It is not unusual for there to be differences between the "intrinsic matrix" of cameras that determines their image, even of the same model. It is a consequence of the manufacturing process, apparently. So if the Razer is made in a different factory from Intel's model, a difference in image would be even more understandable.
RealSense expert Samontab reckoned that there was a time when even temperature could affect the image.
I don't know if the Calibration instruction would be of any use.
I did have a play with the Calibration interface but it just returns pre-set values, doesn't give you an option to actually calibrate it again.
I do know that you can use things like OpenCV to calibrate 3D cameras but a) I don't want to have to code a calibration thing myself (and integrate it with RSSDK to set the new calibration parameters etc), and b) I don't want the users to have to hold up a chessboard every time they need to calibrate. The R200 calibrator is good: easy to use, minimal effort, does the job. Just wondering if there are plans to make another one, or adapt the existing app, for the SR300.
Thanks for your reply Jose.
So if a camera is out of calibration and therefore useless for anything requiring accuracy, that's it, tough luck? The PXCMCalibration interface doesn't have any setters, so even if I did make my own calibration app, it doesn't look like there's any way for me to update the parameters on the device, so I'd have to roll my own Projection algorithms too! The R200 calibrator app works well, why not open that up to the other cameras too?
1 of 1 people found this helpful
You make a lot of sense! If I were in your position I would feel the same way, I will make sure to pass your suggestions to the right Department.
We would like to second the NEED for some calibration tools!
If Intel wants these sensors to be used in serious apps, then the community needs them.
At the moment is impossible to trust the calibrations for sensors.. They appear not to be calibrated at the factory and all have the same intrinsic parameters. This is true for Intel provided sensors as well.
You have no chance for repeatability without external calibration.
We are having to "roll our own" calibration process and mathematics to use the sensors for anything other than relative position detection!
PLEASE ADVOCATE for some tools.. It would really help with the adoption of the technology and make it a serious sensor out of the box.. not just a toy..
BTW.. Check out our nDepth product based on the SR300 here - nDepth 3D Scan - AikenControls
Hi Chris Aiken,
Thanks for your interest in the Intel® RealSense™ Platform.
We appreciate your feedback, the suggestions have already been passed to the respective department.
Please let us know if you have any more suggestions or questions.
Have a nice day!
Thanks! We appreciate your help!
Thanks for your interest in the Intel® Realsense™ Platform.
I did receive feedback from the Engineering Department. Unfortunately, there are no plans for generating a calibration tool for the SR300, we apologize for any inconvenience that this may cause.
Still, I will pass again the suggestion to the Engineering team, so that they know that there are several users in need of a calibration tool.
Is there any update on the calibration tool or method for getting intrinsics and extrinsics of the sr300 rgbd sensor? I have tried opencv calibration tools for instrinsic calibration of the rgb sensor. However the values of the parameters keep changing on iterating the calibration process by 10 pixels. I own 2 sr300 sensors. I see that both of them have different parameters. However still I don't see alignment in the raw rgb and projected depth in rgb frame.
I have tried to use the distortion model on the raw depth image and then used the factory parameters and my estimated intrinsics for alignment. However this doesn't work as I don't see any alignment.
Can somebody outline the algorithm using which they might have succeeded in calibrating the parameters of the camera?