Blinds Feeder will not calibrate

[David Griffiths]

I have relocated my two blinds feeder arrays slightly and now I cannot get either of them to calibrate.  It fails with an error message that says 'Blinds feeder 1 S1 - fiducial not found'. Mainly it fails on the first fid but sometimes gets onto the second fid. I am seeing the diamond displayed each time.

The log entry looks like this:

2021-10-09 16:35:12.353 BlindsFeeder$FindFeatures DEBUG: dismissed fiducal candidate: result { {965.2501220703125, 554.7500610351562} 33x34 * 45.0 }, mmSize (1.394384, 1.427273, -0.000000, 0.000000 mm)

I am guessing it thinks the fiducial is too small (on a friend's machine he is seeing dimensions around 1.7mm), What size is it expecting? Is this configurable? (I have looked in machine.xml and found fid-loc-max-passes="3" fid-loc-tolerance-mm="0.5" but nothing relating to fid-size)

I have checked my pixels per mm.

This was working OK for a couple of weeks - not sure what has changed apart from relocating the arrays. Perhaps my hole size was always marginal?  I don't want to take the knife to them yet though.

[ma...@makr.zone]

Hi David

The fiducial should be 2mm and it accepts down to 1.4mm, up to 2.3mm, both limits are hard-coded. As you see in your result, it is borderline, i.e. it will work/not work unpredictably.

I suspect it's actually good to not be recognized, you may have slightly color-tinted shadows inside the fiducial that are not excluded by MaskHSV (the light is shining through the green PETG and tinting the shadows). Shadows must not be recognized as part of the fiducial, they will throw off accuracy, as they are not necessarily symmetrical.

You are on the testing version, right? The following is only available on testing:

First, if not yet done, switch off USB Automatic White Balance and apply the OpenPnP White Balance? Why? and How? See here:

https://github.com/openpnp/openpnp/wiki/Camera-White-Balance

It is the basis for stable green-screening.

Maybe you can then tune the MaskHSV yourself, to exclude the shadows in the fiducial. When the background of your fiducial is very reflective, you might have to color the inside of the fiducials with a complementary color (red/blue/magenta) or black. Just use a fine permanent felt marker.

If you need help with MaskHSV, please do the following:

1. Place the camera over the fiducial, as the calibration would.
2. Double click the camera view, to save a snapshot
   
3. Send the snapshot from .openpnp2/snapshots to me
   
4. Edit your pipeline, copy & paste and send to me

_Mark

--
You received this message because you are subscribed to the Google Groups "OpenPnP" group.
To unsubscribe from this group and stop receiving emails from it, send an email to openpnp+u...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/openpnp/a0527542-c208-4165-a264-83abda025ac9n%40googlegroups.com.

[Virender Kumar]

Can we use Black color PETG for blinds feeder

[Virender Kumar]

Here is my snapshot

[mark maker]

No.

OpenPnP must be able to discern the blinds covers against parts in their pockets, but those parts may be black too:

https://youtu.be/dGde59Iv6eY?t=393

It needs a unique vivid (=high saturation) color, i.e. it can be red or blue etc., but not black or grey or white. It must be from the upper and outer edge of that color cylinder:

https://en.wikipedia.org/wiki/HSL_and_HSV#Basic_principle

You can then adjust the MaskHsv stage in the pipeline to key the right color.

_Mark

To view this discussion on the web visit https://groups.google.com/d/msgid/openpnp/976ea4d9-e682-4c4a-8d75-a63635e9d0bfn%40googlegroups.com.

[Virender Kumar]

Dear Mark,
i have used green PETG still facing issues. during calibrate fiducials.
I have attached mine snapshot. Please help me out

[mark maker]

Dear Virender,

I see several issues.

• Lighting should be symmetrical. You seem to have a side light and this will continue to give you headaches. Use a LED ring and diffuser around the camera.
• Your camera seem very wide angle. I recommend a longer lens and mouting the camera higher up if at all possible.
  
• Make sure to follow Issues & Solutions to calibrate your camera. I see massive lens distortion.
  
• Make sure to follow Issues & Solutions to calibrated Camera White Balance.
• Your 3D-Print quality might not be tuned enough. The fiducial does not look like a diamond shape at all. There might be settings that affect this, like "outer perimeter first" etc. (I'm not much of an expert, the PRUSA printers I used just always worked perfectly, once I figured out to use 3D-printing spray for adhesion)
  

_Mark

To view this discussion on the web visit https://groups.google.com/d/msgid/openpnp/510daf53-a5fb-46e7-ab34-462851f6074an%40googlegroups.com.

