Systems and methods for detecting traffic signs

• US 9,195,895 B1
• Filed: 05/14/2015
• Issued: 11/24/2015
• Est. Priority Date: 05/14/2014
• Status: Active Grant

• Current Assignee: MobilEye Vision Technologies Ltd. (Intel Corporation)
• Original Assignee: MobilEye Vision Technologies Ltd. (Intel Corporation)
• Inventors: Taieb, Yoav, Kapach, Yair, Krupnik, Yoel
• Primary Examiner(s): Fitzpatrick, Atiba O
• Application Number:
  US14/712,384
• Publication Number:
  US 20150332104A1
• Time in Patent Office: 194 Days
• Field of Search: None
• US Class Current: 1/1
• CPC Class Codes: G06F 18/22 Matching criteria, e.g. proximity measures
• CPC Class Codes: G06V 10/42 Global feature extraction by analysis of the whole pattern, e.g. using frequency domain transformations or autocorrelation
• CPC Class Codes: G06V 10/751 Comparing pixel values or logical combinations thereof, or feature values having positional relevance, e.g. template matching
• CPC Class Codes: G06V 20/582 of traffic signs

Abstract

Systems and methods are provided for detecting traffic signs. In one implementation, a traffic sign detection system for a vehicle include at least one image capture device configured to acquire at least one image of a scene including a traffic sign ahead of the vehicle. The traffic sign detection system also includes a data interface and at least one processing device programmed to receive the at least one image via the data interface, transform the at least one image, sample the transformed at least one image to generate a plurality of images having different sizes, convolve each of the plurality of images with a template image, compare each pixel value of each convolved image to a predetermined threshold, and select local maxima of pixel values within local regions of each convolved image as attention candidates, the local maxima being greater than the predetermined threshold.

Claims

1. A traffic sign detection system for a vehicle, comprising:

at least one image capture device configured to acquire at least one image of an environment including a traffic sign ahead of the vehicle;

a data interface; and

at least one processing device programmed to;

receive the at least one image via the data interface;

transform the at least one image;

sample the transformed at least one image to generate a plurality of images having different sizes;

convolve the plurality of images with a template image;

compare one or more pixel values of the convolved images to a predetermined threshold; and

select local maxima of pixel values within local regions of the convolved images as attention candidates, the local maxima being greater than the predetermined threshold.

2. The traffic sign detection system of claim 1, further comprising detecting the traffic sign based on the attention candidates.

3. The traffic sign detection system of claim 1, wherein the traffic sign includes at least one of a triangular and rhombus shape.

4. The traffic sign detection system of claim 1, wherein the at least one processing device is programmed to transform the at least one image by computing a distance between an angle formed by an edge of the traffic sign and a predetermined angle.

5. The traffic sign detection system of claim 4, wherein the predetermined angle is 60 degrees for detecting a triangular shape.

6. The traffic sign detection system of claim 4, wherein the predetermined angle is 45 degrees for detecting a rhombus shape.

7. The traffic sign detection system of claim 4, wherein the at least one processing device is programmed to transform the at least one image by computing the distance based on a look up table.

8. The traffic sign detection system of claim 4, wherein the angle formed by the edge of the traffic sign is computed from horizontal and vertical gradients of each image point on the edge.

9. The traffic sign detection system of claim 8, wherein the horizontal and vertical gradients of the edge are computed using a nonlinear filter.

10. The traffic sign detection system of claim 9, wherein the nonlinear filter is a Sobel filter.

11. The traffic sign detection system of claim 1, wherein the at least one processing device is programmed to crop the at least one image to reduce its size prior to transforming the at least one image.

12. The traffic sign detection system of claim 1, wherein the template image has a fixed size.

13. The traffic sign detection system of claim 1, wherein the different sizes of the plurality of images vary linearly.

14. The traffic sign detection system of claim 1, wherein the different sizes of the plurality of image vary logarithmically.

15. The traffic sign detection system of claim 1, wherein each of the local regions has a predetermined size smaller than a size of each convolved image.

16. The traffic sign detection system of claim 15, wherein the predetermined size for each of the local regions is 5-pixel by 5-pixel.

17. The traffic sign detection system of claim 1, wherein the at least one processing device is programmed to apply a non-maximum suppression process to the convolved image after it is compared to the threshold.

18. The traffic sign detection system of claim 17, wherein the non-maximum suppression is applied within each of the local regions.

19. A vehicle, comprising:

a body;

at least one image capture device mounted on the body and configured to acquire at least one image of an environment including a traffic sign ahead of the vehicle;

a data interface; and

at least one processing device programmed to;

receive the at least one image via the data interface;

transform the at least one image;

sample the transformed at least one image to generate a plurality of images having different sizes;

convolve the plurality of images with a template image;

compare one or more pixel values of the convolved images to a predetermined threshold; and

select local maxima of pixel values within local regions of the convolved image as attention candidates, the local maxima being greater than the predetermined threshold.

20. The vehicle of claim 19, wherein the at least one processing device is further programmed to detect the traffic sign based on the attention candidates.

21. The vehicle of claim 19, wherein the traffic sign includes at least one of a triangular and rhombus shape.

22. The vehicle of claim 19, wherein the at least one processing device is programmed to transform the at least one image by computing a distance between an angle formed by an edge of the traffic sign and a predetermined angle.

23. The vehicle of claim 22, wherein the angle formed by the edge of the traffic sign is computed from horizontal and vertical gradients of each image point on the edge.

24. The vehicle of claim 22, wherein the at least one processing device is programmed to transform the at least one image by computing the distance based on a look up table.

25. The vehicle of claim 19, wherein the template image has a fixed size.

26. The vehicle of claim 19, wherein the different sizes of the plurality of images vary linearly.

27. The vehicle of claim 19, wherein the different sizes of the plurality of images vary logarithmically.

28. The vehicle of claim 19, wherein the at least one processing device is programmed to apply a non-maximum suppression process to the convolved image after it is compared to the threshold.

29. A method for detecting a traffic sign for a vehicle, comprising:

acquiring, via at least one image capture device, at least one image of an environment including a traffic sign ahead of the vehicle;

receiving, via a processing unit, the at least one image;

transforming, via the processing unit, the at least one image;

sampling, via the processing unit, the transformed at least one image to generate a plurality of images having different sizes;

convolving, via the processing unit, the plurality of images with a template image;

comparing, via the processing unit, one or more pixel values of the convolved image to a predetermined threshold; and

selecting local maxima of pixel values within local regions of the convolved image as attention candidates, the local maxima being greater than the predetermined threshold.

30. The method of claim 29, further comprising detecting the traffic sign based on the attention candidates.