Method and system for detecting traffic lights

1. A method for detecting traffic lights, comprising:

obtaining a color image captured by a camera;

converting the color image into a first monochrome scale image

converting the first monochrome scale image into a first binary image wherein the color image is converted into the first monochrome scale image based on a first threshold of brightness;

identifying a first set of candidate blobs in the first binary image based on at least three predetermined geometric parameters, wherein the at least three predetermined geometric parameters comprise a range of pixel number, a width range, and a range of width to height ratio;

determining whether a first region in the color image, which corresponds to one of the first set of candidate blobs, represents a green traffic light using a green traffic light classifier;

converting the color image into a second binary image, wherein the color image is converted into the second binary image based on a second threshold, the second threshold being determined based on camera parameters including at least one of a color gamut and a color saturation;

identifying a second set of candidate blobs in the second binary image; and

determining whether a second region in the color image, which corresponds to one of the second set of candidate blobs, represents a red traffic light using a red traffic light classifier.

2. The method according to claim 1 , wherein the first monochrome scale image is a gray scale image, and wherein the range of pixel number and the width range are based on a resolution and a focal length of the camera.

3. The method according to claim 2 , wherein the color image is converted into the gray scale image based on green channel value, wherein the range of pixel number is between 5 and 300 pixels, and wherein the width range is between 5 and 25 pixels.

4. The method according to claim 1 further comprising applying a white top-hat morphology operator on the first monochrome scale image to obtain a second monochrome scale image, and converting the second monochrome scale image into the first binary image, and wherein the range of width to height ratio is between 0.8 and 1.25.

5. The method according to claim 4 , wherein the green traffic light classifier is based on one of support vector machine or cascade detection.

6. The method according to claim 5 , wherein the first threshold of brightness is from about 70% to about 80% of a mean brightness of the second monochrome scale image.

7. The method according to claim 1 , wherein the green traffic light classifier is suitable for monochrome scale images, and it is applied on the first monochrome scale image to determine whether the first region represents a green traffic light.

8. The method according to claim 1 , where the color image is converted into the second binary image based on equation (1):

Q= 2 R−G−B   Equation (1)

where R, G, and B respectively represent red channel value, green channel value, and blue channel value of a pixel, if the Q value of the pixel is greater than a second threshold value, the pixel will be set as valid, otherwise the pixel will be set as invalid.

9. The method according to claim 8 , wherein the red traffic light classifier is suitable for monochrome scale images, and it is applied on the first monochrome scale image to determine whether the second region represents a red traffic light.

10. The method according to claim 8 , wherein the second threshold value is from about 210 to about 230.

11. The method of claim 1 , further comprising removing a lower portion of the color image prior to converting the color image into the first monochrome scale image.

12. A system for detecting traffic lights, comprising:

a processing device, configured to:

obtain a color image captured by a camera;

convert the color image into a first monochrome scale image, wherein the color image is converted into the first monochrome scale image based on a first threshold of brightness;

convert the first monochrome scale image into a first binary image;

identify a first set of candidate blobs in the first binary image based on at least three predetermined geometric parameters, wherein the at least three predetermined geometric parameters comprises a range of pixel number, a width range, and a range of width to height ratio;

determine whether a first region in the color image, which corresponds to one of the first set of candidate blobs, represents a green traffic light using a green traffic light classifier;

convert the color image into a second binary image, wherein the color image is converted into the second binary image based on a second, different threshold, the second threshold being determined based on camera parameters including at least one of a color gamut and a color saturation;

identify a second set of candidate blobs in the second binary image; and

determine whether a second region in the color image, which corresponds to one of the second set of candidate blobs, represents a red traffic light using a red traffic light classifier.

13. The system according to claim 12 , wherein the first monochrome scale image is a gray scale image, and wherein the range of pixel number is 5 to 300 pixels.

14. The system according to claim 13 , wherein the color image is converted into the gray scale image based on green channel value, and wherein the width range is 5 to 25 pixels.

15. The system according to claim 12 , wherein the processing device is further configured to apply a white top-hat morphology operator on the first monochrome scale image to obtain a second monochrome scale image, and convert the second monochrome scale image into the first binary image, and wherein the range of width to height ratio is between 0.8 and 1.25.

16. The system according to claim 15 , wherein the range of pixel number and the width range are based on one or more properties of the camera.

17. The system according to claim 16 , wherein the first threshold of brightness is from about 70% to about 80% of a mean brightness of the second monochrome scale image, and wherein the green traffic light classifier comprises template matching, a neural network, a support vector machine, or a cascade detector.

18. The system according to claim 12 , wherein

the color image is converted into the second binary image based on equation (1):

Q= 2 R−G−B   Equation (1)

where R, G, and B respectively represent red channel value, green channel value, and blue channel value of a pixel, if the Q value of the pixel is greater than a second threshold value, the pixel will be set as valid, otherwise the pixel will be set as invalid.

19. The system according to claim 18 , wherein the second threshold value is from about 210 to about 230.

20. The system according to claim 18 , wherein the red traffic light classifier is suitable for monochrome scale images, and is applied on the first monochrome scale image to determine whether the second region represents a red traffic light.

21. The system of claim 12 , wherein the processing device is further configured to remove a lower portion of the color image prior to converting the color image into the first monochrome scale image.