Background Calibration analyzes the empty nozzle tip and background in the bottom camera. The obtained baseline information is later used to remove the background in Bottom Vision. The Background Calibration is automatically (and repeatedly) performed during the Nozzle Tip Calibration.

Background Calibration reuses the images that are already captured by the Nozzle Tip Calibration, so no extra machine time and motion is needed to obtain them.

For Juki (or similar) nozzle tips, it analyzes the background images for the Key Color (Chroma Key). Any dominant, vivid color is automatically detected and a likely bounding box computed in the HSV Color Model. This step is optional.

For all nozzle types, a cutoff brightness is determined where the non-color-keyed parts of the background must all be darker. When we later detect the bright contacts of parts in bottom vision, they must all be significantly brighter. This gives us the minimal brightness cutoff threshold (typically applied in a Threshold stage).

The background calibration automatically ignores a round area in the calibration image, where the nozzle tip was detected. This will eliminate any shiny elements that are typically present on the tip (metallic needle, worn-off tip, etc.). OpenPnP assumes this center area to be always covered by the picked part, i.e. there is no need to include its color in the background calibration. Furthermore, any shiny spots in this area are not treated as a problem (see the Trouble Shooting section). The blotted-out area has a size of Min. Part Diameter - 2 x Max. Pick Tolerance (see the Nozzle Tip Configuration for bottom vision).

The resulting calibration is also validated, and if the detected background quality is not good, diagnostic guidance and visual feed-back is provided for users to remedy the situation (see section Trouble Shooting below).

If the Background Calibration is enabled, the following is controlled in the bottom vision pipeline:

1. The BlurGaussian kernelSize property is controlled by a new Minimum Detail Size field, where the user can enter a length (in system units) that indicates the smallest sensible size of a feature on a part (such as the dimension of a pad, pin, ball of a package). The GaussianBlur is configured accordingly (to half of that size), to suppress image noise, dirt, texture etc. that is not relevant for detection. This is also applied to the Background Calibration itself, to get the same MaskHSV input data as the pipeline.
2. The MaskCircle diameter property is always controlled by the already existing Max. Part Diameter field on the Nozzle tip (this is also true, when the background calibration is not enabled). Peripheral background content that might disrupt the vision operation is masked out. This is also applied to the Background Calibration itself, to get the same MaskHSV input data as the pipeline.
3. The MaskHSV hueMin, hueMax, saturationMin, saturationMax, valueMin and valueMax properties are all controlled by the Background Calibration bounding box.

First make sure your Nozzle Tip Calibration is enabled and working. You can't use the background calibration without the nozzle tip calibration.

If you intend to use Color Keying, be sure to have Camera White Balance set up first.

There is a new Background Calibration section on the Calibration tab:

Method controls the Background Calibration:

• None switches Background Calibration off. The pipeline is no longer controlled, i.e. the user can edit the properties in the Pipeline Editor.
• Brightness only determines the cutoff brightness of the nozzle tip and background, no key color is assumed to be present.
• BrightnessAndKeyColor determines both the nozzle tip key color, and the remaining background cutoff brightness.

Minimum Detail Size (in system units) configures the smallest valid detail of a part to be detected in bottom vision, such as the smallest dimension of a pad, pin, ball of a package. All smaller image details like image noise, textures, dirt etc. are considered irrelevant for detection. The Minimum Detail Size controls a blurring filter to suppress these artifacts.

Minimum and Maximum columns in the Hue (base color), Saturation, Value (Brightness) HSV color model indicate the calibrated bounding box of the key color (if enabled). These values will be computed by the calibration.

The HSV color model can be illustrated like this (Wikipedia):

Tolerance can be used for each channel, to expand the detected bounding box. This provides robustness against changing ambient lighting, shadows cast by the picked parts, etc. Note: practical tolerance values are yet to be determined in testing. The tolerance is applied to both sides of the channel bounds, except for Saturation Maximum, and Value Minimum which are both left unlimited in the bottom vision MaskHSV stage. OpenPnP assumes both darker and more vivid (greener) pixels are always part of the background, even if they were not detected, or cut off in the calibration.

Note, the Tolerance is (currently) not shown in the graphical HSV color indicator.

Once you have selected Method and Minimum Detail Size, you can test the calibration by pressing the Calibrate button. In production use the Background Calibration will be re-triggered together with the Nozzle Tip Calibration, as described in the Calibration Operating Principle section.

Calibrated background removal does only work, if your pipeline actually has a MaskHsv stage. This is the case with the OpenPnP stock pipeline. However, perhaps you modified the pipeline and removed it, so you may need to activate the stock pipeline to get the functionality.

The Background Calibration data is then used to automatically parametrize the bottom vision operation. If the Background Calibration Method is set (not None), the relevant stage properties in the pipeline are fully controlled, as described in the Calibration Operating Principle section.

If the background was successfully calibrated (i.e. no problems indicated), the knockout should now be perfect:

Otherwise see the Trouble Shooting section.

As the Nozzle Tip Calibration is typically renewed at least whenever the machine was homed, or even with each nozzle tip change, the calibration data should be adaptive to changing ambient light and other conditions. A manual recalibration can also always be triggered, if needed.

Furthermore, any (color) differences between nozzle tips are also implicitly handled. This is important, when packages are compatible with multiple such nozzle tips. The pipeline (with is part/package dependent, but not nozzle tip dependent), can therefore dynamically react to these differences, through the calibration control.

The calibration provides diagnostic feed-back regarding the quality of the background. A text describes the problem(s) and provides rough guidance as to how the problem can be remedied. Furthermore, the button Show Problems can be pressed to display the problematic image portions in the camera preview:

Credits: Nozzle tip images by Jan, with thanks!