trained classifier;

a subclass determining step of determining, based on the number of inner outlines of the corresponding candidate traffic sign domain, a subclass to which the corresponding candidate traffic sign domain belongs; and

a traffic sign matching step of determining, by performing matching between the corresponding candidate traffic sign domain and each of the standard traffic signs in the determined subclass, a standard traffic sign in the determined subclass, best matched with the corresponding candidate traffic sign domain as the corresponding traffic sign.

4. The method according to claim 3, wherein, the traffic sign matching step comprises:

calculating a feature distance between an inner outline feature of the corresponding candidate traffic sign domain and an inner outline feature of each of the standard traffic signs in the determined subclass, and

determining a standard traffic sign corresponding to the minimum of the feature distances as best matched with the corresponding candidate traffic sign domain.

5. The method according to claim 3, further comprising:

carrying out, using the pre-trained classifier, training by using traffic sign samples based on rotation and scale-invariance.

6. The method according to claim 1, further comprising:

a distance calculating step of, after recognizing the corresponding traffic sign, calculating an actual distance between the camera and the corresponding traffic sign based on a focal length of the camera, a size of the corresponding traffic sign on an imaging sensor of the camera, and an actual size of the corresponding traffic sign.

7. The method according to claim 6, further comprising:

a notifying step of notifying a driver of the corresponding traffic sign in a way determined according to both a meaning of the corresponding traffic sign and the calculated actual distance between the corresponding traffic sign and the camera.

8. The method according to claim 1, wherein the filtering further comprises:

filtering each of pixels of the HSV color space image according to a predetermined hue pass range and a predetermined saturation pass range of the standard color of the corresponding traffic sign; and

retaining the corresponding pixel only when a hue value and a saturation value of the corresponding pixel are within the predetermined hue pass range and the predetermined saturation pass range, respectively.

9. The method according to claim 1, wherein, in the removing step comprises:

removing at least one of the generated connected domains, not being the corresponding traffic sign according to one or more of traffic sign standard rules, the traffic sign standard rules including,

(a) an aspect ratio of a connected domain;

(b) a position of a connected domain in an image waiting for processing;

(c) an area value of a connected domain;

(d) a color pattern of a connected domain;

(e) a shape complexity of a connected domain; and

(f) a texture complexity of a connected domain.

10. A device for detecting one or more traffic signs in an input image captured by a camera, comprising:

a processor, and

a storage device, the storage device storing instructions that, when executed by the processor, cause the processor to perform operations including,

converting the input image into a HSV color space image;

filtering, based on a predetermined pass range of a standard color of each of the traffic signs, the HSV color space image so as to obtain a filtered image,

generating one or more connected domains, the connected domains being regions of the filtered image that correspond to the standard color, and

adding the connected domains onto the input image;

removing, based on a standard rule of the corresponding traffic sign, one or more of the generated connected domains, not being the corresponding traffic sign; and

recognizing, based on a feature of each of the candidate traffic sign domains, the corresponding traffic sign.