CameraRigLocalization¶
Description
This node retrieves the transformation between the tracked and the optical reference frames.(?) https://alicevision.github.io/#photogrammetry/localization
settings
Name |
Description |
---|---|
SfM Data |
|
Media Path |
The path to the video file, the folder of the image sequence or a text file (one image path per line) for each camera of the rig (eg. –mediapath /path/to/cam1.mov /path/to/cam2.mov) |
Rig Calibration File |
The file containing the calibration data for the rig (subposes) |
Camera Intrinsics |
The intrinsics calibration file for each camera of the rig. (eg. –cameraIntrinsics /path/to/calib1.txt /path/to/calib2.txt) |
Descriptor Path |
Folder containing the .desc |
Match Describer Types |
The describer types to use for the matching (sift’, ‘sift*float’, ‘sift*upright’, ‘akaze’, ‘akaze*liop’, ‘akaze*mldb’, ‘cctag3’, ‘cctag4’, ‘sift*ocv’, ‘akaze*ocv’) |
Preset |
Preset for the feature extractor when localizing a new image (low, medium, normal, high, ultra) |
Resection Estimator |
The type of /sac framework to use for resection (acransac, loransac) |
Matching Estimator |
The type of /sac framework to use for matching (acransac, loransac) |
Refine Intrinsics |
Enable/Disable camera intrinsics refinement for each localized image |
Reprojection Error |
Maximum reprojection error (in pixels) allowed for resectioning. If set to 0 it lets the ACRansac select an optimal value (0 - 10) |
Use Localize Rig Naive |
Enable/Disable the naive method for rig localization: naive method tries to localize each camera separately |
Angular Threshold |
The maximum angular threshold in degrees between feature bearing vector and 3D point direction. Used only with the opengv method (0 - 10) |
Voctree |
[voctree] Filename for the vocabulary tree |
Voctree Weights |
[voctree] Filename for the vocabulary tree weights |
Algorithm |
[voctree] Algorithm type: {FirstBest, AllResults} |
Nb Image Match |
[voctree] Number of images to retrieve in the database |
Max Results |
[voctree] For algorithm AllResults, it stops the image matching when this number of matched images is reached. If 0 it is ignored (0 - 100) |
Matching Error |
[voctree] Maximum matching error (in pixels) allowed for image matching with geometric verification. If set to 0 it lets the ACRansac select an optimal value (0 - 10) |
N Nearest Key Frames |
[cctag] Number of images to retrieve in database (0 - 50) |
Output Alembic |
Filename for the SfMData export file (where camera poses will be stored) desc.Node.internalFolder + ‘trackedcameras.abc |