Hi Team,
I was trying to run targetless calibration on a custom rosbag collected having velodyne VLP32 and zed2i stereo camera inputs. After running the preprocessing by specifying the input parameters for camera intrinsics (the one I got from the manufacturer), I got few intensity files for each rosbag. The outputs does not look very acceptable and I believe this can impack perfroamnce in hte later stages. Any suggestions as to why this might be happening?
the rgb image looks like:
I had manually cleaned the pointclouds by filtering points out side the FOV set by me. I tried visualizing hte pointclouds generated on rviz and its dense.
@ashwinjoseph95 , I believe you might have followed the data collection procedure mentioned in docs.
Basically,
The sensors should be stand still initially and then should be moved up and down.
Try using "Dynamic integrator" in preprocessing.
Check the initial guess automatic correspondences
If the correspondences looks wrong, then i would suggest to use the manual approach as the intensity image is not being used in manual approach.
Hi Team, I was trying to run targetless calibration on a custom rosbag collected having velodyne VLP32 and zed2i stereo camera inputs. After running the preprocessing by specifying the input parameters for camera intrinsics (the one I got from the manufacturer), I got few intensity files for each rosbag. The outputs does not look very acceptable and I believe this can impack perfroamnce in hte later stages. Any suggestions as to why this might be happening? the rgb image looks like:
I had manually cleaned the pointclouds by filtering points out side the FOV set by me. I tried visualizing hte pointclouds generated on rviz and its dense.