How to get 3D coordinates of reflective markers using two cameras?

[Klemen]

You forgot to attach the files?

Best regards,

K

[Anvitha]

sorry Klemen! I tried uploading the zipped files but the size is too large. Is there a way where I can email you the files?

[Klemen]

Hello,

try dividing the files into seperate posts.

Regarding your questions:

1. The position of the cameras depend on your application. There is always a tradeoff between the parameters. For exmple, small baseline is good for easier correspondence estimation, the triangulation angle is smaller, but the depth uncertainty is greater. On the other hand, bigger baseline distance generally needs larger triangulation angle, the depth accuracy is better, but the correspondence estimation is more difficult. Also, more occlusions are possible using this setup (no depth information can be extracted from certain areas).

2. Ideally, the cameras should be synchronized.

3. It is important how you are performing the calibration. For some idea, read "Calibrate Stereo Vision System.vi" in the examples folder. Also, take a look at this link http://forums.ni.com/t5/Machine-Vision/Stereo-library-2012-pointers/td-p/2171812/page/3. For some reason I don't understand it is recommended to use cameras with smaller resolutions (VGA).

4. Regarding the quality of the calibration, read the detailed help of the "IMAQ Learn Binocular Stereo Calibration" for start. For more thorough read, search for papers, books, etc...

Best regards,

K

[Klemen]

You can also take a look here:

https://decibel.ni.com/content/blogs/kl3m3n/2014/07/24/stereovision-in-labview-based-on-correspondin...

Best regards,

K

[Anvitha]

attached are the images acquired from left camera for calibration

[Anvitha]

Attached are the images acquired from right camera for calibration. I tried to do the calibration using the images that I uploaded. It is detecting the circles but it is not saving the calib file and also it is giving all '0' for quality info.

[Klemen]

Hello,

you forgot to specify the distance between the dots - so I can repeat the calibration!

But anyway, your calibration is insufficient. You are covering a very small portion of your image with the calibration grid. Also, not all dots are detected.

It is a bad calibration altogether.

If you take a look at the calibration output, you will see: Insufficient Data returns TRUE if any of the following conditions are met: the number of different projection planes is fewer than 5, the angle difference of the projection planes is less than 20 degrees, or the lens in use is a telecentric lens.

Take a look at the stereo calibration example in the Labview help folder -> "Compute depth image" (runs this and watch the calibration) and "Calibrate Stereo Vision System".

In the calibration software you are using, you can also specify different settings for the grid extraction, so that the number of detected dots is as high as possible.

Best regards,

K