Fiducial designs and pose estimation for augmented reality
Novel fiducial designs that are based on a plane circle are Disclosed. These fiducial designs can provide five or six degree of freedom poses with a single fiducial using efficient closed form solutions. A novel pose estimation method is used to select the appropriate solution from the multiple solutions obtained, given certain application assumptions, is also disclosed.
1. A method of pose estimation for a plurality of regions based on a multiple fiducial pattern having a grid of main fiducials, the method comprising:
calculating a plurality of parameters, using a processor for one or more 5D poses of each region to generate a plurality of 5D poses by
calculating a pose normal for each 5D pose; and
calculating a pose center for each 5D pose;
grouping the 5D poses into one or more planar groups by
creating a set of planar groups wherein each planar group has a plane normal and a plane point; and
assigning each 5D pose to a planar group based on the pose normal, the plane normal, the pose center and the plane point;
selecting the planar group with the most members as the plane of the multiple fiducial pattern and
adding a new planar group to the set of planar groups when a current 5D pose does not fit any planar group of the set of planar groups.
2. The method of claim 1 , wherein the plane normal of the new planar group is set equal to the pose normal of the current 5D pose.
3. The method of claim 1 wherein the plane point of the new planar group is set equal to the pose center of the current 5D pose.
4. The method of claim 1 , further comprising using identifying features to match one or more regions of the plurality of regions to fiducials in the multiple fiducial pattern.
5. The method of claim 4 , wherein the identifying features provide orientation information.
6. A method of pose estimation for a plurality of regions based on a multiple fiducial pattern having a grid of main fiducials, the method comprising:
calculating a plurality of parameters, using a processor, for one or more 5D poses of each region to generate a plurality of 5D poses;
grouping the 5D poses into one or more planar groups;
selecting the planar group with the most members as the plane of the multiple fiducial pattern;
calculating a center difference between each pair of 5D poses to form a plurality of center differences; and
orienting the multiple fiducial pattern based on the 5D poses with a center distance that most closely equals a grid distance of the multiple fiducial pattern by
assigning the center difference whose length most closely equals a grid distance as a first direction of the multiple fiducial pattern; and
assigning a cross product of the first direction with a plane normal as a second direction of the multiple fiducial pattern.
7. A system of pose estimation for a plurality of regions based on a multiple fiducial pattern having a grid of main fiducials, the system comprising:
means for calculating a plurality of parameters for one or more 5D poses of each region to generate a plurality of 5D poses having:
means for calculating a pose normal for each 5D pose; and
means for calculating a pose center for each 5D pose;
means for grouping the 5D poses into one or more planar groups having:
means for creating a set of planar groups wherein each planar group has a plane normal and a plane point; and
means for assigning each 5D pose to a planar group based on the pose normal, the plane normal, the pose center and the plane point;
means for selecting the planar group with the most members as the plane of the multiple fiducial pattern; and
means for adding a new planar group to the set of planar groups when a current 5D pose does not fit any planar group of the set of planar groups.
8. The system of claim 7 , wherein the plane normal of the new planar group is set equal to the pose normal of the current 5D pose.
9. The system of claim 7 , wherein the plane point of the new planar group is set equal to the pose center of the current 5D pose.
10. The system of claim 7 , further comprising means for using identifying features to match one or more regions of the plurality of regions to fiducials in the multiple fiducial pattern.
11. The system of claim 10 , wherein the means for identifying features provide orientation information.
12. A system of pose estimation for a plurality of regions based on a multiple fiducial pattern having a grid of main fiducials, the system comprising:
means for calculating a plurality of parameters for one or more 5D poses of each region to generate a plurality of 5D poses;
means for grouping the 5D poses into one or more planar groups;
means for selecting the planar group with the most members as the plane of the multiple fiducial pattern;
means for calculating a center difference between each pair of 5D poses to form a plurality of center differences; and
means for orienting the multiple fiducial pattern based on the 5D poses with a center distance that most closely equals a grid distance of the multiple fiducial pattern having
means for assigning the center difference whose length most closely equals a grid distance as a first direction of the multiple fiducial pattern; and
means for assigning a cross product of the first direction with a plane normal as a second direction of the multiple fiducial pattern.