ethz-asl
Singular Matrix issue on camera validation
I have two Fisheye cameras with 180 Deg FOV. I was trying to do stereo calibration using omni-radtan model but encountered the following issue on validation. Any suggestions on how to overcome this ?
➜ kalibr_bags rosrun kalibr kalibr_camera_validator --cam omni-radtan-test-camchain.yaml --target april_6x6_80x80cm.yaml
Initializing calibration target:
Type: aprilgrid
Tags:
Rows: 6
Cols: 6
Size: 0.064 [m]
Spacing 0.0192 [m]
initializing camera geometry
Camera /stereo_inertial_publisher/left/image_raw:
Camera model: omni
Focal length: [1002.5692444857015, 1005.4935095836079]
Principal point: [637.6776275100682, 358.86173277919795]
Omni xi: 2.0443492843408433
Distortion model: radtan
Distortion coefficients: [-0.11604615769371034, 0.45292288562081956, -0.00037304896853662407, -0.0003430209815233101]
initializing camera geometry
Camera /stereo_inertial_publisher/right/image_raw:
Camera model: omni
Focal length: [991.1462040077117, 993.9745275717806]
Principal point: [634.8044972035357, 359.9922330463282]
Omni xi: 2.034611648180097
Distortion model: radtan
Distortion coefficients: [-0.12403769230569024, 0.49585542981612835, -0.0009685380713238949, -0.0005033730597604979]
Traceback (most recent call last):
File "/home/sachin/Desktop/luxonis/kalibr_workspace/devel/lib/kalibr/kalibr_camera_validator", line 15, in <module>
exec(compile(fh.read(), python_script, 'exec'), context)
File "/home/sachin/Desktop/luxonis/kalibr_workspace/src/kalibr/aslam_offline_calibration/kalibr/python/kalibr_camera_validator", line 498, in <module>
chain_validator = CameraChainValidator(camchain, targetConfig)
File "/home/sachin/Desktop/luxonis/kalibr_workspace/src/kalibr/aslam_offline_calibration/kalibr/python/kalibr_camera_validator", line 102, in __init__
edge["A"][cidx_dest] = self.prepareStereoRectificationMaps(cidx_src, cidx_dest)
File "/home/sachin/Desktop/luxonis/kalibr_workspace/src/kalibr/aslam_offline_calibration/kalibr/python/kalibr_camera_validator", line 330, in prepareStereoRectificationMaps
c1 = -np.linalg.inv(Poa[:,0:3]) * Poa[:,3]
File "<__array_function__ internals>", line 180, in inv
File "/home/sachin/.local/lib/python3.8/site-packages/numpy/linalg/linalg.py", line 552, in inv
ainv = _umath_linalg.inv(a, signature=signature, extobj=extobj)
File "/home/sachin/.local/lib/python3.8/site-packages/numpy/linalg/linalg.py", line 89, in _raise_linalgerror_singular
raise LinAlgError("Singular matrix")
numpy.linalg.LinAlgError: Singular matrix
➜ kalibr_bags
I don't think the validator supports omni directional cameras. It should only support pinhole ones since it relies on opencv for undistortion.
Does this issue help you https://github.com/ethz-asl/kalibr/issues/153?
Ah got it. Thanks for the reply @goldbattle. And sorry for the delay on follow-up. I was on vacation.
Does this issue help you https://github.com/ethz-asl/kalibr/issues/153?
Not really. It talks about how pinhole-radtan and pinhole-equi is supported. But I need omni-radtan undistort to check how well the undistort has happened here.
Do you have any suggestions on that?
The end goal for me here is to have a stereo setup using 2 cameras which are 180 Deg fov. While retaining atleast 140 deg FOV.
Can you try this package? Sorry that I have not added this link to the wiki yet. https://github.com/ethz-asl/image_undistort#image_undistort_node
Can you try this package? Sorry that I have not added this link to the wiki yet. https://github.com/ethz-asl/image_undistort#image_undistort_node
Yes. My mad. Forgot to reply. I tried that. it helped. needed some modifications. Thanks for all the help.