Camera pair calibration using non-standard calibration objects
First Claim
1. A method to calibrate cameras, comprising:
- obtaining a first image of an object from a first image capture unit and, concurrently, a second image of the object from a second image capture unit;
identifying a first set of landmark points based on the first image and a second set of landmark points based on the second image;
determining a first pose of the object based on the first set of landmark points and a putative structure of the object and a second pose of the object based on the second set of landmark points and the putative structure of the object;
determining an estimated structure of the object based on a model of the object and the first and second poses;
determining a projection error based on the first pose, the second pose and the estimated structure of the object;
evaluating whether the projection error meets a threshold value;
calibrating the first and second image capture units based on the estimated structure when the projection error meets the threshold value; and
repeating each of the determining actions and the evaluating action when the projection error does not meet the threshold value, wherein the putative structure of the object is replaced by the estimated structure of the object.
1 Assignment
0 Petitions
Accused Products
Abstract
Systems, methods, and computer readable media for calibrating two cameras (image capture units) using a non-standard, and initially unknown, calibration object are described. More particularly, an iterative approach to determine the structure and pose of an target object in an unconstrained environment are disclosed. The target object may be any of a number of predetermined objects such as a specific three dimensional (3D) shape, a specific type of animal (e.g., dogs), or the face of an arbitrary human. Virtually any object whose structure may be expressed in terms of a relatively low dimensional parametrized model may be used as a target object. The identified object (i.e., its pose and shape) may be used as input to a bundle adjustment operation resulting in camera calibration.
27 Citations
19 Claims
-
1. A method to calibrate cameras, comprising:
-
obtaining a first image of an object from a first image capture unit and, concurrently, a second image of the object from a second image capture unit; identifying a first set of landmark points based on the first image and a second set of landmark points based on the second image; determining a first pose of the object based on the first set of landmark points and a putative structure of the object and a second pose of the object based on the second set of landmark points and the putative structure of the object; determining an estimated structure of the object based on a model of the object and the first and second poses; determining a projection error based on the first pose, the second pose and the estimated structure of the object; evaluating whether the projection error meets a threshold value; calibrating the first and second image capture units based on the estimated structure when the projection error meets the threshold value; and repeating each of the determining actions and the evaluating action when the projection error does not meet the threshold value, wherein the putative structure of the object is replaced by the estimated structure of the object. - View Dependent Claims (2, 3, 4, 5, 6, 7)
-
-
8. An electronic device, comprising:
-
a first image capture unit; a second image capture unit; a memory communicatively coupled to the first and second image capture units; and one or more processors communicatively coupled to the first and second image capture units and the memory, the one or more processors configured to execute instructions stored in the memory to cause the electronic device to — obtain a first image of an object from the first image capture unit and, concurrently, a second image of the object from the second image capture unit; identify a first set of landmark points based on the first image and a second set of landmark points based on the second image; determine a first pose of the object based on the first set of landmark points and a putative structure of the object and a second pose of the object based on the second set of landmark points and the putative structure of the object; determine an estimated structure of the object based on a model of the object and the first and second poses; determine a projection error based on the first pose, the second pose and the estimated structure of the object; evaluate whether the projection error meets a threshold value; calibrate the first and second image capture units based on the estimated structure when the projection error meets the threshold value; and repeat each of the instructions to determine and the instructions to evaluate when the projection error does not meet the threshold value, wherein the putative structure of the object is replaced by the estimated structure of the object. - View Dependent Claims (9, 10, 11, 12, 13, 14)
-
-
15. A non-transitory program storage device, readable by one or more processors and comprising instructions stored thereon to cause the one or more processors to:
-
obtain a first image of an object from a first image capture unit and, concurrently, a second image of the object from a second image capture unit; identify a first set of landmark points based on the first image and a second set of landmark points based on the second image; determine a first pose of the object based on the first set of landmark points and a putative structure of the object and a second pose of the object based on the second set of landmark points and the putative structure of the object; determine an estimated structure of the object based on a model of the object and the first and second poses; determine a projection error based on the first pose, the second pose and the estimated structure of the object; evaluate whether the projection error meets a threshold value; calibrate the first and second image capture units based on the estimated structure when the projection error meets the threshold value; and repeat each of the instructions to determine and the instructions to evaluate when the projection error does not meet the threshold value, wherein the putative structure of the object is replaced by the estimated structure of the object. - View Dependent Claims (16, 17, 18, 19)
-
Specification