|
This is the sample implementation of the region based absolute pose estimation algorithm for central cameras, described in the following paper:
- Robert Frohlich, Levente Tamas and Zoltan Kato. Absolute Pose Estimation of Central Cameras Using Planar Regions. IEEE Transactions on Pattern Analysis and Machine Intelligence, ISSN 0162-8828, ESSN: 1939-3539, July 2019
Download
The code can be downloaded here.
The code has been written in Matlab and is released under the GNU General Public License.
|
Usage Notes
Use the pami_run_registration_persp.m script found in the src folder to run the algorithm for a perspective test data.
Use the pami_run_registration_omni.m script to run the algorithm with an omnidirectional test data.
Running the algorithm also requires the intrinsic camera parameters for the used central camera. In case of a perspective camera this is stored in the Kc variable, while for the omnidirectional camera it is the polinomial coefficients stored in the omni_g variable. For more details about the camera model used, please refer to the article [1].
|
Data
The test data contains the triangulated 3D template regions and their corresponding 2D observation image as a binary image. For convenience, the region used for the initial registration relying on a single region is stored separately with the _init tag in its name, while the full template having 3 regions and the full observation image are stored with Final tags in their names.
Results
The results provided by the algorithm contain the estimated pose, expressed as a vector of 6 elements, namely the rotation angles around the X,Y,Z axes in degrees, and the translation along X,Y,Z axes, and the bakcprojection error expressed on the 2D region by the delta error measure. The visualization of this error measure is also provided as a binary image, depicting only the non-overlapping pixels of the 2D regions and the projected 3D regions.
|