Machine Vision System Calibration

The accuracy and repeatability of an inspection system for dimensioning (more so than classification) depends heavily on the match between the imaging conditions for which it was programmed and tested to the current imaging conditions.

 

A variety of factors affect this match:​

​

Vibration of Camera/Optics with respect to the Unit Under Inspection (UUI) (more so for rolling shutter cameras)

Position of the Camera/Optics/UUI and mounting fasteners

Quality of focus

Even aspects such as oil vapors , moisture, dust on the lens

Working distance

Skew/Tilt/Rotation

Level of illumination

Aperture (iris) openness

Sensor temperature and performance over time

​

With these many factors affecting the quality of the image and therefore the accuracy and repeatability, it becomes more important than say in a purely electrical or mechanical system alone to carefully:

 

-calibrate the conditions at the time of capturing samples for programming the system with the system itself setting boundaries that limit the repeatability and accuracy impact of non-ideal conditions to acceptable vales (usually 0.5 pixels or less)

-adopt a smart, simple and accurate strategy to quantify and qualify the imaging conditions in the production line or arena of operation with actionable advice on how to restore those conditions if disturbed

-If possible, integrate the calibration with the real time scene of imaging itself but if not, employ suitable standalone calibration targets that can be used periodically such as once every shift to benchmark and if needed restore the imaging conditions.

 

I2SeeBots approaches:

​

I2SeeBots uses 2 possible approaches:

​

-When field of view provides sufficient mechanical play to mount real-time calibration artifacts, then the fixture is designed so as to enable calibration and report actions to restore the conditions based on the fixture's visual artifacts themselves.

​

-When field of view does not provide the play to mount real time targets, targets such as the above are used to be imaged manually or through automation on a periodic basis. From different attributes on the features in the calibration targets (see pictures), the system will advise the operator on what to do, if any, to restore the factory imaging conditions that guarantee <0.5 pixel loss in repeatability from the speced out factory values.

 

Calibration API:

​

I2SeeBots provides a DLL or LabVIEW API  as well as CAD files for Calibration Templates or Targets that look like the planar surface in the pictures above to perform this kind of calibration. The API will allow:

​

Benchmarking of ideal imaging conditions

Reporting the actions that a user can undertake to restore the factory imaging conditions