Rmse value is 66

[AkhilaPerumalla123]

Hi, I captured images using 2 webcams of the same specifications simultaneously. rmse value returned when each camera is calibrated is around 0.11. When they are setreocalibrated, the rmse value returned is around 66.

Is it a good rmse value? if not how to achieve good rmse value?

[AkhilaPerumalla123]

Hi @TemugeB, Can you please help me with this?

[TemugeB]

Hi,

Sorry for the delay. An rmse of 66 is definitely very bad. Did you make sure the calibration settings file was updated correctly? Also, when doing stereo calibration step, it is very important that you keep your hand steady. Are the calibration patterns large enough in both camera views at the same time? If the calibration pattern is too small, you will run into problems.

[AkhilaPerumalla123]

I captured steady images. But, for some of the images the the direction is horizontal and for some of the images the direction is vertical.

[TemugeB]

This is happening because your calibration pattern has equal number of rows and columns. I suggest you print out a calibration pattern that has different number of columns and rows. Or you can simply skip the frames where this is happening. The "O" in images refer to the origin of the calibration points. So the "O"s and the directions should match in both images.

Additionally, your camera is introducing too much radial distortion. To fix this, you need to call openCV's undistort function or find a camera that does not introduce radial distortion. Here is OpenCV's tutorial on how to undistort: link. It is quite simple to use.

[AkhilaPerumalla123]

Hi @TemugeB,

By taking a calibration pattern with different numbers of rows and columns, I reduced it to 2.88.

Any suggestions to improve it further?

[AkhilaPerumalla123]

Hi @TemugeB,

Can you mention the specifications of the camera you used to capture the pictures and videos?

It will be really helpful.