[Virender Kumar]

Dear mark,
I have sorted out 3D print issue.
Light is in the center now.
Camera is a bit far away from the previous one.
Here is my pipeline

<cv-pipeline>
   <stages>
      <cv-stage class="org.openpnp.vision.pipeline.stages.ImageRead" name="00" enabled="false" file="test.png" color-space="Bgr" handle-as-captured="false"/>
      <cv-stage class="org.openpnp.vision.pipeline.stages.ImageCapture" name="0" enabled="true" default-light="true" settle-option="Settle" count="1"/>
      <cv-stage class="org.openpnp.vision.pipeline.stages.BlurGaussian" name="1" enabled="true" kernel-size="3" property-name="BlurGaussian"/>
      <cv-stage class="org.openpnp.vision.pipeline.stages.AffineWarp" name="ocr1" enabled="true" length-unit="Millimeters" x-0="0.0" y-0="0.0" x-1="0.0" y-1="0.0" x-2="0.0" y-2="0.0" scale="1.0" rectify="true" region-of-interest-property="regionOfInterest"/>
      <cv-stage class="org.openpnp.vision.pipeline.stages.ConvertColor" name="ocr2" enabled="false" conversion="Bgr2Gray"/>
      <cv-stage class="org.openpnp.vision.pipeline.stages.SimpleOcr" name="OCR" enabled="true" alphabet="0123456789.-+_RCLDQYXJIVAFH%GMKkmuµnp" font-name="Liberation Mono" font-size-pt="7.0" font-max-pixel-size="28" auto-detect-size="false" threshold="0.75" draw-style="OverScaledImage" debug="false" property-name="SimpleOcr"/>
      <cv-stage class="org.openpnp.vision.pipeline.stages.ImageRecall" name="11" enabled="true" image-stage-name="1"/>
      <cv-stage class="org.openpnp.vision.pipeline.stages.ConvertColor" name="2" enabled="true" conversion="Bgr2HsvFull"/>
      <cv-stage class="org.openpnp.vision.pipeline.stages.Normalize" name="3" enabled="false"/>
      <cv-stage class="org.openpnp.vision.pipeline.stages.MaskHsv" name="4" enabled="true" auto="false" fraction-to-mask="0.0" hue-min="65" hue-max="115" saturation-min="50" saturation-max="255" value-min="40" value-max="255" soft-edge="0" soft-factor="1.0" invert="false" binary-mask="true" property-name="MaskHsv"/>
      <cv-stage class="org.openpnp.vision.pipeline.stages.BlurMedian" name="5" enabled="true" kernel-size="13"/>
      <cv-stage class="org.openpnp.vision.pipeline.stages.FindContours" name="7" enabled="true" retrieval-mode="List" approximation-method="Simple"/>
      <cv-stage class="org.openpnp.vision.pipeline.stages.FilterContours" name="8" enabled="true" contours-stage-name="7" min-area="1000.0" max-area="100000.0" property-name="FilterContours"/>
      <cv-stage class="org.openpnp.vision.pipeline.stages.MinAreaRectContours" name="results" enabled="true" contours-stage-name="8"/>
      <cv-stage class="org.openpnp.vision.pipeline.stages.ImageRecall" name="10" enabled="true" image-stage-name="0"/>
      <cv-stage class="org.openpnp.vision.pipeline.stages.DrawRotatedRects" name="12" enabled="false" rotated-rects-stage-name="results" thickness="3" draw-rect-center="true" rect-center-radius="20" show-orientation="false"/>
   </stages>
</cv-pipeline>

[mark maker]

I suspect a scale issue, i.e. Units per Pixel.

You should still use Issues & Solutions to calibrate the camera. And White Balance. If all these things are okay, you should be able to use the stock pipeline unchanged.

I wrote down the steps in the Wiki:

https://github.com/openpnp/openpnp/wiki/BlindsFeeder#openpnp-preparation

If it does still not work, send the native image and a log at TRACE level.

https://github.com/openpnp/openpnp/wiki/FAQ#how-can-i-get-a-native-camera-image

https://github.com/openpnp/openpnp/wiki/FAQ#where-are-configuration-and-log-files-located

_Mark

To view this discussion on the web visit https://groups.google.com/d/msgid/openpnp/349dc2cb-ba6a-4c40-b613-54ece04ca911n%40googlegroups.com.