CAMERA CALIBRATION METHOD USING A CALIBRATION TARGET

• US 20170287166A1
• Filed: 03/23/2017
• Published: 10/05/2017
• Est. Priority Date: 03/29/2016
• Status: Active Grant

• Current Assignee: Institut national d'optique
• Original Assignee: Institut national d'optique
• Inventors: CLAVEAU, Fabien, SAMSON, ric, CADORET, Yannick, PRVOST, Donald, ST-LAURENT, Louis, MIKHNEVICH, Maxim
• Granted Patent: US 9,965,870 B2
• Time in Patent Office: Days
• CPC Class Codes: G06T 2207/30208 Marker matrix
• CPC Class Codes: G06T 7/80 Analysis of captured images to determine intrinsic or extrinsic camera parameters, i.e. camera calibration
• CPC Class Codes: H04N 17/002 for television cameras

Abstract

Calibration methods use a calibration target for obtaining the intrinsic and extrinsic camera parameters of one or more cameras are. The methods can include acquiring, with each camera, a sequence of target images representing the calibration target in different target poses and at different acquisition times. The methods can include identifying reference images from the target images, and defining volume bins, angle bins and multi-camera bins into which the reference images are stored. The reference images can be used to determine the intrinsic and extrinsic parameters of the one or more cameras. In some implementations, the calibration methods can enable a user to monitor the progress of the calibration process, for example by providing an interactive calibration target including an input/output user interface to guide the user in real-time during the acquisition of the target images and/or sensors to provide positional information about the target poses.

Claims

1. A method for intrinsic calibration of a camera using a calibration target, the camera having a field of view covering a scene, the method comprising the steps of:

providing a plurality of target images acquired with the camera, each target image representing a view of the calibration target in a respective one of a plurality of target poses, each target pose corresponding to a certain location and orientation of the calibration target relative to the camera;

partitioning a volume of interest of the scene into a set of volume bins;

defining a set of angle bins, each angle bin encompassing a respective range of possible orientation values for the calibration target;

identifying, among the plurality of target images, reference images of the calibration target;

assigning each reference image to either or both of one of the volume bins and one of the angle bins based on the respective target pose corresponding to the reference image; and

obtaining intrinsic camera parameters based on the reference images.

2. The method of claim 1, wherein the providing step comprises the steps of:

acquiring, with the camera, the plurality of target images of the calibration target in the respective plurality of target poses; and

displacing the calibration target between acquiring successive ones of the plurality of target images.

3. The method of claim 2, further comprising, during the providing step, a step of measuring at least one of location information and orientation information about the calibration target with at least one sensor mounted on the calibration target, and wherein the assigning step comprises determining at least one of the location and the orientation of the calibration target corresponding to each reference image based on the at least one of the location information and orientation information measured by the at least one sensor.

4. The method of claim 2, further comprising a step of monitoring, in real-time, a current number of reference images assigned to each one of the volume bins and each one of the angle bins.

5. The method of claim 4, further comprising a step of communicating filling-level information related to the current numbers of reference images assigned to the volume and angle bins.

6. The method of claim 5, wherein the displacing step is performed based on the filling-level information.

7. The method of claim 5, wherein the communicating step comprises displaying the filling-level information on a visual display mounted on the calibration target.

8. The method of claim 2, wherein the providing, identifying and assigning steps are performed until, for each one of the volume bins and each one of the angle bins, the current number of assigned reference images reaches a respective predetermined threshold amount.

9. The method of claim 2, wherein the obtaining step comprises a step of determining a calibration error associated with the intrinsic camera parameters, and wherein the providing, identifying, assigning and obtaining steps are stopped if the calibration error gets lower than a predetermined error value.

10. The method of claim 2, wherein the providing, identifying and assigning steps are performed at least partly concurrently with one another.

11. The method of claim 2, wherein the obtaining step is performed iteratively and at least partly concurrently with the providing, identifying and assigning steps.

12. The method of claim 1, further comprising preliminary steps of:

acquiring a series of initial images of the calibration target at different initial target poses; and

obtaining initial estimates for the intrinsic camera parameters based on the series of initial images.

