Tracking road boundaries

Patent metadata

Field	Value
Publication number	US-2016379064-A1
Application number	US-201514754488-A
Country	US
Kind code	A1
Filing date	Jun 29, 2015
Priority date	Jun 29, 2015
Publication date	Dec 29, 2016
Grant date	—

Abstract

Systems and methods of tracking a road boundary are provided. According to one aspect, a method of tracking a road boundary may include capturing an image from a camera, identifying a pair of regions of interest (ROI) in the image on each side of a candidate boundary position, extracting a color profile from each of the ROIs, generating a weighted color difference score by comparing the color profiles and weighting a difference between the color profiles based on a color similarity between colors in the color profiles, and outputting a determination of a detected boundary based upon the weighted color difference score.

First claim

1 . A method of tracking a road boundary, comprising:
capturing an image from a camera; identifying a pair of regions of interest (ROI) in the image on each side of a candidate boundary position; extracting a color profile from each of the ROIs; generating a weighted color difference score by comparing the color profiles and weighting a difference between the color profiles based on a color similarity between colors in the color profiles; and outputting a determination of a detected boundary based upon the weighted color difference score.

2 . The method of claim 1 , wherein the camera is positioned onboard a vehicle.

3 . The method of claim 1 , wherein the camera is positioned onboard a robot.

4 . The method of claim 1 , further comprising:
extracting an edge feature score at the candidate boundary position from the image.

5 . The method of claim 4 , wherein the edge feature score is extracted based upon grayscale information of a part of the image that contains the pair of ROIs.

6 . The method of claim 1 , further comprising:
transforming the captured image using inverse-perspective image mapping prior to identifying the pair of regions of interest.

7 . The method of claim 1 , wherein extracting the color profile includes identifying clustered colors in a part of the image that contains the pair of ROIs.

8 . The method of claim 7 , wherein the image includes a plurality of segments, each segment having an associated luminance value and chrominance value, and wherein the colors are clustered by, for each segment in the plurality of segments, classifying the segment according to its chrominance value into one of a plurality of chrominance ranges.

9 . The method of claim 1 , wherein each color profile includes a respective color histogram which represents the clustered colors in each ROI of the pair of ROIs, and the weighted color difference score represents a dissimilarity in the respective color histograms.

10 . The method of claim 1 , wherein the weighted color difference score for the candidate boundary position and a plurality of other weighted color difference scores for other candidate boundary positions in the image are stored in a weighted color difference score vector, and the detected boundary is determined by identifying a peak in the weighted color difference score vector.

11 . The method of claim 10 , wherein the detected boundary is determined by filtering the weighted color difference score to remove values above a threshold prior to identifying the peak.

12 . The method of claim 1 , wherein the detected boundary is a first boundary at an edge of a path or roadway.

13 . The method of claim 1 , wherein the detected boundary is a first detected boundary, and the method further comprises:
identifying a second pair of regions of interest (ROI) in the image on each side of a second candidate boundary position; extracting a second color profile from each of the ROIs of the second pair; generating a second weighted color difference score by comparing the second color profiles and weighting a second difference between the second color profiles based on a second color similarity between colors in the second color profiles; and outputting a second determination of a second detected boundary based upon the second weighted color difference score.

14 . The method of claim 1 , wherein the image is a first image, the method further comprising:
capturing a second image from a camera, the second image being a subsequent frame in a video stream to the first image; transforming the second via inverse perspective mapping; performing color clustering on the second image; and determining a detected boundary in the second image based on a pair of regions of interest (ROIs) positioned on each side of a candidate boundary position in the second image, wherein the candidate boundary position in the second image is at least partially based upon a position of the first detected boundary.

15 . The method of claim 1 , wherein the detected boundary is further determined based upon a lateral offset of the camera position relative to its position at an earlier time instant.

16 . A method of tracking a road boundary, comprising:
capturing an image from a camera; transforming the captured image using inverse perspective image mapping; identifying a pair of regions of interest (ROI) in the image on each side of a candidate boundary position; extracting a color profile from each of the ROIs; generating a weighted color difference score by comparing a difference between the color profiles based on a color similarity between colors in the color profiles; and outputting a determination of a detected boundary based upon the weighted color difference score.

17 . A road boundary tracking system, comprising:
a camera for capturing an image; and a processor configured to:
extract an edge feature at a candidate boundary position from the image;
identify a pair of regions of interest (ROI) in the image on each side of the candidate boundary position;
extract a color profile from each of the ROIs;
generate a weighted color difference score by comparing the color profiles, and weighting a difference between the color profiles based on a color similarity between colors in the color profiles; and
output a determination of a detected boundary based upon the weighted color difference score and the extracted edge feature.

18 . The road boundary tracking system of claim 17 , wherein the processor extracts the color profile by identifying clustered colors in a part of the image that contains the pair of ROIs.

19 . The road boundary tracking system of claim 17 , wherein each color profile includes a respective color histogram which represents the clustered colors in each ROI of the pair of ROIs, and the weighted color difference score represents a dissimilarity in the respective color histograms.

20 . The road boundary tracking system of claim 17 , comprising:
an onboard computing system which includes the camera and the processor, and wherein the processor is further configured to execute a navigation module to receive the output and control travel of the vehicle based upon the output.

Assignees

• Sharp Laboratories America Inc

Inventors

• Van Beek Petrus J L

Classifications

• G06V10/56

  relating to colour · CPC title

• G06V10/443

  by matching or filtering · CPC title

• G06K9/4652

  Physics · mapped topic

• G06K9/00798 Primary

  Physics · mapped topic

• G06T2207/10004

  Still image; Photographic image · CPC title