Systems and methods for initiating calibration of a multi-view camera pair deployed in a robot
View Patent ↗Systems and methods described herein relate to initiating calibration of a multi-view camera pair deployed in a robot. In one embodiment, a calibration initiation system in the robot monitors the reprojection error between cameras of the multi-view camera pair, the reprojection error measuring the calibration accuracy of the multi-view camera pair. The system also performs, when the reprojection error exceeds a predetermined threshold, one of (1) notifying a user that manual calibration of the multi-view camera pair is recommended and (2) initiating automatic calibration of the multi-view camera pair.
1 . A system, comprising:
a processor; and
a memory storing machine-readable instructions that, when executed by the processor, cause the processor to:
calculate a reprojection error as a combination of first and second reprojection errors between a first camera and a second camera of a multi-view camera pair deployed in a robot, wherein:
the first reprojection error is computed by converting a first image from the first camera to a first reprojected image reprojected from a three-dimensional (3D) point cloud transformed to a frame of reference of the second camera and comparing the first reprojected image with an image from the second camera that corresponds to the first image; and
the second reprojection error is computed by converting a second image from the second camera to a second reprojected image reprojected from a 3D point cloud transformed to a frame of reference of the first camera and comparing the second reprojected image with an image from the first camera that corresponds to the second image; and
perform, when the reprojection error exceeds a predetermined threshold, one of:
notifying a user that manual calibration of the multi-view camera pair is recommended; and
initiating automatic calibration of the multi-view camera pair.
2 . The system of claim 1 , wherein the multi-view camera pair is a stereo camera pair.
3 . The system of claim 1 , wherein the first image is converted to the 3D point cloud transformed to the frame of reference of the second camera and the second image is converted to the 3D point cloud transformed to the frame of reference of the first camera based on one of estimated and ground-truth depth information.
4 . The system of claim 3 , wherein the combination is an average.
5 . The system of claim 1 , wherein notifying the user that manual calibration of the multi-view camera pair is recommended includes transmitting a notification from the robot to a device associated with the user.
6 . The system of claim 1 , wherein the automatic calibration includes minimizing the reprojection error.
7 . The system of claim 1 , wherein the robot is one of an autonomous vehicle and a humanoid robot.
8 . A non-transitory computer-readable medium storing instructions that, when executed by a processor, cause the processor to:
calculate a reprojection error as a combination of first and second reprojection errors between a first camera and a second camera of a multi-view camera pair deployed in a robot, wherein:
the first reprojection error is computed by converting a first image from the first camera to a first reprojected image reprojected from a three-dimensional (3D) point cloud transformed to a frame of reference of the second camera and comparing the first reprojected image with an image from the second camera that corresponds to the first image; and
the second reprojection error is computed by converting a second image from the second camera to a second reprojected image reprojected from a 3D point cloud transformed to a frame of reference of the first camera and comparing the second reprojected image with an image from the first camera that corresponds to the second image; and
perform, when the reprojection error exceeds a predetermined threshold, one of:
notifying a user that manual calibration of the multi-view camera pair is recommended; and
initiating automatic calibration of the multi-view camera pair.
9 . The non-transitory computer-readable medium of claim 8 , wherein the multi-view camera pair is a stereo camera pair.
10 . The non-transitory computer-readable medium of claim 8 , wherein the first image is converted to the 3D point cloud transformed to the frame of reference of the second camera and the second image is converted to the 3D point cloud transformed to the frame of reference of the first camera based on one of estimated and ground-truth depth information.
11 . The non-transitory computer-readable medium of claim 8 , wherein notifying the user that manual calibration of the multi-view camera pair is recommended includes transmitting a notification from the robot to a device associated with the user.
12 . The non-transitory computer-readable medium of claim 8 , wherein the automatic calibration includes minimizing the reprojection error.
13 . The non-transitory computer-readable medium of claim 8 , wherein the robot is one of an autonomous vehicle and a humanoid robot.
14 . A method, comprising:
calculating a reprojection error as a combination of first and second reprojection errors between a first camera and a second camera of a multi-view camera pair deployed in a robot, wherein:
the first reprojection error is computed by converting a first image from the first camera to a first reprojected image reprojected from a three-dimensional (3D) point cloud transformed to a frame of reference of the second camera and comparing the first reprojected image with an image from the second camera that corresponds to the first image; and
the second reprojection error is computed by converting a second image from the second camera to a second reprojected image reprojected from a 3D point cloud transformed to a frame of reference of the first camera and comparing the second reprojected image with an image from the first camera that corresponds to the second image; and
performing, when the reprojection error exceeds a predetermined threshold, one of:
notifying a user that manual calibration of the multi-view camera pair is recommended; and
initiating automatic calibration of the multi-view camera pair.
15 . The method of claim 14 , wherein the multi-view camera pair is a stereo camera pair.
16 . The method of claim 14 , wherein the first image is converted to the 3D point cloud transformed to the frame of reference of the second camera and the second image is converted to the 3D point cloud transformed to the frame of reference of the first camera based on one of estimated and ground-truth depth information.
17 . The method of claim 16 , wherein the combination is an average.
18 . The method of claim 14 , wherein notifying the user that manual calibration of the multi-view camera pair is recommended includes transmitting a notification from the robot to a device associated with the user.
19 . The method of claim 14 , wherein the automatic calibration includes minimizing the reprojection error.
20 . The method of claim 14 , wherein the robot is one of an autonomous vehicle and a humanoid robot.