13. The method of claim 1, wherein, for at least one of the target images, the identifying step comprises:

searching the calibration target in the target image;

if the calibration target is found, evaluating at least one image quality parameter of the target image; and

if each image quality parameter meets a respective quality criterion, classifying the target image as one of the reference images.

14. The method of claim 13, wherein the searching step comprises looking for one or more fiducial features present on the calibration target.

15. The method of claim 1, wherein the assigning step comprises, for at least one of the reference images, determining the respective target pose based on a pixel-based location of the calibration target in the reference image.

16. A method for extrinsic calibration of a network of cameras using a calibration target, each camera of the network having a field of view that covers a portion of a scene and partially overlaps the field of view of at least another camera of the network, the method comprising the steps of:

providing, for each camera, a plurality of target images acquired with the camera, each target image representing a view of the calibration target in a respective one of a plurality of target poses and at a respective one of a plurality of acquisition times;

defining at least one multi-camera bin, each multi-camera bin being associated with two or more of the cameras of the network, the two or more cameras having partially overlapping fields of view;

identifying, for each camera, qualified target images among the plurality of target images;

for each multi-camera bin;

assigning sets of multi-camera reference images to the multi-camera bin, the multi-camera reference images in each set having a same acquisition time and consisting of one of the qualified target images acquired by each one of the two or more associated cameras; and

obtaining, for each one of the two or more associated cameras and based on the multi-camera reference images, estimates of extrinsic camera parameters in a bin reference frame associated with the multi-camera bin; and

if the at least one multi-camera bin consists of more than one multi-camera bin, obtaining, for each camera of the network and based on the estimates of the extrinsic camera parameters obtained for every camera of the network, calibrated values of the extrinsic camera parameters in a same global reference frame;

otherwise, using, for each camera of the network, the estimates of the extrinsic camera parameters as the calibrated values of the extrinsic camera parameters.

17. The method of claim 16, wherein the providing step is performed at least partly concurrently for all cameras of the network and comprises, for each camera, the steps of:

acquiring, with the camera, the plurality of target images of the calibration target; and

displacing the calibration target between acquiring successive ones of the plurality of target images.

18. The method of claim 17, further comprising, during the providing step, a step of measuring at least one of location information and orientation information about the calibration target with at least one sensor mounted on the calibration target, and wherein, for each multi-camera bin, the assigning step comprises assigning the sets of multi-camera reference images to the multi-camera bin based on the at least one of the location information and orientation information measured by the at least one sensor.

19. The method of claim 17, further comprising a step of monitoring, in real-time, a current number of sets of multi-camera reference images assigned to each multi-camera bin.

20. The method of claim 19, further comprising a step of communicating filling-level information related to the current number of sets of multi-camera reference images assigned to each multi-camera bin.

21. The method of claim 20, wherein the communicating step comprises displaying the filling-level information on a visual display mounted on the calibration target.

22. The method of claim 17, wherein, for each multi-camera bin, the providing, identifying and assigning steps are performed at least partly concurrently with one another.

23. The method of claim 16, further comprising the steps of:

assigning, for each multi-camera bin, sets of validation images to the multi-camera bin, the validation images in each set having a same acquisition time and consisting of one of the qualified target images acquired by each one of the two or more associated cameras, the validation images being different from the multi-camera reference images; and

validating the calibrated values of the extrinsic camera parameters based on the validation images.

24. The method of claim 16, wherein each multi-camera bin is a two-camera bin associated with two of the cameras of the network.

25. The method of claim 16, wherein the network of cameras includes more than two cameras.

26. The method of claim 16, further comprising, for each camera, a step of determining intrinsic camera parameters, comprising the steps of:

partitioning a volume of interest of the scene into a set of volume bins;

defining a set of angle bins, each angle bin encompassing a respective range of possible orientation values for the calibration target;

identifying, among the qualified target images, single-camera reference images of the calibration target;

assigning each single-camera reference image to either or both of one of the volume bins and one of the angle bins based on the respective target pose corresponding to the single-camera reference image; and

obtaining the intrinsic camera parameters based on the single-camera reference images.