Your experience in calibrating the Kinect?
Hi there,
I would be interested in your experience regarding the calibration of the kinect depth camera to the build-in RGB camera. I followed all the openni_launch tutorials for the intrinsic and extrinsic calibration of the kinect. The results of the intrinsic calibration seem to be promising, but those of the extrinsic cannot compete with the given registration of the openni drivers based on what I "see". Especially if I move an object within the scene (away from the point where the extrinsic calibration took place). So I would like to know if it is possible to do the extrinsic calibration with more than just one view/pose of the checkerboard using camera_pose_calibration? I could use the matlab calibration toolbox with a set of paired rgb/ir images to calibrate offline or do you happen to know an alternative approach?
If you turn on the built-in calibration (which, as I recall, is the default), you'll get pretty good registration performance. Do you have the built-in calibration turned on? (Or are you in a case where the built-in calibration isn't good enough?)
I'm surprised to hear that the built-ins aren't good enough; I've had good luck with them (of course, I don't know the details of your problem). If they really are being quite bad, it might be worth trying a different Kinect; maybe it got stepped on. (Not that I've ever done that, ahem.)
When doing VSLAM, the built-in is good enough with 5 cm of error. However, doing small and very fine structures such as figurines, 5 cm of error is bad. I need it to be down to millimeter accuracy.
@loam, I'd actually be curious to know if you've got multiple Kinects working with the camera_pose_calibration package or just one?