PIXEL-BASED TEXTURE-RICH CLEAR PATH DETECTION

• US 20090268948A1
• Filed: 05/29/2009
• Published: 10/29/2009
• Est. Priority Date: 04/24/2008
• Status: Active Grant

• Current Assignee: GM Global Technology Operations LLC (General Motors Company)
• Original Assignee: GM Global Technology Operations Incorporated (Motors Liquidation Co. GUC Trust)
• Inventors: Naik, Sanjeev M., Zhang, Wende
• Granted Patent: US 8,452,053 B2
• Time in Patent Office: Days
• US Class Current: 382/104
• CPC Class Codes: G06V 10/757 Matching configurations of points or features
• CPC Class Codes: G06V 20/588 Recognition of the road, e.g. of lane markings; Recognition of the vehicle driving pattern in relation to the road
• CPC Class Codes: G08G 1/165 for passive traffic, e.g. including static obstacles, trees
• CPC Class Codes: G08G 1/166 for active traffic, e.g. moving vehicles, pedestrians, bikes

Abstract

A method for detecting a clear path of travel for a vehicle utilizing analysis of a plurality of images generated by a camera device located upon the vehicle includes monitoring the images wherein each image comprises a plurality of pixels, identifying a set of interest points from the plurality of pixels in a current image, finding their corresponding points in a preceding image through correspondence matching, filtering the matched pairs of interest points to select a preferential set of matched pairs, generating a three dimensional map of features in the view based upon the preferential set of matched pairs, determining the clear path based upon the mapped features, and utilizing the clear path to operate the vehicle.

Claims

1. Method for detecting a clear path of travel for a vehicle utilizing analysis of a plurality of images of a view generated by a camera device located upon the vehicle, the method comprising:

monitoring the images, each image comprising a plurality of pixels;

identifying a set of interest points from the plurality of pixels in each of two images, the two images comprising a current image and a preceding image;

matching pairs of interest points from the set of interest points from the current image and the set of interest points from the preceding image, wherein each of the matched pairs is estimated to correspond to a single viewed feature;

filtering the matched pairs of interest points to select a preferential set of matched pairs;

generating a three dimensional map of features in the view based upon the preferential set of matched pairs; and

determining the clear path based upon the mapped features.

2. The method of claim 1, wherein identifying a set of interest points comprises filtering each of the images with interest point extraction.

3. The method of claim 2, wherein filtering each of the images with interest point extraction comprises applying a scale-invariant feature transform.

4. The method of claim 2, wherein filtering each of the images with interest point extraction comprises applying corner detection.

5. The method of claim 2, wherein filtering each of the images with interest point extraction comprises applying a Sobel filter.

6. The method of claim 1, wherein matching pairs of interest points from the set of interest points from the current image and the set of interest points from the preceding image comprises:

using correspondence matching to match interest points from the current image and the previous image.

7. The method of claim 1, wherein matching pairs of interest points from the set of interest points from the current image and the set of interest points from the preceding image comprises:

using template matching to match interest points from the current image and the previous image.

8. The method of claim 1, wherein filtering the matched pairs of interest points to select the preferential set of matched pairs comprises filtering out poorly matched pairs.

9. The method of claim 1, wherein filtering the matched pairs of interest points to select the preferential set of matched pairs comprises analyzing the matched pairs with quality control criteria.

10. The method of claim 9, wherein analyzing the matched pairs with the quality control criteria comprises:

determining if a distance between interest points of a particular matched pair of interest points is less than a threshold distance; and

including the particular matched pair of interest points in the preferential set of matched pairs if the distance between the interest points is less than the threshold distance.

11. The method of claim 10, wherein the threshold distance is based upon a speed of the vehicle.

12. The method of claim 9, wherein analyzing the matched pairs with the quality control criteria comprises:

determining if the interest points of a particular matched pair are within a predetermined threshold distance from an image boundary; and

including the particular matched pair of interest points in the preferential set of matched pairs if either of the interest points of the particular matched pair are within the predetermined threshold distance from the image boundary.

13. The method of claim 9, wherein analyzing the matched pairs with the quality control criteria comprises:

defining an area around an interest point of a particular matched pair;

defining an equivalent area around the other interest point of the particular matched pair; and

determining if a difference in color between the area and the equivalent area is less than a threshold difference in color.

14. The method of claim 1, wherein generating the three dimensional map of the features in the view based upon the preferential set of matched pairs comprises:

mapping each of the features in the view corresponding to a pair of interest points within the preferential set of matched pairs based upon a distance a vehicle traveled between the generation of the images containing the pair of interest points within the preferential set of matched pairs, vehicle yaw, and elapsed time between the generation of the images containing the pair of interest points within the preferential set of matched pairs.

15. The method of claim 1, further comprising:

classifying the features based upon a height above a ground level.

16. The method of claim 15, wherein determining the clear path based upon the mapped features is based upon the classifications of the features.

17. Method for detecting a clear path of travel for a vehicle utilizing analysis of a plurality of images generated by a camera device located upon the vehicle, the method comprising:

monitoring the images, each image comprising a plurality of pixels;

utilizing a texture rich processing scheme to analyze the images comprising

identifying a set of interest points from the plurality of pixels in each of two images, the two images comprising a current image and a preceding image,

matching pairs of interest points from the set of interest points from the current image and the set of interest points from the preceding image, wherein each of the matched pairs is estimated to correspond to a single viewed feature,

filtering the matched pairs of interest points to select a high quality set of matched pairs, and

generating a three dimensional map of features in the view based upon the high quality set of matched pairs; and

utilizing a texture-less processing scheme to analyze the images comprising

in each image, filtering the image to separate portions of the image indicating a clear surface upon which a potential road surface can be estimated from other portions of the image that do not indicate a potential road surface,

determining a clear path based upon the texture rich processing scheme and the texture-less processing scheme, and

utilizing the clear path to operate the vehicle.

18. The method of claim 17, wherein filtering the image to separate the portions of the image indicating the clear surface upon which the potential road surface can be estimated from other portions of the image that do not indicate a potential road surface comprises:

filtering an area below a horizon line on the image from an area above the horizon line.

19. The method of claim 17, wherein filtering the image to separate the portions of the image indicating the clear surface upon which the potential road surface can be estimated from other portions of the image that do not indicate a potential road surface comprises:

filtering pixels in the image based upon comparing a color intensity of the pixels to a color intensity threshold based upon a color distribution mean.

20. The method of claim 17, wherein filtering the image to separate the portions of the image indicating the clear surface upon which the potential road surface can be estimated from other portions of the image that do not indicate a potential road surface comprises:

analyzing based upon differencing sequential images.

21. The method of claim 17, wherein filtering the image to separate the portions of the image indicating the clear surface upon which the potential road surface can be estimated from other portions of the image that do not indicate a potential road surface comprises:

filtering based upon identifying pixels representing edges.

22. Apparatus for detecting a clear path of travel for a vehicle utilizing analysis of a plurality of images generated by a camera device located upon the vehicle, the apparatus comprising:

the camera device; and

a control module monitoring the images from the camera device and identifying a set of interest points from a plurality of pixels in each of two images from the camera device, the two images comprising a current image and a preceding image,

matching pairs of interest points from the set of interest points from the current image and the set of interest points from the preceding image, wherein each of the matched pairs is estimated to correspond to a single viewed feature

filtering the matched pairs of interest points to select a preferential set of matched pairs,

generating a three dimensional map of features in the view based upon the preferential set of matched pairs, and

determining the clear path based upon the mapped features.