Corner point extraction system and method for image guided stereo camera optical axes alignment
A method of aligning optical axes of cameras for a non-transitory computer readable storage medium storing one or more programs is disclosed. The one or more programs comprise instructions, which when executed by a computing device, cause the computing device to perform by one or more autonomous vehicle driving modules execution of processing of images using the following steps comprising: calibrating intrinsic parameters of a set of cameras; extracting corner points associated with a pattern; and computing a vanishing point based on information on the extracted corner points.
1. A method of aligning cameras used for autonomous driving, the method comprising:
determining intrinsic parameters for a pair of cameras including a first camera and a second camera, wherein the determining the intrinsic parameters comprises obtaining a single image of multiple patterns, disposed at different orientations in a scene, from each of the pair of cameras;
extracting four corner points from a pair of images of a checkerboard pattern taken by the pair of cameras;
determining a vanishing point in each of the pair of images in pixel coordinates, based on the extracted corner points, from the intersection of two lines constructed by connecting corner points in pairs on a same side of the checkerboard pattern in the pair of images;
transforming the pixel coordinates of the vanishing point into feature points in image coordinates using the intrinsic parameters; and
aligning the pair of images by aligning the feature points in the pair of images.
2. The method according to claim 1 , further comprising:
adjusting the first camera and the second camera to align their optical axes to be parallel to each other based on the vanishing point in each of the pair of images.
3. The method according to claim 1 , wherein the determining the intrinsic parameters comprises:
disposing another pattern at different orientations in the scene; and
obtaining another image of the other pattern at each of the different orientations from each of the pair of cameras.
4. The method according to claim 1 , wherein the determining the intrinsic parameters comprises:
obtaining an image of a planar pattern including corners;
recovering from the image the corners of the planar pattern and assigning 3D positions to the recovered corners; and
calibrating the first camera and the second camera based on correspondences of corner locations between a 2D pixel coordinate and the 3D positions.
5. The method according to claim 1 , wherein the determining the vanishing point in each image comprises:
constructing a pair of parallel lines based on 2D positions of the extracted corner points.
6. A system for aligning cameras in an autonomous driving system, the system comprising:
a memory;
one or more processing units; and
one or more programs stored in the memory configured for execution by the one or more processing units, the one or more programs including instructions to at least:
determine intrinsic parameters for a pair of cameras including a first camera and a second camera, wherein the determining the intrinsic parameters comprises obtaining a single image of multiple patterns, disposed at different orientations in a scene, from each of the pair of cameras;
extract four corner points from a pair of images of a checkerboard pattern taken by the pair of cameras;
determine a vanishing point in each of the pair of images in pixel coordinates, based on the extracted corner points, from the intersection of two lines constructed by connecting corner points in pairs on a same side of the checkerboard pattern in the pair of images;
transform the pixel coordinates of the vanishing point into feature points in image coordinates using the intrinsic parameters; and
align the pair of images by aligning the feature points in the pair of images.
7. The system according to claim 6 , further comprising:
adjusting the pair of cameras to align their optical axes to be parallel to each other based on the vanishing point in each image.
8. The system according to claim 6 , wherein the determining the intrinsic parameters comprises:
disposing another pattern at different orientations in the scene; and
obtaining another image of the other pattern at each of the different orientations.
9. The system according to claim 6 , wherein the determining the intrinsic parameters comprises:
obtaining an image of a planar pattern including corners;
recovering from the image the corners of the planar pattern and assigning 3D positions to the recovered corners;
calibrating the first camera and the second camera based on correspondences of corner locations between a 2D pixel coordinate and the 3D positions; and
determining a first set of intrinsic parameters for a first camera and a second set of intrinsic parameters for a second camera, wherein a first set of intrinsic parameters for the first camera link first pixel coordinates of a first image point to corresponding image coordinates in a first camera reference frame and a second set of intrinsic parameters for the second camera link second pixel coordinates of a second image point to corresponding image coordinates in a second camera reference frame.
10. The system according to claim 6 , wherein the determining the vanishing point in each image comprises:
constructing a pair of parallel lines including a first line and a second line based on 2D positions of the extracted corner points, wherein the vanishing point in each of the pair of images is determined from the intersection of a first line generated by connecting the corner points on a first side of the pattern and a second line generated by connecting corner points on a second side of the pattern.
11. The method according to claim 1 , wherein the vanishing point in each of the pair of images is determined from the intersection of a first line generated by connecting the corner points on a first side of the pattern and a second line generated by connecting corner points on a second side of the pattern.
12. The method according to claim 1 , wherein a first set of intrinsic parameters for the first camera link first pixel coordinates of a first image point to corresponding image coordinates in a first camera reference frame and a second set of intrinsic parameters for the second camera link second pixel coordinates of a second image point to corresponding image coordinates in a second camera reference frame.
13. The method according to claim 1 , wherein the feature points further include a plane, a straight line, and a sparse point in three dimensions.
14. The system according to claim 6 , wherein the feature points further include one or more of a plane, a straight line, or a sparse point in three dimensions.
15. A non-transitory computer readable information recording medium including executable instructions that when executed by at least one processor perform operations comprising:
determining intrinsic parameters for a pair of cameras including a first camera and a second camera, wherein the determining the intrinsic parameters comprises obtaining a single image of multiple patterns, disposed at different orientations in a scene, from each of the pair of cameras;
extracting corner points from a pair of images of a checkerboard pattern taken by the pair of cameras;
determining a vanishing point in each of the pair of images in pixel coordinates, based on the extracted corner points, from the intersection of two lines constructed by connecting corner points in pairs on a same side of the checkerboard pattern in the pair of images;
transforming the pixel coordinates of the vanishing point into feature points in image coordinates using the intrinsic parameters; and
aligning the pair of images by aligning the feature points in the pair of images.
16. The non-transitory computer readable information recording medium according to claim 15 , further comprising:
adjusting the first camera and the second camera to align their optical axes to be parallel to each other based on the vanishing point.
17. The non-transitory computer readable information recording medium according to claim 15 , wherein the feature points further include a plane, a straight line, and a sparse point in three dimensions.
18. The non-transitory computer readable information recording medium according to claim 15 , wherein the determining the intrinsic parameters comprises:
recovering from the image corners of the one or more patterns;
assigning 3D positions to the recovered corners; and
calibrating the pair of cameras based on correspondences of corner locations between a 2D pixel coordinate and a 3D position.