Vision method of detecting lane boundaries and obstacles

• US 4,970,653 A
• Filed: 04/06/1989
• Issued: 11/13/1990
• Est. Priority Date: 04/06/1989
• Status: Expired due to Term

• Current Assignee: Delphi Technologies, Inc. (Aptiv PLC)
• Sponsoring Entity: General Motors Corporation (Motors Liquidation Co. GUC Trust)
• Inventors: Kenue, Surender K.
• Primary Examiner(s): Chin, Gary
• Time in patent office: 586 Days
• Field of Search: 364/424.02, 364/461, 364/424.02, 180/167, 180/168, 180/169, 318/587, 318/103
• US Class Current: 701/301
• CPC Class Codes: G05D 1/0246 using a video camera in combination with image processing means
• CPC Class Codes: G06T 7/00 Image analysis

Abstract

An image processing method operates on an image from a ccd camera viewing a roadway scene in front of a vehicle to detect lane markers and determine the relationship of the vehicle to the lane. Obstacles in the lane near the vehicle are detected and a warning is given to the driver. The method uses template matching techniques or a Hough algorithm to detect the lane markers or road edges.

Claims

1. In an automotive vehicle having a computer vision system and an associated camera for viewing a scene ahead of the vehicle, a method of detecting lane markers on a roadway comprising the steps of:

obtaining an image of the scene and digitizing the image,

normalizing the image,

defining a search area in the image,

searching for lane markers in the search area of the image to locate lane marker positions,

estimating the position of any missing lane marker from said located lane marker positions, and

defining lane boundaries from the said located lane marker positions and the estimated position of the missing lane marker.

2. The invention as defined in claim 1 wherein the step of estimating the position of missing lane markers comprises averaging the corresponding marker positions from previous images.

5. The invention as defined in claim 2 including the steps of:

checking the validity of a current lane boundary position by comparing a current lane boundary angle and lane width with that of the previous image wherein a discrepancy of said current lane boundary angle or said current lane width greater than threshold values reveals an invalid boundary, and

when said current lane boundary position is invalid, substituting a preceding boundary position.

6. The invention as defined in claim 2 wherein the step of searching for lane markers comprises the steps of:

generating edge data from the image,

correlating a template with the image in the search area, and

if insufficient match points are found to define a marker then correlating said template with the edge data in the search area.

7. The invention as defined in claim 2 wherein the steps of defining a search area and searching for lane markers comprise the steps of:

generating edge data from the image,

performing a boundary tracing operation on the edge data to generate a boundary traced image,

performing a line Hough transform on the boundary traced image to yield pixels in Hough space,

defining a search area in Hough space based on lane boundaries defined in a previous frame, and

searching for lane markers in Hough space to locate said lane marker positions.

8. The invention as defined in claim 7 wherein the step of search for lane markers in Hough space comprises the steps of:

reducing the number of said pixels in Hough space by filtering said pixels in Hough space with a non-maximum suppression filter,

deleting pixels in Hough space representing vertical and horizontal lines,

removing pixels in Hough space closer than a set amount, and

selecting pixels in Hough space with the most contributing edge elements as said lane markers.

3. In an automotive vehicle having a computer vision system and an associated camera for viewing a scene ahead of the vehicle and obtaining a series of images of the scene, a method of locating lane boundaries on a roadway comprising for each image the steps of:

digitizing the image and normalizing the image,

dynamically defining a search area in the image based on the lane boundaries located in a previous image,

searching for lane markers in the search area of the image to locate lane marker positions,

estimating the position of any missing lane marker based on information selected from the current image and previous images, and

defining lane boundaries from the said located lane marker positions and the estimated position of the missing lane marker.

4. The invention as defined in claim 3 wherein the step of estimating the position of missing lane markers comprises the steps of:

determining the lane width from said previous image, and

calculating the estimated position of a missing lane marker from the position of a marker on the opposite side of the lane and the lane width.

9. In an automotive vehicle having a computer vision system and an associated camera for viewing a scene ahead of the vehicle and obtaining a series of images of the scene, a method of locating lane boundaries and close obstacles on a roadway comprising for each image the steps of:

digitizing the image and normalizing the image,

generating edge data from the image,

determining an edge pixel count by counting edge pixels between lane boundaries defined in the previous image ad in the image portion representing space close to the vehicle,

comparing the edge pixel count to a threshold value and issuing an obstacle warning when the count is above the threshold value,

when the edge pixel count is below the threshold value, dynamically defining a search area in the image based on the lane boundaries located in said previous image,

searching for lane markers in the search area of the image to locate lane marker positions,

estimating the position of any missing lane marker based on information selected from the current image and previous images, and

defining lane boundaries from the said located lane marker positions and the estimated position of the missing lane marker.

10. The invention as defined in claim 9 including the step of:

before searching for said lane markers, effectively deleting obstacles from far field portions of the image by removing vertical and horizontal lines within the said search area in the image.