Optimize Calibration Distortion Model Using Multiple Images

Updated Jun 8, 2018

Reported In

Software

  • Vision Development Module

Driver

  • Vision Acquisition Software

Issue Details

I am using a large field of view (FOV) for my application and cannot print a full-sized calibration guide. What can I do to optimize my calibration results? 

Solution

You should use multiple calibration grid images to maximize the accuracy of your distortion model. Please see the following considerations when using multiple calibration images: 
  1. Coverage is key
Having ​as much coverage as possible of your FOV will increase the calibration accuracy. In general, if you want to use multiple images of your grid to have a better representation of your plane. The more points you have available to build your model, the more accurate it will be. If you have a very wide FOV, make sure that the combination of all your images cover your FOV as much as possible. 
 
  1. Create overlapping regions
Typically, you would want your grid the cover the whole image and have minimal black areas. Black areas will result in the highest error percentage and will increase the mean error. Therefore, you want as much overlap as possible in the images that you use. 
 
The ideal case would be to use complete grid images or as much of the grid as possible (40-50% of the points visible) in different angles or shifts with strong overlaps. 
 
  1. Multiple images are calculated separately, but their results are automatically integrated 
The Learn Distortion Model VI is internally applied to each image individually, but the results are put together in a way to take into account the location of the dots and their coverage over the whole FOV. For more information, see the Using Multiple Calibration Grid Images section in Calibration Concepts

Additional Information

For more information, please see the attachment about Prevision Metrology from the 2011 NI Week Vision Summit. 

WAS THIS ARTICLE HELPFUL?

Not Helpful