|
How to Unwarp Image
JHao from United States [10 posts] |
7 year
|
I want to unwarp my images with "Intrinsic and Extrinsic Parameters", which are obtained from "Automatic Image Calibration" with a known checkerboard image. I plugged in the parameters to "Transform Image" and "Radial Distortion" modules but without luck.
In the previous post, "http://www.roborealm.com/forum/index.php?thread_id=5843#", it seems to me that you suggested to unwarp the image using corner points. Could you provide an example to show how this works.
Thanks
|
|
|
|
Steven Gentner from United States [1446 posts] |
7 year
|
JHao,
Can you post the image that you are working with?
STeven.
|
|
|
|
JHao from United States [10 posts] |
7 year
|
Hi STeve,
Attached are checkerboard image in our application and the script to get Intrinsic and Extrinsic parameters from the checkerboard.
Our objective is to develop a script to unwarp the images taken by the same camera used for the checkerboard. To verify whether the script works or not, the first step is to see if the script can transform the original checkerboard image correctly. An example would help me greatly.
Thanks
|
|
|
|
JHao from United States [10 posts] |
7 year
|
Here are the checkerboard image and script to get the parameters.
 program.robo
|
|
|
|
Steven Gentner from United States [1446 posts] |
7 year
|
JHao,
Its possible that the slightly older version you are using doesn't have the new updates to this module. Just adding the module and pressing the Calibrate button (without the rotation module as this is redundant (the calibration will take care of that) I get the attached image (but using a square size of 20 pixels just to shrink the image).
I assume you are not getting this result? There isn't really an example since you just add the module and press the calibrate button. From then on, any image will undergo the same warping to correct for lens distortion. This calibration information is kept in the module and saved when a .robo file is saved. If you want to use this calibration in a different robofile and don't want to update the module you can export the calibration info as a file and reimport it in the other pipeline.
You should be able to run the attached robofile and see that good calibration assuming you upgrade to the latest version of RR in case that update was not in yours.
Let me know if you are able to reproduce these results.
STeven.
program.robo
|
|
|
|
Anonymous |
7 year
|
Thanks, STeven.
Two follow up questions. The first is U0=0 and V0=0 after I clicked "Calibrate" button. Could you explain this? The other is the Radio Button to select "Radian" or "Degree" in "Extrinsic Parameters". When I select "Degree", the "Radian" is still selected.
Thanks
|
|
|
|
Steven Gentner from United States [1446 posts] |
7 year
|
U0 and V0 will be the offset to center of the image relative to its lens. If the CCD sensor and the lens are perfectly aligned those will be zero. So that's good!
Radian and Degree has a small error. If you select one and press Calibrate it will update the values in that particular mode. This GUI error will be updated in the next version. Regardless, its easy to convert from one to the other ( a quick search of radian to degree will get it).
STeven.
|
|
|
|
JHao from United States [10 posts] |
7 year
|
Hi STeve,
I am running the Auto_Image_Calibration module today. The good news is that the issue of Degree and Radian is fixed. However, the parameters produced by Auto_Image_Calibration are different from before. If I click "Show Detected Grid (Purple)", the detected grids are off considerably. The version I am testing is "Commercial Ver. 2.83.17".
Thanks
|
|
|
|
Steven Gentner from United States [1446 posts] |
7 year
|
JHao,
Try upgrading to the latest version ... there were additional updates to that module since that version.
If that doesn't do it, can you post the raw calibration image here to allow us to test why that isn't working?
STeven.
|
|
|
|
JHao from United States [10 posts] |
7 year
|
Hi STeve,
The most recent upgrading doesn't fix the issue. The calibration image is same as I posted before. Please see attachment.
Thanks,
|
|
|
|
Steven Gentner from United States [1446 posts] |
7 year
|
JHao,
There was an issue in the display logic of those points that wasn't taking into account missing grid marks (it can happen once in a while). This did not effect the calibration but just the graphical drawing of those connecting lines. If you download the latest version that will have been corrected.
thanks!
STeven.
|
|
This Thread
